Compare commits
10
Commits
d09829bdb8
..
master
| Author | SHA1 | Date | |
|---|---|---|---|
  chenmoranzhiandCursor
|
7185f25dec | ||
|
chenmoranzhiandCursor
|
66ed1eaca8 | ||
|
chenmoranzhiandCursor
|
cf6cb68429 | ||
|
chenmoranzhiandCursor
|
ff5c77bfb1 | ||
|
chenmoranzhiandCursor
|
820397e578 | ||
|
chenmoranzhiandCursor
|
1cc4ae557a | ||
|
chenmoranzhiandCursor
|
d0a48f9431 | ||
|
chenmoranzhiandCursor
|
bb4f934b3d | ||
 fazhu
|
d0274afb80 | ||
 ociflab-bot
|
48801c656d |
@@ -1,84 +1,92 @@
|
||||
# Writing Agent 使用说明
|
||||
# Writing Agent
|
||||
|
||||
解压或下载完成后，按下面的顺序即可：双击启动、打开浏览器、写一张最简单的卡，然后游玩一句。
|
||||
本地运行的多 Agent 文本创作与游玩平台：先为一部作品搭好可复用的运行规格，再进入游玩——你说一句，它按规格写一轮。适合交互式角色扮演、短局/长文、仿写扩写等场景。
|
||||
|
||||
需要先安装 [Node.js](https://nodejs.org/)（勾选安装时附带的 npm）。第一次启动会自动安装依赖，请等黑窗口跑完。
|
||||
数据默认落在本机用户目录（`~/.writing-agent`），通过浏览器或 Electron 桌面窗口使用；模型走你自己配置的 API。
|
||||
|
||||
## 1. 双击 `start.bat`
|
||||
## 它能做什么
|
||||
|
||||
在文件夹里双击 `start.bat`。会弹出一个黑窗口，并自动打开浏览器。地址是：
|
||||
- **创作**：为新作品选进料方式（工序编排 / 对话落盘（默认选项））与配方，由编排器调度技能，落成可复用的运行规格与开场
|
||||
- **游玩**：在钉住的规格下回合推进；可切换呈现壳，保存游玩进度
|
||||
- **技能包**：能力以磁盘上的技能包装载（当前主包为「世界模拟器」相关配方与模块）（目前不需要手动选择）
|
||||
- **预设与库**：兼容 SillyTavern 风格的提示词预设；文风库 / 偏好库可注入创作节点，种子可从仓库恢复
|
||||
- **多接口**：分组管理 API、嗅探型号、切换同一接口下的模型
|
||||
- **本机优先**：Web（`localhost:23337`）或 `start-app.bat` 桌面壳；运维可用 `console.bat`
|
||||
|
||||
http://localhost:23337
|
||||
## 快速开始
|
||||
|
||||
窗口不要关。关窗口或按 `Ctrl+C` 就是停止软件。
|
||||
1. 安装 [Node.js](https://nodejs.org/)（含 npm）
|
||||
2. 双击 `start.bat`（桌面窗口则用 `start-app.bat`），浏览器打开 [http://localhost:23337](http://localhost:23337)
|
||||
3. 在 **设置** 里新增模型接口（Base URL、API Key、型号）；根目录已有 `.env` 时可跳过
|
||||
4. **新建作品** → 对话写出一张简单可玩卡 → **保存定稿** → 左侧菜单选 **游玩**
|
||||
|
||||
|
||||
完整图文步骤（含截图与示例提示词）见 **[从启动到开玩](docs/guide/getting-started.md)**。
|
||||
|
||||
想用桌面窗口而不是浏览器，改双击 `start-app.bat`。后面的操作一样。
|
||||
|
||||
## 2. 第一次要填模型接口
|
||||
|
||||
点右上角 **设置**，再点 **新增**。填上这四项后点 **保存并选用**：
|
||||
|
||||
- 名称：自己认得就行，例如 DeepSeek
|
||||
- Base URL：接口地址，例如 `https://api.deepseek.com/v1`
|
||||
- API Key：你的密钥
|
||||
- 型号：这一行要用的模型名，例如 `deepseek-chat`
|
||||
|
||||
填完型号后也可以点 **嗅探**，从列表里点一个型号。
|
||||
|
||||
|
||||
|
||||
项目根目录如果已经有填好的 `.env`，启动时会自动带上，这一步可以跳过。
|
||||
|
||||
## 3. 新建作品
|
||||
|
||||
还没有任何作品时，打开页面就会弹出 **新建作品**。输入名称，点 **创建**。以后再开新的，点左侧 **+**。
|
||||
|
||||
|
||||
|
||||
这里用「雨天小书店」做例子。
|
||||
|
||||
## 4. 写一张最简单的卡
|
||||
|
||||
在底部输入框里原样贴上这段，点发送（或按 Enter）：
|
||||
## 核心回路
|
||||
|
||||
```text
|
||||
请写一张最简单的可玩卡，写完就停。
|
||||
|
||||
题材：雨天小书店。玩家是进来躲雨的客人，店员是一只戴蓝色围巾的猫。
|
||||
只要短对话。不要数值、地图、复杂世界观。
|
||||
开场只要两三句：雨还在下，猫店员抬起头。
|
||||
新建作品 → 选进料方式与配方
|
||||
→ 创作（编排技能 / 对话落盘）→ 运行规格 + 开场
|
||||
→ 保存定稿
|
||||
→ 游玩（执行单元按规格上场）→ 保存游玩
|
||||
```
|
||||
|
||||
|
||||
术语以 `[docs/ui-glossary.md](docs/ui-glossary.md)` 为准（进料方式、配方、编排器、技能、运行规格、执行单元）。
|
||||
|
||||
等它写完。中间的回复会说明卡已经好了，右侧 **已持久化** 是落下来的几张卡片。
|
||||
## 仓库一览
|
||||
|
||||
|
||||
|
||||
写完它如果还追问，回复「都不用再加，写完开场就可以开玩」。
|
||||
| 路径 | 用途 |
|
||||
| ------------------- | ------------------------------ |
|
||||
| `src/` | 运行时、会话、技能装载、Book/预设/库、HTTP API |
|
||||
| `web/` | 前端静态页 |
|
||||
| `skills/` | 技能包（见 `skills/README.md`） |
|
||||
| `examples/presets/` | 示例预设 |
|
||||
| `seeds/libraries/` | 文风/偏好库种子 |
|
||||
| `electron/` | 桌面壳 |
|
||||
| `docs/` | 架构与契约文档 |
|
||||
| `tests/` | Vitest |
|
||||
|
||||
## 5. 保存并开玩
|
||||
|
||||
右上角出现 **保存定稿** 时，点它，给这份稿起个名字。没有这个按钮时，点输入框右侧的保存图标，效果相同。
|
||||
用户数据目录可用环境变量 `WRITING_AGENT_DATA_DIR` 覆盖；默认 `~/.writing-agent`（Windows 多为 `%USERPROFILE%\.writing-agent`）。
|
||||
|
||||
然后点左侧的作品名，在弹出的菜单里选 **游玩**。
|
||||
## 维护者
|
||||
|
||||
会先看到开场白：
|
||||
开发与排障常用命令：
|
||||
|
||||
|
||||
|
||||
在底部写你要做的事，再发送。例如：
|
||||
|
||||
```text
|
||||
我收起伞，问店员今天适合躲雨看的书。
|
||||
```bash
|
||||
npm install
|
||||
npm run web # 启动 HTTP 服务（不监视文件）
|
||||
npm run web:watch # 开发时监视 src/ 与 skills/
|
||||
npm test # Vitest
|
||||
npm run libraries:restore # 从 seeds 恢复库
|
||||
npm run libraries:export # 导出库到 seeds
|
||||
```
|
||||
|
||||
它会接下去写这一轮：
|
||||
Windows 运维脚本：`console.bat`（查询 / 启停 / 日志）。入口实现：`src/server/web-server.ts` → `session-manager.ts` → `src/runtime/` + `src/main-agent/` + `skills/`。
|
||||
|
||||
|
||||
**安全注意**：本地 HTTP 控制面默认无鉴权，且会监听本机端口；请勿把服务暴露到不可信网络。API Key 以明文保存在用户数据目录的配置里。
|
||||
|
||||
之后就一直这样：你说一句，它写一轮。右上角 **保存游玩** 可以把进度存下来。
|
||||
### 文档索引
|
||||
|
||||
想换你在故事里的名字，到 **设置 → 用户角色** 里新增一个角色并选用。创作时如果希望开场先写占位，把角色名写成 `@玩家`。
|
||||
|
||||
| 文档 | 内容 |
|
||||
| --------------------------------------------------------------- | -------------------- |
|
||||
| [architecture.md](docs/architecture.md) | 系统架构与模块边界 |
|
||||
| [ui-glossary.md](docs/ui-glossary.md) | 用户可见术语 |
|
||||
| [implementation-guide.md](docs/implementation-guide.md) | 文档地图与实现约定 |
|
||||
| [runtime-state-machine.md](docs/runtime-state-machine.md) | 相位机 |
|
||||
| [tool-contracts.md](docs/tool-contracts.md) | Tool 契约 |
|
||||
| [run-snapshot.md](docs/run-snapshot.md) | 存档 / 快照（以代码现状为准） |
|
||||
| [play-presentation-shells.md](docs/play-presentation-shells.md) | 游玩呈现壳 |
|
||||
| [preset-format.md](docs/preset-format.md) | 预设格式 |
|
||||
| [skill-format.md](docs/skill-format.md) 等 | 技能 / 编排器 / Worker 格式 |
|
||||
| [px-roadmap.md](docs/px-roadmap.md) | 产品路线图 |
|
||||
| [skills/README.md](skills/README.md) | 技能包目录说明 |
|
||||
|
||||
|
||||
`docs/book-storage.md` 描述的是目标存储模型，**不是**当前磁盘布局；当前以 Book JSON + session + run-snapshots 为准。
|
||||
|
||||
## 许可与第三方
|
||||
|
||||
见 `[THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md)`。项目版本见 `package.json`（当前 `0.1.0`，`private`）。
|
||||
@@ -0,0 +1,84 @@
|
||||
# 创作轮次与前缀花费
|
||||
|
||||
> 状态：结论已记录。缓存粘滞已接入：本局 `sessionId` 在 OpenRouter 请求体写成 `session_id`。不手选节点导致的「先选再写」尚无做法。
|
||||
> 记录：2026-10-01。用量取自本地 `~/.writing-agent/stats`，2026-09-20 起。
|
||||
> 相关：`src/dictate/prompt.ts`、`src/dictate/turn.ts`、`src/dictate/read-module.ts`、`src/llm/client.ts`。
|
||||
|
||||
## 已记录的花费
|
||||
|
||||
9 月 20 日以来，OpenRouter 上的 token 主要在创作落盘的输入，不在游玩输出。
|
||||
|
||||
| 路径 | 请求 | 会话约 | prompt | 其中缓存 | 未缓存 | 输出 |
|
||||
|---|---:|---:|---:|---:|---:|---:|
|
||||
| 创作落盘 `dictate_agent` | 416 | 5 | 1198 万 | 782 万 | 417 万 | 38 万 |
|
||||
| 游玩 `worker:world-simulator` | 28 | 2 | 77 万 | 0 | 77 万 | 7 万 |
|
||||
|
||||
输出大约是 prompt 的 3%–9%。输出单价更高，仍然要算，但排在输入后面。缓存命中仍按折扣价计，缩体积会同时降低全价和折扣价。
|
||||
|
||||
每轮花费可以看成：
|
||||
|
||||
```text
|
||||
请求次数 ×（未缓存输入 × 全价 + 缓存输入 × 折扣价 + 输出 × 输出价）
|
||||
```
|
||||
|
||||
创作这 416 次是最大的一块。大约五个会话，一句用户话后面跟着一串读和写。
|
||||
|
||||
## 缓存粘滞：必做
|
||||
|
||||
OpenRouter 请求体现由统计层带上本局 `sessionId`，字段名 `session_id`（最长 256 字）。直连其它渠道不写这个字段。有 `session_id` 后，粘滞从该会话第一次成功请求就开始，不再等缓存命中，也不再因为第一条 user 被重写而换上游。`provider.order` 仍不发送，它会关掉这层粘滞。
|
||||
|
||||
同一条前缀，落点稳定时能命中，换上游就是 0：
|
||||
|
||||
- 创作工具循环间隔大约 7–15 秒。Gemini 上同一轮里缓存从约 3.8k 爬到 7.6k、11.5k、15k，整段 prompt 大约 50%–68% 命中。burst 的第一下经常是 0。中间偶尔整段掉回 0，是换了上游。
|
||||
- 游玩 world-simulator 在 OpenRouter 上 28 次全是 0。gemini-3.7 间隔中位数约 5.5 分钟，gemini-3.8 约 22 分钟，prompt 1.8 万–3.4 万。
|
||||
- 直连 glm-5.3 的游玩，间隔 8–30 分钟仍常见 80%–91%。某一轮从约 11k 掉到 6656 并停住，说明前缀中段改过，只保住开头约 6.6k。
|
||||
|
||||
因此缓存粘滞是必做项。不传稳定的 `session_id`，OpenRouter 每一轮按新开头重新选上游，游玩前缀缓存等于没有。`provider.order` 会关掉这层粘滞，不能拿来代替 `session_id`。
|
||||
|
||||
上下文排序对游玩前缀是有效的：人设和静态产物在前，历史居中，变量和本轮输入在后。它只在同一上游上表现得出来。
|
||||
|
||||
## 未决：不手选节点，选和写至少两轮
|
||||
|
||||
用户不手动指定本轮写哪个能力。系统提示里只有能力目录和短说明（何时用、何时不用、档位）。`prompt.md` 里的任务、原则、产物规格、评分不在这条提示里。
|
||||
|
||||
模型选定能力后调用 `read_module`。程序把这些段落放进工具结果，并在 `运行.对话落盘.已读能力` 记下该能力当前版本。版本是这些段落的哈希。随后的 `insert` 要核对这个账：没读过，或磁盘上的提示已经改过，就拒绝写入。`self_score` 也按这次读到的评分维度检查。`narrative`、`variable-context`、`context-order`、`worker-spec`、`refine` 以及 `auto` 能力不走这道门，其余内容能力都走。库条目同理：目录只是摘要，正文要等 `read_library_entry` 回来。
|
||||
|
||||
读的那一轮，完整写法还不在上下文里，模型给不出合格正文。同一项能力至少是：
|
||||
|
||||
```text
|
||||
第 1 次请求：凭短说明选定，并 read_module
|
||||
第 2 次请求：看见任务 / 原则 / 规格 / 评分之后 insert
|
||||
```
|
||||
|
||||
前缀发送两遍。第二遍还多带上那份写法。手选节点时，程序可以把该节点的完整写法放进唯一一次请求，模型当场写完。不手选时，第一次请求不能把所有节点的完整提示都放进去，这第二次省不掉。
|
||||
|
||||
一次只返回一个工具时，倍数再乘上去。提示要求一轮通常写 2～3 个已经说清的能力，模型常做成「读甲、写甲、读乙、写乙」，最后再单独回用户一句。每一拍都把已经变长的前缀再送一遍。「选和写拆成两次」是翻倍的原因；「每次只动一个节点」把两倍变成好几倍。
|
||||
|
||||
驱动程序已经会在同一步执行模型这一次返回的全部工具。接口没有禁止一轮多个 `insert`。日志里偶尔同轮两次写入。多数时候模型一次只吐一个工具。
|
||||
|
||||
结构化输出可以把一次生成拆成多份产物：一份 JSON 里多条 `{ position, content }`，程序再落盘，用户可见回复也可以放在同一份里。这省的是写入侧重复发送前缀。规格说明仍须在生成这批产物之前到达模型；否则模型是在没看见写法的情况下填内容。读和写如何合成更少的请求，本文不规定做法。
|
||||
|
||||
## 一本书上的拆账
|
||||
|
||||
本机作品 `4c13b2b4` 的当前会话（25 份产物、22 轮用户话）按现在的 `buildDictateMessages` 拼出来，不发请求。字数大约按 2 字 1 token 对照：同书末次 OpenRouter 请求 prompt 为 46897 token。
|
||||
|
||||
| 块 | 字数 | 占当前 messages |
|
||||
|---|---:|---:|
|
||||
| 已有产物合计 | 60518 | 69% |
|
||||
| 全量对话 | 19975 | 23% |
|
||||
| 系统提示（含方案进度 4392） | 6695 | 8% |
|
||||
| 工具定义 JSON | 4917 | 另计，约等于系统提示 |
|
||||
| 当前选用预设里启用的正文 | 362 | 可忽略 |
|
||||
|
||||
产物没有一份独自占满。最大的是 `设计.正文组成` 8879 字，其次变量规则 6005、实现机制 4569。其余多为两三千字，加在一起才是大头。会话中段的检查点（第 119 条消息之前）产物已经是 56102 字，和结尾同一量级。
|
||||
|
||||
按前面两段收窄再量同一份材料：
|
||||
|
||||
| 方案 | 字数 |
|
||||
|---|---:|
|
||||
| 选择：方案进度 + 用户最后一句（130 字） | 4522 |
|
||||
| 填写若仍带全部产物 + 最近 4 条对话（1770 字） | 约 6.2 万 |
|
||||
| 再注入 1 份节点提示（正文约 5 千字，叙事指南 5557） | 约 6.7 万 |
|
||||
| 现在这一次请求的 messages | 88218 |
|
||||
|
||||
选择请求很小，目录也稳定，缓存几乎打在整段上。填写请求的大头仍是已有产物，不是对话，也不是刚插入的节点提示。砍掉全量对话大约省两成，改不了产物那七成。节点提示每份约 5 千字，一次写两三份会在产物之上再加一两成，但只发送一次。
|
||||
Binary file not shown.
@@ -288,7 +288,7 @@
|
||||
|
||||
写开场或结束创作前，用仍未落的「必须」和「有必要」问用户哪些还没设计。同轮写入多个已说清产物时综合回应整轮，不逐模块重复模板。
|
||||
|
||||
写正文组成或 0 层正文组成时，回复里直接给出一份隔离内容供预览：与本局前端同一套骨架，示例已经写进去。完整 JSON 仍由产物卡展示。
|
||||
正文组成、0 层正文组成和开场的画面由产物卡按 frontend 渲染。完整 JSON 仍由产物卡展示。
|
||||
|
||||
## 6. 工作包 D · 游玩投影编译〔未开始〕
|
||||
|
||||
|
||||
Binary file not shown.
@@ -0,0 +1,128 @@
|
||||
# 提示缓存核对：记账是否可信、首填为何仍≈0
|
||||
|
||||
> 状态：核对结论。承接 `docs/briefs/token-cost-report-2026-10-06.md` §1.5 / §4.P2。
|
||||
> 记录：2026-10-07。
|
||||
> 疑问：若 OpenRouter 上 DeepSeek 缓存能挂住，是否说明「首填几乎不吃缓存」是本地记账错了，或粘滞/TTL 才是主因？
|
||||
|
||||
## 1. 结论（先看这个）
|
||||
|
||||
1. **缓存记账可信。** 我们只是把 OpenRouter 的 `usage.prompt_tokens_details.cached_tokens`（及 DeepSeek 的 `prompt_cache_hit_tokens`）原样写入 `~/.writing-agent/stats`；不是自估、不是漏字段。
|
||||
2. **「首填≈0、retry 高」是上游真值，不是漏记。** 同轮对照里，绝大多数是 fill 几乎无缓存、fill-retry 立刻上到八成上下。若解析坏了，两边应同假。
|
||||
3. **DeepSeek 探测成功，不能推翻 Gemini 首填账。** 花销报告窗口主模型是 `google/gemini-3.*-flash`；DeepSeek 只证明「同 baseUrl + 同解析链路」在另一模型上能读到非零缓存。
|
||||
4. **粘滞/`session_id` 不是当前主账缺口。** 同会话内 fill→retry 已能吃到缓存，说明粘到了同一上游且前缀可复用；贵在 **每一轮第一枪大前缀仍全价**，以及 **重试常态**。
|
||||
5. **优化优先级不变：** 仍以 P0（fill 去工具循环、消灭 retry）为主；P2（跨 select/fill 稳定前缀）对 Gemini 有意义，但不是「先修记账」或「再抠 TTL」。
|
||||
|
||||
---
|
||||
|
||||
## 2. 记账链路（我们记了什么）
|
||||
|
||||
| 层 | 行为 |
|
||||
|---|---|
|
||||
| OpenRouter 响应 | `prompt_tokens` / `prompt_tokens_details.cached_tokens` / 可选 `cache_write_tokens` |
|
||||
| `parseUsage`（`src/llm/client.ts`） | 读 `cached_tokens`；DeepSeek 另认 `prompt_cache_hit_tokens` |
|
||||
| stats jsonl | `cachedTokens` / `cacheWriteTokens` 等落盘 |
|
||||
| 报告命中率 | `cached / prompt`（按请求平均或合计） |
|
||||
|
||||
单元测试覆盖 OpenAI details、DeepSeek hit/miss、cache write。探测脚本与生产走同一 `complete` → `parseUsage`。
|
||||
|
||||
**原始对照（2026-10-07，Gemini 探测）：** 直接 `fetch` OpenRouter，两次固定大前缀请求的 `usage` 均为：
|
||||
|
||||
```json
|
||||
"prompt_tokens_details": {
|
||||
"cached_tokens": 0,
|
||||
"cache_write_tokens": 0
|
||||
}
|
||||
```
|
||||
|
||||
落盘若记成 0，与上游一致。
|
||||
|
||||
---
|
||||
|
||||
## 3. 生产窗口再切一刀（缓存行为）
|
||||
|
||||
范围：`~/.writing-agent/stats` 中 2026-10-03、10-04、10-06、10-07 的 `dictate_agent`（与花销报告同一量级窗口，略含 10-07）。
|
||||
|
||||
| step | 请求 | 有缓存占比 | 均命中率 | 中位命中 |
|
||||
|---|---:|---:|---:|---:|
|
||||
| select | 79 | 1.3% | ~0.8% | 0% |
|
||||
| **fill** | 79 | **1.3%** | **~0.5%** | **0%** |
|
||||
| **fill-retry** | 83 | **84.3%** | **~71.9%** | **~85%** |
|
||||
| reply | 30 | 0% | 0% | 0% |
|
||||
|
||||
同轮 `fill` + 至少一次 `fill-retry` 的配对约 **45**：
|
||||
|
||||
| 模式 | 对数 |
|
||||
|---|---:|
|
||||
| 首填≈0 且 retry 命中 >40% | **34** |
|
||||
| 两者都≈0 | 11 |
|
||||
| 首填命中 >20% | **0** |
|
||||
|
||||
样本：同轮 `fill` `cached=0`、随后 `fill-retry` `prompt≈90k / cached≈82k`（命中约 91%）。这只能解释为：**上游在 retry 上报了高缓存，我们照记。**
|
||||
|
||||
模型侧：窗口几乎全是 `google/gemini-3.7-flash` / `gemini-3.8-flash`；与花销报告一致。
|
||||
|
||||
---
|
||||
|
||||
## 4. 探测：DeepSeek vs Gemini（同 OpenRouter）
|
||||
|
||||
核对方式：固定大前缀连打 OpenRouter，读 `usage.prompt_tokens_details.cached_tokens`（脚本已删，结论仍有效）。
|
||||
|
||||
### 4.1 `deepseek-v4-pro`
|
||||
|
||||
| 调用 | prompt | cached | 命中 |
|
||||
|---|---:|---:|---:|
|
||||
| A#1 | 10602 | 7650 | 72.2% |
|
||||
| A#2～A#4 | 10602 | 9743 | 91.9% |
|
||||
| 等 60s 后 B#1～B#2 | 10602 | 9743 | 91.9% |
|
||||
|
||||
→ 隐式缓存可用；短间隔爬升；**约 1 分钟间隔未掉光**。说明「同 baseUrl + 粘滞 + 记账」这条链对 DeepSeek 成立。
|
||||
|
||||
### 4.2 `google/gemini-3.8-flash`
|
||||
|
||||
| 调用 | prompt | cached | 命中 |
|
||||
|---|---:|---:|---:|
|
||||
| A#1～A#3（间隔 5s） | ~12.7k～15k | **0** | **0%** |
|
||||
|
||||
原始 `usage` 同上，`cached_tokens=0` 且 `cache_write_tokens=0`。
|
||||
|
||||
→ **不是我们没记上**，是这次探测形态下上游就报 0。与生产里「仅当 fill 刚跑过、retry 几乎同前缀」才高频命中，并不矛盾：retry 复用的是**刚发过的整包 fill 前缀**；探测的「新会话 + 纯文本 system」未复现那种命中。
|
||||
|
||||
OpenRouter 文档对 Gemini 同时提到隐式缓存与 `cache_control` 显式断点；本仓库**未打** `cache_control`。生产 retry 仍能命中，说明至少部分 Gemini 路由在「极近、极同前缀」下会报 `cached_tokens>0`；**不能**因此假设跨 select→fill、或跨轮首填也会自动吃到折扣。
|
||||
|
||||
---
|
||||
|
||||
## 5. 对花销报告结论的修正与保留
|
||||
|
||||
| 命题 | 判定 |
|
||||
|---|---|
|
||||
| 首填缓存几乎失效 | **保留**（记账复核后仍成立） |
|
||||
| 贵主要因为 fill 大前缀 ×（1+retry） | **保留** |
|
||||
| 怀疑「其实有缓存但没记」 | **否定** |
|
||||
| 怀疑「DeepSeek 能缓存 ⇒ 当前账单不该贵」 | **否定**（账单模型是 Gemini；且 DeepSeek≠本窗口主路径） |
|
||||
| 怀疑「再抠 TTL / session 就能救首填」 | **弱化**：同会话 retry 已证明粘滞可用；缺口在 **select≠fill 前缀** 与 **每轮第一枪**，不是「完全没粘住」 |
|
||||
| P2 共享稳定前缀 | **仍值得做**（对 Gemini 跨步命中），但收益次于 P0 去循环 |
|
||||
| 换 DeepSeek 指望隐式缓存 | **可作实验**，质量/工具/价格需另验；不能单靠本次探测定案 |
|
||||
|
||||
相对花销直觉（未缓存=1、缓存≈0.25）：retry 已打折，**未打折的是几乎每次的首填大包**。消灭 retry 或把首填变小，都比「怀疑仪表」更贴账单。
|
||||
|
||||
---
|
||||
|
||||
## 6. 建议（本核对后的落地顺序）
|
||||
|
||||
与 `token-cost-report-2026-10-06.md` §5 对齐，本文件只钉缓存相关验收：
|
||||
|
||||
| 步 | 动作 | 验收 |
|
||||
|---|---|---|
|
||||
| A | **不改记账**；需要时用原始 `usage` 或 `/generation` 抽查 | 抽查与 jsonl `cachedTokens` 一致 |
|
||||
| B | P0：fill 改为一次 structured，去掉写工具循环 | 每轮调用 p50→2；fill-retry 占比 <10%（主收益） |
|
||||
| C | P2：select/fill 共用不变字节前缀，易变块置尾 | 同会话第 2 轮起 **fill 首次** `cached/prompt` 明显上升（Gemini 目标仍可订 ≥40%，以实测为准） |
|
||||
| D | 若坚持 Gemini 显式缓存 | 单独评估 OpenRouter `cache_control`；先小流量探测，再决定是否进主路径 |
|
||||
| E | 可选：用 DeepSeek 等隐式缓存友好模型做同书对照周 | 比的是 **有效单价 × 质量**，不是单看 hit% |
|
||||
|
||||
**明确不做：** 为「修缓存」去动 `parseUsage` 估命中；用 `provider.order` 换 `session_id`；把探测脚本的 DeepSeek 命中率写进创作账单当成已省钱。
|
||||
|
||||
---
|
||||
|
||||
## 7. 一句话答复原疑问
|
||||
|
||||
> 缓存记录是正确的。DeepSeek 能挂缓存，只说明链路和仪表没坏；当前创作窗口里 Gemini **首填几乎不打折是上游如实返回**，retry 高命中反而证明记对了。钱仍主要花在「每轮第一枪大前缀 + 填写真的在重试」，优先去 Agent 循环，而不是先怀疑统计或死磕 TTL。
|
||||
@@ -0,0 +1,219 @@
|
||||
# 创作对话 Token 花销报告与优化建议
|
||||
|
||||
> 状态：体验核对结论。数据来自本机 `~/.writing-agent/stats`（2026-10-02～10-06），对照 `docs/briefs/creation-round-cost.md` 旧基线与当前 `src/dictate/round.ts` 实现。
|
||||
> 记录：2026-10-06。
|
||||
> 诉求：对话不要继续走「自由 Agent 工具循环」，花销过大。
|
||||
|
||||
## 1. 结论（先看这个）
|
||||
|
||||
1. **主战场仍是创作落盘，不是游玩。** 本窗口 `dictate_agent` 约占总 prompt 的 **99%+**；游玩 Worker 几乎可忽略。
|
||||
2. **已经不是旧版「读甲写甲再读乙」自由 Agent**，而是 `select → 程序注入 → fill 工具循环（可 retry）→ 偶发 lean reply`。但 **fill 仍是 agent 形态**（`toolChoice: required` + 多工具 + 最多 4 次重试），这是当前贵的原因。
|
||||
3. **一句用户话平均约 3.3 次模型请求**，中位 3、P90 5、最多 6。其中 **60% 的轮次至少 1 次 fill-retry**，约 **29% 要 2 次及以上重试**。
|
||||
4. **相对花销（未缓存=1、缓存=0.25、输出=4）**：fill **53%**、fill-retry **28%**、select **15%**、reply **4%**。去 Agent 的收益应打在 **fill 一次写完、消灭重试、砍历史**，而不是再抠游玩。
|
||||
5. **首填缓存几乎失效**：fill 首次命中率约 **1%**，retry 才爬到约 **70%**。大前缀每一轮至少全价付一遍。
|
||||
|
||||
---
|
||||
|
||||
## 2. 本窗口花销账
|
||||
|
||||
统计范围：本地 jsonl，2026-10-02、10-03、10-04、10-06（10-05 无落盘）。共 **298** 条用量记录。
|
||||
|
||||
### 2.1 按 caller
|
||||
|
||||
| 路径 | 请求 | prompt | 其中缓存 | 命中率 | completion |
|
||||
|---|---:|---:|---:|---:|---:|
|
||||
| `dictate_agent` | 279 | **636.7 万** | 202.6 万 | 31.8% | 57.9 万 |
|
||||
| `worker:*` | 13 | 2.5 万 | ~0.1 万 | — | 0.3 万 |
|
||||
| `main_agent` | 6 | 0.06 万 | — | — | — |
|
||||
|
||||
创作 : 游玩（prompt）≈ **250 : 1**（本窗口游玩样本很少；方向与旧基线「创作远大于游玩」一致）。
|
||||
|
||||
### 2.2 按日（仅创作）
|
||||
|
||||
| 日 | 请求 | prompt（百万） | 缓存（百万） | completion（千） |
|
||||
|---|---:|---:|---:|---:|
|
||||
| 10-02 | 37 | 0.30 | 0.04 | 117 |
|
||||
| 10-03 | 86 | 2.25 | 0.76 | 187 |
|
||||
| 10-04 | 23 | 0.31 | 0.07 | 45 |
|
||||
| 10-06 | 133 | 3.50 | 1.15 | 229 |
|
||||
|
||||
### 2.3 创作按 step（关键）
|
||||
|
||||
| step | 请求 | prompt 合计 | **单次均 prompt** | 缓存命中 | completion |
|
||||
|---|---:|---:|---:|---:|---:|
|
||||
| **fill-retry** | 89 | 284.1 万 | **31.9k** | **70.1%** | 16.4 万 |
|
||||
| **fill** | 82 | 242.5 万 | **29.6k** | **1.1%** | 33.7 万 |
|
||||
| select | 85 | 89.6 万 | 10.5k | 0.9% | 5.0 万 |
|
||||
| reply | 23 | 20.5 万 | 8.9k | 0% | 2.7 万 |
|
||||
|
||||
要点：
|
||||
|
||||
- retry 请求数 **多于** 首次 fill（89 > 82）→ 重试是常态，不是边角。
|
||||
- 首次 fill 几乎不吃缓存；钱主要花在「每一轮第一枪大前缀」。
|
||||
- select 已不带产物正文（实现正确），但仍带全量更早对话，均约 10k prompt。
|
||||
- lean reply 只覆盖约 **28%** 有 fill 的轮次；其余多在 fill 同请求里 `reply_user`（设计意图），或失败后走 prose 回退。
|
||||
|
||||
### 2.4 按用户轮次（`turnId`）
|
||||
|
||||
| 指标 | 值 |
|
||||
|---|---:|
|
||||
| 轮次数 | 85 |
|
||||
| 每轮模型调用 avg / p50 / p90 / max | **3.28 / 3 / 5 / 6** |
|
||||
| 每轮 prompt 合计 avg / p50 / p90 / max | **7.5 万 / 6.0 万 / 14.9 万 / 25.9 万** |
|
||||
| 无 retry | 34（40%） |
|
||||
| ≥1 次 retry | 51（60%） |
|
||||
| ≥2 次 retry | 25（29%） |
|
||||
|
||||
### 2.5 上下文块（`segmentChars`，字符，非 token）
|
||||
|
||||
| step | n | rules | history | **products** | injected | tail | 均 prompt tok |
|
||||
|---|---:|---:|---:|---:|---:|---:|---:|
|
||||
| fill | 82 | ~2.1k | ~9.3k | **~23.0k** | ~5.6k | ~5.0k | 29.6k |
|
||||
| fill-retry | 89 | ~2.1k | ~8.7k | **~21.9k** | ~6.3k | ~4.9k | 31.9k |
|
||||
| select | 85 | ~0.3k | ~9.0k | **0** | 0 | ~4.9k | 10.5k |
|
||||
| reply | 23 | ~0.5k | 0 | ~6.7k（仅本轮） | 0 | ~6.3k | 8.9k |
|
||||
|
||||
近期 fill 样本（10-06）：products 已到 **4.3～5.0 万字**，history **2.6～3.2 万字**，单次 prompt **5.5～6.9 万 token**，且 `cached=0`。
|
||||
|
||||
### 2.6 模型
|
||||
|
||||
| 模型 | 请求 | prompt（百万） | 命中率 |
|
||||
|---|---:|---:|---:|
|
||||
| `google/gemini-3.8-flash` | 171 | 4.13 | 29.6% |
|
||||
| `google/gemini-3.7-flash` | 102 | 2.17 | 36.4% |
|
||||
| `openai/gpt-6.1-sol` | 6 | 0.07 | 16.1% |
|
||||
|
||||
金额粗算（按 Flash 量级：未缓存输入 $0.15/M、缓存 $0.0375/M、输出 $0.60/M，**仅作数量级**）：本窗口创作约 **$1.1**，约 **$0.013/用户轮**。体感「贵」主要来自 **轮次×大前缀×重试**，以及若换成更贵模型会线性放大；不是输出写爆。
|
||||
|
||||
### 2.7 与旧基线对照
|
||||
|
||||
旧账（`creation-round-cost.md`，约 9/20 起、创作 416 次 / prompt 1198 万）：自由工具循环、先 `read_module` 再 `insert`。
|
||||
|
||||
| | 旧自由 Agent | 本窗口（已 planned） |
|
||||
|---|---|---|
|
||||
| 形状 | 读/写交织、工具定义常驻 | select + 注入 + fill 循环 |
|
||||
| 同轮请求倍数 | 常「每节点×2」再加回复 | 中位 **3**（已封顶感更好） |
|
||||
| 产物进选题 | 会进（旧整包） | **已去掉** |
|
||||
| 填写大头 | 全产物 + 全对话 | **仍是全产物 + 近乎全对话** |
|
||||
| 首填缓存 | 依赖粘滞 | **仍≈0**（select/fill 前缀形状不同） |
|
||||
|
||||
**程序已省下「读模块那一轮」和选题侧产物**；**未省下「大前缀×（1+重试）」**。用户觉得「还是 Agent、还是贵」，与数据一致。
|
||||
|
||||
---
|
||||
|
||||
## 3. 当前链路（体验视角）
|
||||
|
||||
```text
|
||||
用户一句
|
||||
→ select（小上下文，强制 select_nodes） ≈1 次，~10k prompt
|
||||
→ 程序注入本轮节点写法 / 库正文（确定性，不计模型）
|
||||
→ fill（大上下文，强制工具：insert/delete/变量/映射 + reply_user）
|
||||
└─ 未过则 append 工具结果，再 fill-retry（最多约 4）
|
||||
→ 若填完却未 reply_user → lean reply（只带本轮产物） ≈9k
|
||||
```
|
||||
|
||||
与「不要 Agent」的落差：
|
||||
|
||||
| 已做到 | 仍像 Agent |
|
||||
|---|---|
|
||||
| 选题与填写拆开 | fill 仍多工具 + required |
|
||||
| 规格由程序注入，不再 `read_module` | 校验失败靠模型再跑一整轮大前缀 |
|
||||
| 回复可瘦身 | 变量/映射/删除仍走 tool loop |
|
||||
| `session_id` 粘滞已接 | select→fill 系统提示与正文布局不同，首填缓存打不穿 |
|
||||
|
||||
---
|
||||
|
||||
## 4. 优化建议（按性价比）
|
||||
|
||||
### P0 — 对话填写去 Agent（最贴合诉求）
|
||||
|
||||
**目标形状：一句用户话 = 2 次模型请求（澄清时 1 次），禁止 fill 工具循环。**
|
||||
|
||||
1. **select**：保持现状（或改为纯 JSON structured output，去掉 tools 税）。只出 `module_id[]` / 澄清题。
|
||||
2. **fill**：改为 **一次 structured complete**：`{ products: [...], reply, declares?, deletes? }`，程序本地切开落盘与校验。
|
||||
3. **失败只重写失败 tag**：上下文 = 规则摘要 + 该 tag 注入规格 + 相关产物白名单 + 错误信息；**不要**再塞全黑板 + 全历史重跑。
|
||||
4. **变量/映射**：优先由程序从 structured 字段落盘，或极窄工具面；不要与长正文 insert 共用大循环。
|
||||
|
||||
预期（相对本窗口相对花销）：
|
||||
|
||||
- 去掉 fill-retry 常态 → 相对花销约 **-25%～35%**
|
||||
- 首填仍大，但每轮从 3.3 枪降到 ~2 枪 → 再 **-20%～30%**（视澄清/失败率）
|
||||
- 合计数量级：**创作输入花销降到现在的一半左右**（前提：一次交多份成功率够高）
|
||||
|
||||
与相位机约束不冲突：仍在同一次 `running` 内、用户不手选节点、方案不是执行序。
|
||||
|
||||
### P1 — 砍 history（实现便宜、收益稳）
|
||||
|
||||
现状：select/fill 的「更早对话」均值约 9k 字，会话后期到 **3 万字+**，占填写真份额。
|
||||
|
||||
建议：
|
||||
|
||||
- select：方案进度 + **最近 2～4 轮** + 末句（旧 brief 已论证「只留末句会误选」，不能只留一句）。
|
||||
- fill：同样只留最近对话；产物冲突靠产物正文，不靠早期闲聊。
|
||||
- 清空对话能力保持；长记忆另走总结产物，不靠无限 chat。
|
||||
|
||||
预期：fill/select 各再 **-10%～20%** prompt；会话越长收益越大。
|
||||
|
||||
### P2 — 稳首填缓存前缀
|
||||
|
||||
`session_id` 只保证同上游；**select 与 fill 的 system + 前部布局不一致 → 首填命中≈0**。
|
||||
|
||||
建议（不改业务语义）：
|
||||
|
||||
- 抽出 **跨 select/fill 不变的规则前缀**（身份、写入规则骨架），两边共用同一字节前缀。
|
||||
- 易变块（方案进度、checklist、注入、末句）固定在末尾（现已部分如此）。
|
||||
- 工具定义：若 fill 改 structured，则工具 JSON 不再每轮跟着大请求。
|
||||
|
||||
预期：在粘滞有效时，同一会话后续轮次的「规则+稳定壳」可从全价变折扣；**不降低发送体积，但降低未缓存账单**。对 OpenRouter+Gemini 块粒度（约 4k）敏感，前缀差几百 token 可能整块失效。
|
||||
|
||||
### P3 — 产物相关性（大头，风险高）
|
||||
|
||||
products 占 fill 字符大头（样本约一半以上）。旧 brief 已说明：误裁会导致冲突设定。
|
||||
|
||||
建议分两档：
|
||||
|
||||
1. **安全档**：按本轮 `module_id` 维护「必带 tag」白名单（依赖边写在 catalog）；其余只带 **标题+一行摘要+hash**。
|
||||
2. **激进档**：检索/摘要。必须配套：写入前冲突扫描、失败则自动补全文重试（否则质量回退会被当成「省钱事故」）。
|
||||
|
||||
预期：做好可再砍 fill **30%～50%**；做不好会表现为反复 rewrite，把 P0 收益吃掉。
|
||||
|
||||
### P4 — 游玩与总管
|
||||
|
||||
本窗口不必优先。游玩保持「一次生成、非创作 tool loop」。OpenRouter 上游玩缓存曾为 0，靠 `session_id` 已具备前提；前缀顺序现已正确。总管 `main_agent` 可忽略。
|
||||
|
||||
### 明确不做 / 慎做
|
||||
|
||||
| 项 | 原因 |
|
||||
|---|---|
|
||||
| 改回用户手选节点以省一轮 | 与产品「自然语言推进」冲突 |
|
||||
| 相位机按固定工序跑完全部节点 | 与「方案不是执行序」冲突 |
|
||||
| 选题带全产物 | 已证明浪费；会毁掉 select 的小请求优势 |
|
||||
| 用 `provider.order` 代替 `session_id` | 会关掉粘滞 |
|
||||
| 只砍输出 | 输出约是 prompt 的 3%～9%，不是主账 |
|
||||
|
||||
---
|
||||
|
||||
## 5. 建议落地顺序（可验收）
|
||||
|
||||
| 步 | 改动 | 验收指标（同书同对话长度） |
|
||||
|---|---|---|
|
||||
| A | fill 改为一次 structured，去掉 write tool loop | 每轮模型调用 p50→**2**；fill-retry 调用占比 **<10%** |
|
||||
| B | select/fill history 窗口 | fill `segmentChars.history` 中位下降 **≥50%** |
|
||||
| C | 共享稳定 system 前缀 | 同会话第 2 轮起 fill 首次 `cached/prompt` **≥40%**（Gemini/OpenRouter） |
|
||||
| D | 产物白名单/摘要 | fill 均 prompt **再降 ≥25%**，且 insert 拒绝率不升高 |
|
||||
|
||||
体验验收（主观）：
|
||||
|
||||
- 用户感知仍是「说一句 → 落几块 → 一段综合回复」，无「工具来回思考很久」的 Agent 感。
|
||||
- 澄清轮仍可 1 次请求结束（仅选题提问）。
|
||||
- 失败时只看到「某几块重写」，而不是整轮长时间空转。
|
||||
|
||||
---
|
||||
|
||||
## 6. 和「不要 Agent」的直接答复
|
||||
|
||||
- **创作对话：应该停止以 Agent tool loop 作为填写主路径。** 选题可以极窄工具或 JSON；填写应程序注入规格 + 一次结构化交卷 + 本地校验。
|
||||
- **游玩：已经不是创作那种 Agent**；继续单次 Worker 生成即可。
|
||||
- **总管 Agent：不是当前账单**；可维持。
|
||||
|
||||
若只做一件事：做 **P0（fill 去循环）**。它同时回应「体感像 Agent」和「token 太大」两件事。
|
||||
@@ -0,0 +1,132 @@
|
||||
这份文档是自包含的。接收方看不到代码仓库、聊天记录和作品正文。请只根据本文做事。
|
||||
|
||||
请你做两件事，都要写成我们可以逐条核对的文字：
|
||||
|
||||
1. 列出业界在「长上下文 Agent、提示缓存、工具循环」上的常见优化方案。每条写清：做法、它节省的是请求次数、未缓存输入、缓存折扣价还是输出，以及它依赖什么前提（例如上游必须支持前缀缓存、模型必须能一次返回多个工具）。
|
||||
2. 针对下面描述的这个程序，给出你建议采用的组合方案。说明每一步落在哪一次模型请求、上下文里留哪些块、块的顺序、相位机的哪一段执行。不要假装读过仓库。标出你不确定、需要我们核对的点。
|
||||
|
||||
我们会用仓库和现有规则检查你的方案。与本文第三节冲突的做法请单独标出，不要藏进推荐里。
|
||||
|
||||
|
||||
一、这是什么程序
|
||||
|
||||
这是一个写作代理，分两段使用。
|
||||
|
||||
创作期：用户用自然语言描述想要的作品，不手动点选节点。模型决定这一轮要落实哪些「能力」，把结果写成黑板上的产物。产物是规格（舞台、规则、变量、正文如何呈现、开场），不是游玩回合的正文。聊天只做确认和追问。清空对话后产物还在，产物才是创作期的持久记忆。
|
||||
|
||||
游玩期：另一套 Worker 按已经写好的规格往下演。一次用户输入通常对应一次生成，不是创作那种多步工具循环。
|
||||
|
||||
主要渠道是 OpenRouter 上的 Gemini 与 OpenAI 兼容模型。也有直连的单一上游做对照。费用近似为：
|
||||
|
||||
请求次数 ×（未缓存输入 × 全价 + 缓存输入 × 折扣价 + 输出 × 输出价）
|
||||
|
||||
缓存命中不是免费的，只是更便宜。输出大约是 prompt 的 3% 到 9%，单价更高，但当前主账仍是输入。最近约三天（2026-09-28 10:35 到 2026-10-01 10:35，东八区）创作落盘 416 次请求、prompt 1199 万 token，游玩 29 次、prompt 80 万。创作大约是游玩的 14 倍。优化主战场是创作。
|
||||
|
||||
|
||||
二、会有哪些上下文
|
||||
|
||||
下面这些块都会进模型，但不是每一次请求全有。
|
||||
|
||||
创作落盘，一句用户话开始时拼一次，工具循环中途只在末尾追加：
|
||||
|
||||
- 系统提示：身份、写入规则、推进原则。尚无本局方案时附整份能力短目录（何时用、何时不用、档位）。已有方案时改为方案进度：每个能力一行，已落或未落、档位、理由、缺口。不含产物正文。样本里系统提示 6695 字，其中方案进度 4392 字。
|
||||
- 一条用户消息：全部已有产物，按相对顺序，然后全量对话。样本 25 份产物共 60518 字（约 69%），全量对话 19975 字（约 23%）。没有一份产物独大，最大的正文组成 8879 字，其次变量规则 6005、实现机制 4569，其余多为两三千字。会话中段产物已经约 5.6 万字。
|
||||
- 工具定义：每次都带，约 4917 字，一轮内不变。工具有读取能力、读取库条目、准备开场、写入产物、删除、声明变量、声明映射、清空对话。
|
||||
- 预设：包在系统侧的文风或格式片段。当前选用预设启用正文约 362 字，可忽略。换长预设时它变成稳定前缀，应放在易变内容之前。
|
||||
- 工具结果：循环后半才出现。读取能力的返回体是该能力的任务、原则、产物规格、评分和库目录，一份约 5000 字。库条目正文同理，目录里只有摘要。这些全文一开始不在上下文里。
|
||||
|
||||
游玩一次生成：
|
||||
|
||||
- 预设与该 Worker 的系统说明（相对稳定）。
|
||||
- 槽位人设。
|
||||
- 静态规格投影（舞台、规则、叙事、正文组成等，开局后很少改）。
|
||||
- 对话历史（每轮变长，按投影裁剪）。
|
||||
- 变量当前值与本轮用户输入（每轮都变）。
|
||||
|
||||
总管工具循环（决定何时调用哪个 Worker）另有一套 message 历史。最近三天的账几乎不在这条上，优化时不要把它和创作落盘混成同一次请求。
|
||||
|
||||
能力的完整写法存放在各自的提示文件里，不在系统目录里。不把所有能力的完整提示放进第一次请求。完整写法只在已经确定本轮要写哪几项之后进入上下文。
|
||||
|
||||
|
||||
三、相位机之内，程序可以动哪里
|
||||
|
||||
运行分两层。
|
||||
|
||||
业务阶段：创作（设计规格）然后游玩。这是作品生命周期，不是每一次模型调用。
|
||||
|
||||
运行相位：idle、running、waiting_user、done、error。相位机是纯函数：给定会话和事件，决定谁在场、哪些工具可用、何时必须等用户、产物何时算数。它不决定「下一步叫哪个能力」。下一步由 running 里的模型工具循环决定。
|
||||
|
||||
一句用户话是硬事件：相位进入 running，这一轮的模型调用次数从零计。创作落盘在这次 running 里跑工具循环，上限大约三十多步。循环开始时拼好开头，之后只追加助手工具调用和工具结果。模型不再欠请求、或写出给用户的回复之后，相位回到 waiting_user，等待下一句。
|
||||
|
||||
因此：
|
||||
|
||||
- 减少「读一份、写一份、再读下一份」必须发生在同一次 running 里，也就是两次用户话之间。
|
||||
- 不能把相位机改成按固定工序自动跑完全部节点。用户不手选节点，方案也不是执行顺序。
|
||||
- Worker 执行时总管暂停。游玩的那一次生成是 Worker 自己的请求，上下文按该 Worker 的投影拼，不复用创作工具循环的消息历史。
|
||||
- 程序可以在两次模型请求之间插入确定性步骤：注入选中能力的当前提示、核对版本哈希、逐条校验、只重试失败的那一条。这些不是相位变化。
|
||||
|
||||
|
||||
四、缓存优化
|
||||
|
||||
目标是稳定前缀留在同一上游的提示缓存里。不追求两次请求的整包字节相同。Agent 每一轮消息都会变。
|
||||
|
||||
已经接上：OpenRouter 请求体带本局 session_id（最长 256 字）。直连其它渠道不带，避免未知字段被拒绝。有 session_id 后，粘滞从该会话第一次成功请求就开始，不必先看到缓存命中，也不再因为第一条用户消息被重写就换上游。不要发送 provider.order，它会关掉粘滞。
|
||||
|
||||
接上之前的测量（不要把「再做一遍粘滞」当成主体）：
|
||||
|
||||
- 创作工具循环间隔约 7 到 15 秒。Gemini 上同一次循环内缓存从约 3800 token 按大约 4000 一块爬到 7600、11500、15000，整段 prompt 约 50% 到 68% 命中。每个 burst 的第一下经常是 0。中途偶尔整段掉回 0，是换了上游。
|
||||
- 游玩在 OpenRouter 上 28 次、prompt 1.8 万到 3.4 万，间隔约 5 到 22 分钟，缓存为 0。直连单一上游时，间隔 8 到 30 分钟仍常见命中 80% 到 91%。有一轮从约 1.1 万掉到 6656 并停住，说明前缀中段改过，只保住开头。
|
||||
- Gemini 隐式缓存大约 4000 token 一块记账。前缀差几百 token，账单上可以差一整块。
|
||||
|
||||
session_id 只保证落到同一上游。命中到第几个 token，取决于前缀从哪一字开始和上一次不同。
|
||||
|
||||
|
||||
五、上下文编排顺序
|
||||
|
||||
原则：越少改的越靠前，每轮都变的放在最后。一处改动只作废它后面的缓存。
|
||||
|
||||
游玩已按此排列：人设与静态规格、对话历史、变量与本轮输入。排序本身在直连上游上有效；OpenRouter 上曾经是 0，是因为当时没有粘住上游，不是因为顺序反了。
|
||||
|
||||
创作当前顺序是：系统提示（含会随落盘改写的方案进度）、然后一条用户消息里先全部产物再全量对话、工具定义在消息之外、工具结果接在末尾。产物任何一份被覆盖，如果它不在整包的末尾，它后面的产物和对话都会离开缓存。工具定义一轮内稳定，适合留在前缀可缓存的区域。方案进度放在系统提示里，方案一改，系统提示之后全部重计。
|
||||
|
||||
选择与填写若拆成两次请求，稳定的能力目录应在选择请求的前部，方案进度、上一问和本轮用户输入在末尾。填写请求里，少改的产物在前，本轮新注入的节点提示和最近对话在后。节点提示每轮都会换，不能放在全部产物之前，否则一换节点，后面整包产物都按未缓存计。
|
||||
|
||||
|
||||
六、上下文发送哪些
|
||||
|
||||
用同一本书量过，不发请求。当前一次创作请求的 messages 约 88218 字，末次实际 prompt 约 46897 token（大约 2 字 1 token）。
|
||||
|
||||
- 选择若只带方案进度加用户最后一句：约 4522 字。再加助手上一问仍然很小。产物正文不要放进选择。只看用户这句原文不够，否则回答「第二个」或修改已有地图时会对错节点，已写过的能力会被再选。
|
||||
- 填写若仍带全部产物，对话只留最近约两轮（样本 1770 字）：约 6.2 万字。砍掉全量对话大约只省两成。
|
||||
- 再注入一份节点提示约 5000 字。一次注入两三份，是在产物之上加一两成，而且只发送一次。
|
||||
|
||||
所以「少发对话」省得有限。「少发哪些产物」才碰到大头，但填写时不带相关产物就会和已有设定冲突。哪些产物是这份能力的材料、哪些只是其它规格，本文没有一份可执行的白名单。常见的相关性裁剪、摘要、检索如果要用，必须说明误裁后怎样发现。
|
||||
|
||||
模型在看见该能力当前版本的任务、原则、规格、评分之前，写出来的正文不算数。程序用这些段落的哈希做版本。同一轮生成里先调用读取、同时把正文交上去，不算看见过写法。不要求先读的只有 narrative、variable-context、context-order、worker-spec、refine 和标记为自动的程序步骤。库条目正文同样要先看见再写进产物，默认 1 到 2 条。
|
||||
|
||||
|
||||
七、程序上的优化，仍在同一次 running 里
|
||||
|
||||
现在的倍数来自工具循环，不是来自相位来回切换。
|
||||
|
||||
同一项内容能力至少两次模型请求：第一次凭短说明选定并读取，第二次看见写法之后写入。提示要求一轮通常写这句已经说清的 2 到 3 个能力。模型常一次只返回一个工具，于是变成读甲、写甲、读乙、写乙，最后再单独回用户一句。驱动程序其实会在同一步执行该次响应里的全部工具，日志里偶发一次写两份。
|
||||
|
||||
讨论过、尚未定案的程序形状，供你对照业界做法，不是已经同意的设计：
|
||||
|
||||
- 同一次 running 里两次模型请求。第一次只输出能力 id 和库条目 id。程序注入对应提示并记下版本。第二次一次交回多份产物和一段给用户的回复。某一条校验失败只重写那一条。
|
||||
- 若模型一次只能调用一个工具，用一次结构化输出携带多条产物，程序在本地切开。这只去掉写入侧的重复前缀。规格说明仍须在这次生成之前就在上下文里。
|
||||
- 写开场前有一次程序准备：编译信息可见范围、游玩执行方式、上下文排序和运行配置。这一步不应再变成一次让模型自由发挥的长上下文请求。
|
||||
|
||||
必须仍然成立的行为：用户不手选节点；方案不是执行顺序；没说清的「必须」和「有必要」本轮不写，先问 2 到 3 个选项；同轮多份产物只给一次综合回复，不把完整 JSON 贴进回复；变量和映射不走普通写入。
|
||||
|
||||
|
||||
八、请你列出的常见方案，以及针对我们的组合
|
||||
|
||||
常见方案请覆盖至少这四类，每类不要只写一句口号。我们没给仓库，所以请把业界做法写全，包括你最后不推荐的。我们会删掉不适合的。
|
||||
|
||||
1. 缓存：粘滞路由、前缀稳定性、缓存块粒度、显式缓存断点、把变化赶出前缀。session_id 已经有了，请写它解决不了的部分。
|
||||
2. 编排顺序：稳定段、历史、本轮输入的常见排序，以及中段一改、后段全失效时业界怎么处理。
|
||||
3. 发送哪些：全文产物、摘要、按能力取材料、检索、滑动窗口、把目录和正文拆开。请写误删材料的后果。
|
||||
4. 程序：一次多工具、结构化输出一次交多份、选择与填写拆请求、把规格预注入从而取消读取轮、限制工具循环次数。请写哪种依赖模型能力，哪种只依赖程序。
|
||||
|
||||
然后给出你建议我们采用的组合：一句用户话几次请求、每次上下文块的顺序、预计相对现在少发送几次大前缀、产物那约 6 万字还在不在填写请求里。故意不做的项写原因。和第三节、第六节冲突的地方单独列出，不要写进推荐步骤。
|
||||
@@ -0,0 +1,84 @@
|
||||
# 从启动到开玩
|
||||
|
||||
解压或下载完成后，按下面的顺序即可：双击启动、打开浏览器、写一张最简单的卡，然后游玩一句。
|
||||
|
||||
需要先安装 [Node.js](https://nodejs.org/)（勾选安装时附带的 npm）。第一次启动会自动安装依赖，请等黑窗口跑完。
|
||||
|
||||
## 1. 双击 `start.bat`
|
||||
|
||||
在文件夹里双击 `start.bat`。会弹出一个黑窗口，并自动打开浏览器。地址是：
|
||||
|
||||
http://localhost:23337
|
||||
|
||||
窗口不要关。关窗口或按 `Ctrl+C` 就是停止软件。
|
||||
|
||||
|
||||
|
||||
想用桌面窗口而不是浏览器，改双击 `start-app.bat`。后面的操作一样。
|
||||
|
||||
## 2. 第一次要填模型接口
|
||||
|
||||
点右上角 **设置**，再点 **新增**。填上这四项后点 **保存并选用**：
|
||||
|
||||
- 名称：自己认得就行，例如 DeepSeek
|
||||
- Base URL：接口地址，例如 `https://api.deepseek.com/v1`
|
||||
- API Key：你的密钥
|
||||
- 型号：这一行要用的模型名，例如 `deepseek-chat`
|
||||
|
||||
填完型号后也可以点 **嗅探**，从列表里点一个型号。
|
||||
|
||||
|
||||
|
||||
项目根目录如果已经有填好的 `.env`，启动时会自动带上，这一步可以跳过。
|
||||
|
||||
## 3. 新建作品
|
||||
|
||||
还没有任何作品时，打开页面就会弹出 **新建作品**。输入名称，点 **创建**。以后再开新的，点左侧 **+**。
|
||||
|
||||
|
||||
|
||||
这里用「雨天小书店」做例子。
|
||||
|
||||
## 4. 写一张最简单的卡
|
||||
|
||||
在底部输入框里原样贴上这段，点发送（或按 Enter）：
|
||||
|
||||
```text
|
||||
请写一张最简单的可玩卡，写完就停。
|
||||
|
||||
题材：雨天小书店。玩家是进来躲雨的客人，店员是一只戴蓝色围巾的猫。
|
||||
只要短对话。不要数值、地图、复杂世界观。
|
||||
开场只要两三句：雨还在下，猫店员抬起头。
|
||||
```
|
||||
|
||||
|
||||
|
||||
等它写完。中间的回复会说明卡已经好了，右侧 **已持久化** 是落下来的几张卡片。
|
||||
|
||||
|
||||
|
||||
写完它如果还追问，回复「都不用再加，写完开场就可以开玩」。
|
||||
|
||||
## 5. 保存并开玩
|
||||
|
||||
右上角出现 **保存定稿** 时，点它，给这份稿起个名字。没有这个按钮时，点输入框右侧的保存图标，效果相同。
|
||||
|
||||
然后点左侧的作品名，在弹出的菜单里选 **游玩**。
|
||||
|
||||
会先看到开场白：
|
||||
|
||||
|
||||
|
||||
在底部写你要做的事，再发送。例如：
|
||||
|
||||
```text
|
||||
我收起伞，问店员今天适合躲雨看的书。
|
||||
```
|
||||
|
||||
它会接下去写这一轮：
|
||||
|
||||
|
||||
|
||||
之后就一直这样：你说一句，它写一轮。右上角 **保存游玩** 可以把进度存下来。
|
||||
|
||||
想换你在故事里的名字，到 **设置 → 用户角色** 里新增一个角色并选用。创作时如果希望开场先写占位，把角色名写成 `@玩家`。
|
||||
Generated
+455
-8
@@ -8,6 +8,7 @@
|
||||
"name": "writing-agent",
|
||||
"version": "0.1.0",
|
||||
"dependencies": {
|
||||
"msedge-tts": "^2.0.9",
|
||||
"yaml": "^2.9.0"
|
||||
},
|
||||
"devDependencies": {
|
||||
@@ -1046,6 +1047,18 @@
|
||||
"url": "https://opencollective.com/vitest"
|
||||
}
|
||||
},
|
||||
"node_modules/agent-base": {
|
||||
"version": "6.0.2",
|
||||
"resolved": "https://registry.npmjs.org/agent-base/-/agent-base-6.0.2.tgz",
|
||||
"integrity": "sha512-RZNwNclF7+MS/8bDg70amg32dyeZGZxiDuQmZxKLAlQjr3jGyLx+4Kkk58UO7D2QdgFIQCovuSuZESne6RG6XQ==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"debug": "4"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 6.0.0"
|
||||
}
|
||||
},
|
||||
"node_modules/assertion-error": {
|
||||
"version": "2.0.1",
|
||||
"resolved": "https://registry.npmjs.org/assertion-error/-/assertion-error-2.0.1.tgz",
|
||||
@@ -1056,6 +1069,44 @@
|
||||
"node": ">=12"
|
||||
}
|
||||
},
|
||||
"node_modules/asynckit": {
|
||||
"version": "0.4.0",
|
||||
"resolved": "https://registry.npmjs.org/asynckit/-/asynckit-0.4.0.tgz",
|
||||
"integrity": "sha512-Oei9OH4tRh0YqU3GxhX79dM/mwVgvbZJaSNaRk+bshkj0S5cfHcgYakreBjrHwatXKbz+IoIdYLxrKim2MjW0Q==",
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/axios": {
|
||||
"version": "1.20.0",
|
||||
"resolved": "https://registry.npmjs.org/axios/-/axios-1.20.0.tgz",
|
||||
"integrity": "sha512-r8aOh8j9cGKpgQAqpzrUHnSIc6a59Y3Xf/cv8sy1DrHCkZHzQGEuoq1tARk6qSyDdtQGSDgpb9kFlruzPvrgwg==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"follow-redirects": "^1.16.0",
|
||||
"form-data": "^4.0.6",
|
||||
"https-proxy-agent": "^5.0.1",
|
||||
"proxy-from-env": "^2.1.0"
|
||||
}
|
||||
},
|
||||
"node_modules/base64-js": {
|
||||
"version": "1.5.1",
|
||||
"resolved": "https://registry.npmjs.org/base64-js/-/base64-js-1.5.1.tgz",
|
||||
"integrity": "sha512-AKpaYlHn8t4SVbOHCy+b5+KKgvR4vrsD8vbvrbiQJps7fKDTkjkDry6ji0rUJjC0kzbNePLwzxq8iypo41qeWA==",
|
||||
"funding": [
|
||||
{
|
||||
"type": "github",
|
||||
"url": "https://github.com/sponsors/feross"
|
||||
},
|
||||
{
|
||||
"type": "patreon",
|
||||
"url": "https://www.patreon.com/feross"
|
||||
},
|
||||
{
|
||||
"type": "consulting",
|
||||
"url": "https://feross.org/support"
|
||||
}
|
||||
],
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/boolean": {
|
||||
"version": "3.2.0",
|
||||
"resolved": "https://registry.npmjs.org/boolean/-/boolean-3.2.0.tgz",
|
||||
@@ -1065,6 +1116,30 @@
|
||||
"license": "MIT",
|
||||
"optional": true
|
||||
},
|
||||
"node_modules/buffer": {
|
||||
"version": "6.0.3",
|
||||
"resolved": "https://registry.npmjs.org/buffer/-/buffer-6.0.3.tgz",
|
||||
"integrity": "sha512-FTiCpNxtwiZZHEZbcbTIcZjERVICn9yq/pDFkTl95/AxzD1naBctN7YO68riM/gLSDY7sdrMby8hofADYuuqOA==",
|
||||
"funding": [
|
||||
{
|
||||
"type": "github",
|
||||
"url": "https://github.com/sponsors/feross"
|
||||
},
|
||||
{
|
||||
"type": "patreon",
|
||||
"url": "https://www.patreon.com/feross"
|
||||
},
|
||||
{
|
||||
"type": "consulting",
|
||||
"url": "https://feross.org/support"
|
||||
}
|
||||
],
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"base64-js": "^1.3.1",
|
||||
"ieee754": "^1.2.1"
|
||||
}
|
||||
},
|
||||
"node_modules/buffer-crc32": {
|
||||
"version": "0.2.13",
|
||||
"resolved": "https://registry.npmjs.org/buffer-crc32/-/buffer-crc32-0.2.13.tgz",
|
||||
@@ -1114,6 +1189,19 @@
|
||||
"node": ">=8"
|
||||
}
|
||||
},
|
||||
"node_modules/call-bind-apply-helpers": {
|
||||
"version": "1.0.2",
|
||||
"resolved": "https://registry.npmjs.org/call-bind-apply-helpers/-/call-bind-apply-helpers-1.0.2.tgz",
|
||||
"integrity": "sha512-Sp1ablJ0ivDkSzjcaJdxEunN5/XvksFJ2sMBFfq6x0ryhQV/2b/KwFe21cMpmHtPOSij8K99/wSfoEuTObmuMQ==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"es-errors": "^1.3.0",
|
||||
"function-bind": "^1.1.2"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/chai": {
|
||||
"version": "5.3.3",
|
||||
"resolved": "https://registry.npmjs.org/chai/-/chai-5.3.3.tgz",
|
||||
@@ -1154,11 +1242,22 @@
|
||||
"url": "https://github.com/sponsors/sindresorhus"
|
||||
}
|
||||
},
|
||||
"node_modules/combined-stream": {
|
||||
"version": "1.0.8",
|
||||
"resolved": "https://registry.npmjs.org/combined-stream/-/combined-stream-1.0.8.tgz",
|
||||
"integrity": "sha512-FQN4MRfuJeHf7cBbBMJFXhKSDq+2kAArBlmRBvcvFE5BB1HZKXtSFASDhdlz9zOYwxh8lDdnvmMOe/+5cdoEdg==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"delayed-stream": "~1.0.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.8"
|
||||
}
|
||||
},
|
||||
"node_modules/debug": {
|
||||
"version": "4.4.3",
|
||||
"resolved": "https://registry.npmjs.org/debug/-/debug-4.4.3.tgz",
|
||||
"integrity": "sha512-RGwwWnwQvkVfavKVt22FGLw+xYSdzARwm0ru6DhTVA3umU5hZc28V3kO4stgYryrTlLpuvgI9GiijltAjNbcqA==",
|
||||
"dev": true,
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"ms": "^2.1.3"
|
||||
@@ -1259,6 +1358,15 @@
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/delayed-stream": {
|
||||
"version": "1.0.0",
|
||||
"resolved": "https://registry.npmjs.org/delayed-stream/-/delayed-stream-1.0.0.tgz",
|
||||
"integrity": "sha512-ZySD7Nf91aLB0RxL4KGrKHBXl7Eds1DAmEdcoVawXnLD7SDhpNgtuII2aAkg7a7QS41jxPSZ17p4VdGnMHk3MQ==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">=0.4.0"
|
||||
}
|
||||
},
|
||||
"node_modules/detect-node": {
|
||||
"version": "2.1.0",
|
||||
"resolved": "https://registry.npmjs.org/detect-node/-/detect-node-2.1.0.tgz",
|
||||
@@ -1267,6 +1375,20 @@
|
||||
"license": "MIT",
|
||||
"optional": true
|
||||
},
|
||||
"node_modules/dunder-proto": {
|
||||
"version": "1.0.1",
|
||||
"resolved": "https://registry.npmjs.org/dunder-proto/-/dunder-proto-1.0.1.tgz",
|
||||
"integrity": "sha512-KIN/nDJBQRcXw0MLVhZE9iQHmG68qAVIBg9CqmUYjmQIhgij9U5MFvrqkUL5FbtyyzZuOeOt0zdeRe4UY7ct+A==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"call-bind-apply-helpers": "^1.0.1",
|
||||
"es-errors": "^1.3.0",
|
||||
"gopd": "^1.2.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/electron": {
|
||||
"version": "35.7.5",
|
||||
"resolved": "https://registry.npmjs.org/electron/-/electron-35.7.5.tgz",
|
||||
@@ -1310,9 +1432,7 @@
|
||||
"version": "1.0.1",
|
||||
"resolved": "https://registry.npmjs.org/es-define-property/-/es-define-property-1.0.1.tgz",
|
||||
"integrity": "sha512-e3nRfgfUZ4rNGL232gUgX06QNyyez04KdjFrF+LTRoOXmrOgFKDg4BCdsjW8EnT69eqdYGmRpJwiPVYNrCaW3g==",
|
||||
"dev": true,
|
||||
"license": "MIT",
|
||||
"optional": true,
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
@@ -1321,9 +1441,7 @@
|
||||
"version": "1.3.0",
|
||||
"resolved": "https://registry.npmjs.org/es-errors/-/es-errors-1.3.0.tgz",
|
||||
"integrity": "sha512-Zf5H2Kxt2xjTvbJvP2ZWLEICxA6j+hAmMzIlypy4xcBg1vKVnx89Wy0GbS+kf5cwCVFFzdCFh2XSCFNULS6csw==",
|
||||
"dev": true,
|
||||
"license": "MIT",
|
||||
"optional": true,
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
@@ -1335,6 +1453,33 @@
|
||||
"dev": true,
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/es-object-atoms": {
|
||||
"version": "1.1.2",
|
||||
"resolved": "https://registry.npmjs.org/es-object-atoms/-/es-object-atoms-1.1.2.tgz",
|
||||
"integrity": "sha512-HWcBoN6NileqtSydK2FqHbS/LoDd2pqrnQHLyJzBj4kOp/ky2MWMN694xOfkK8/SnUsW2DH7EfyVlydKCsm1Zw==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"es-errors": "^1.3.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/es-set-tostringtag": {
|
||||
"version": "2.1.0",
|
||||
"resolved": "https://registry.npmjs.org/es-set-tostringtag/-/es-set-tostringtag-2.1.0.tgz",
|
||||
"integrity": "sha512-j6vWzfrGVfyXxge+O0x5sh6cvxAog0a/4Rdd2K36zCMV5eJ+/+tOAngRO8cODMNWbVRdVlmGZQL2YS3yR8bIUA==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"es-errors": "^1.3.0",
|
||||
"get-intrinsic": "^1.2.6",
|
||||
"has-tostringtag": "^1.0.2",
|
||||
"hasown": "^2.0.2"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/es6-error": {
|
||||
"version": "4.1.1",
|
||||
"resolved": "https://registry.npmjs.org/es6-error/-/es6-error-4.1.1.tgz",
|
||||
@@ -1450,6 +1595,42 @@
|
||||
"pend": "~1.2.0"
|
||||
}
|
||||
},
|
||||
"node_modules/follow-redirects": {
|
||||
"version": "1.16.1",
|
||||
"resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.16.1.tgz",
|
||||
"integrity": "sha512-FNvFGzoMLWmE6Yj9spb/zjd7yiNCHiAW9/Tg9CXrQ8wuu32HtlJOwWO11OJafl5FfY3DxTdQ0vj42zU1kvv5jg==",
|
||||
"funding": [
|
||||
{
|
||||
"type": "individual",
|
||||
"url": "https://github.com/sponsors/RubenVerborgh"
|
||||
}
|
||||
],
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">=4.0"
|
||||
},
|
||||
"peerDependenciesMeta": {
|
||||
"debug": {
|
||||
"optional": true
|
||||
}
|
||||
}
|
||||
},
|
||||
"node_modules/form-data": {
|
||||
"version": "4.0.6",
|
||||
"resolved": "https://registry.npmjs.org/form-data/-/form-data-4.0.6.tgz",
|
||||
"integrity": "sha512-vKatAh4SlVfgbv+YtmhiRjhEMJsYpsG1Y2rMQtR+SVSbytsSD1YGzDIcrAJmdFec88u/+VoGmxnl+80gL1tRCQ==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"asynckit": "^0.4.0",
|
||||
"combined-stream": "^1.0.8",
|
||||
"es-set-tostringtag": "^2.1.0",
|
||||
"hasown": "^2.0.4",
|
||||
"mime-types": "^2.1.35"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 6"
|
||||
}
|
||||
},
|
||||
"node_modules/fs-extra": {
|
||||
"version": "8.1.0",
|
||||
"resolved": "https://registry.npmjs.org/fs-extra/-/fs-extra-8.1.0.tgz",
|
||||
@@ -1480,6 +1661,52 @@
|
||||
"node": "^8.16.0 || ^10.6.0 || >=11.0.0"
|
||||
}
|
||||
},
|
||||
"node_modules/function-bind": {
|
||||
"version": "1.1.2",
|
||||
"resolved": "https://registry.npmjs.org/function-bind/-/function-bind-1.1.2.tgz",
|
||||
"integrity": "sha512-7XHNxH7qX9xG5mIwxkhumTox/MIRNcOgDrxWsMt2pAr23WHp6MrRlN7FBSFpCpr+oVO0F744iUgR82nJMfG2SA==",
|
||||
"license": "MIT",
|
||||
"funding": {
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/get-intrinsic": {
|
||||
"version": "1.3.0",
|
||||
"resolved": "https://registry.npmjs.org/get-intrinsic/-/get-intrinsic-1.3.0.tgz",
|
||||
"integrity": "sha512-9fSjSaos/fRIVIp+xSJlE6lfwhES7LNtKaCBIamHsjr2na1BiABJPo0mOjjz8GJDURarmCPGqaiVg5mfjb98CQ==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"call-bind-apply-helpers": "^1.0.2",
|
||||
"es-define-property": "^1.0.1",
|
||||
"es-errors": "^1.3.0",
|
||||
"es-object-atoms": "^1.1.1",
|
||||
"function-bind": "^1.1.2",
|
||||
"get-proto": "^1.0.1",
|
||||
"gopd": "^1.2.0",
|
||||
"has-symbols": "^1.1.0",
|
||||
"hasown": "^2.0.2",
|
||||
"math-intrinsics": "^1.1.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
},
|
||||
"funding": {
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/get-proto": {
|
||||
"version": "1.0.1",
|
||||
"resolved": "https://registry.npmjs.org/get-proto/-/get-proto-1.0.1.tgz",
|
||||
"integrity": "sha512-sTSfBjoXBp89JvIKIefqw7U2CCebsc74kiY6awiGogKtoSGbgjYE/G/+l9sF3MWFPNc9IcoOC4ODfKHfxFmp0g==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"dunder-proto": "^1.0.1",
|
||||
"es-object-atoms": "^1.0.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/get-stream": {
|
||||
"version": "5.2.0",
|
||||
"resolved": "https://registry.npmjs.org/get-stream/-/get-stream-5.2.0.tgz",
|
||||
@@ -1551,9 +1778,7 @@
|
||||
"version": "1.2.0",
|
||||
"resolved": "https://registry.npmjs.org/gopd/-/gopd-1.2.0.tgz",
|
||||
"integrity": "sha512-ZUKRh6/kUFoAiTAtTYPZJ3hw9wNxx+BIBOijnlG9PnrJsCcSjs1wyyD6vJpaYtgnzDrKYRSqf3OO6Rfa93xsRg==",
|
||||
"dev": true,
|
||||
"license": "MIT",
|
||||
"optional": true,
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
},
|
||||
@@ -1608,6 +1833,45 @@
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/has-symbols": {
|
||||
"version": "1.1.0",
|
||||
"resolved": "https://registry.npmjs.org/has-symbols/-/has-symbols-1.1.0.tgz",
|
||||
"integrity": "sha512-1cDNdwJ2Jaohmb3sg4OmKaMBwuC48sYni5HUw2DvsC8LjGTLK9h+eb1X6RyuOHe4hT0ULCW68iomhjUoKUqlPQ==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
},
|
||||
"funding": {
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/has-tostringtag": {
|
||||
"version": "1.0.2",
|
||||
"resolved": "https://registry.npmjs.org/has-tostringtag/-/has-tostringtag-1.0.2.tgz",
|
||||
"integrity": "sha512-NqADB8VjPFLM2V0VvHUewwwsw0ZWBaIdgo+ieHtK3hasLz4qeCRjYcqfB6AQrBggRKppKF8L52/VqdVsO47Dlw==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"has-symbols": "^1.0.3"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
},
|
||||
"funding": {
|
||||
"url": "https://github.com/sponsors/ljharb"
|
||||
}
|
||||
},
|
||||
"node_modules/hasown": {
|
||||
"version": "2.0.4",
|
||||
"resolved": "https://registry.npmjs.org/hasown/-/hasown-2.0.4.tgz",
|
||||
"integrity": "sha512-T2UbfbBEF32wiepXIsMlTW9+dDYC6wMh/t/vYA4tuOMKqWz/n3vr1NFSxQiyP+zk2mXsoMA/i/7qV6LKut1t1A==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"function-bind": "^1.1.2"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/http-cache-semantics": {
|
||||
"version": "4.2.0",
|
||||
"resolved": "https://registry.npmjs.org/http-cache-semantics/-/http-cache-semantics-4.2.0.tgz",
|
||||
@@ -1629,6 +1893,54 @@
|
||||
"node": ">=10.19.0"
|
||||
}
|
||||
},
|
||||
"node_modules/https-proxy-agent": {
|
||||
"version": "5.0.1",
|
||||
"resolved": "https://registry.npmjs.org/https-proxy-agent/-/https-proxy-agent-5.0.1.tgz",
|
||||
"integrity": "sha512-dFcAjpTQFgoLMzC2VwU+C/CbS7uRL0lWmxDITmqm7C+7F0Odmj6s9l6alZc6AELXhrnggM2CeWSXHGOdX2YtwA==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"agent-base": "6",
|
||||
"debug": "4"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 6"
|
||||
}
|
||||
},
|
||||
"node_modules/ieee754": {
|
||||
"version": "1.2.1",
|
||||
"resolved": "https://registry.npmjs.org/ieee754/-/ieee754-1.2.1.tgz",
|
||||
"integrity": "sha512-dcyqhDvX1C46lXZcVqCpK+FtMRQVdIMN6/Df5js2zouUsqG7I6sFxitIC+7KYK29KdXOLHdu9zL4sFnoVQnqaA==",
|
||||
"funding": [
|
||||
{
|
||||
"type": "github",
|
||||
"url": "https://github.com/sponsors/feross"
|
||||
},
|
||||
{
|
||||
"type": "patreon",
|
||||
"url": "https://www.patreon.com/feross"
|
||||
},
|
||||
{
|
||||
"type": "consulting",
|
||||
"url": "https://feross.org/support"
|
||||
}
|
||||
],
|
||||
"license": "BSD-3-Clause"
|
||||
},
|
||||
"node_modules/inherits": {
|
||||
"version": "2.0.4",
|
||||
"resolved": "https://registry.npmjs.org/inherits/-/inherits-2.0.4.tgz",
|
||||
"integrity": "sha512-k/vGaX4/Yla3WzyMCvTQOXYeIHvqOKtnqBduzTHpzpQZzAskKMhZ2K+EnBiSM9zGSoIFeMpXKxa4dYeZIQqewQ==",
|
||||
"license": "ISC"
|
||||
},
|
||||
"node_modules/isomorphic-ws": {
|
||||
"version": "5.0.0",
|
||||
"resolved": "https://registry.npmjs.org/isomorphic-ws/-/isomorphic-ws-5.0.0.tgz",
|
||||
"integrity": "sha512-muId7Zzn9ywDsyXgTIafTry2sV3nySZeUDe6YedVd1Hvuuep5AsIlqK+XefWpYTyJG5e503F2xIuT2lcU6rCSw==",
|
||||
"license": "MIT",
|
||||
"peerDependencies": {
|
||||
"ws": "*"
|
||||
}
|
||||
},
|
||||
"node_modules/json-buffer": {
|
||||
"version": "3.0.1",
|
||||
"resolved": "https://registry.npmjs.org/json-buffer/-/json-buffer-3.0.1.tgz",
|
||||
@@ -1705,6 +2017,36 @@
|
||||
"node": ">=10"
|
||||
}
|
||||
},
|
||||
"node_modules/math-intrinsics": {
|
||||
"version": "1.1.0",
|
||||
"resolved": "https://registry.npmjs.org/math-intrinsics/-/math-intrinsics-1.1.0.tgz",
|
||||
"integrity": "sha512-/IXtbwEk5HTPyEwyKX6hGkYXxM9nbj64B+ilVJnC/R6B0pH5G4V3b0pVbL7DBj4tkhBAppbQUlf6F6Xl9LHu1g==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">= 0.4"
|
||||
}
|
||||
},
|
||||
"node_modules/mime-db": {
|
||||
"version": "1.52.0",
|
||||
"resolved": "https://registry.npmjs.org/mime-db/-/mime-db-1.52.0.tgz",
|
||||
"integrity": "sha512-sPU4uV7dYlvtWJxwwxHD0PuihVNiE7TyAbQ5SWxDCB9mUYvOgroQOwYQQOKPJ8CIbE+1ETVlOoK1UC2nU3gYvg==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">= 0.6"
|
||||
}
|
||||
},
|
||||
"node_modules/mime-types": {
|
||||
"version": "2.1.35",
|
||||
"resolved": "https://registry.npmjs.org/mime-types/-/mime-types-2.1.35.tgz",
|
||||
"integrity": "sha512-ZDY+bPm5zTTF+YpCrAU9nK0UgICYPT0QtT1NZWFv4s++TNkcgVaT0g6+4R2uI4MjQjzysHB1zxuWL50hzaeXiw==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"mime-db": "1.52.0"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 0.6"
|
||||
}
|
||||
},
|
||||
"node_modules/mimic-response": {
|
||||
"version": "1.0.1",
|
||||
"resolved": "https://registry.npmjs.org/mimic-response/-/mimic-response-1.0.1.tgz",
|
||||
@@ -1719,9 +2061,25 @@
|
||||
"version": "2.1.3",
|
||||
"resolved": "https://registry.npmjs.org/ms/-/ms-2.1.3.tgz",
|
||||
"integrity": "sha512-6FlzubTLZG3J2a/NVCAleEhjzq5oxgHyaCU9yYXvcLsvoVaHJq/s5xXI6/XXP6tz7R9xAOtHnSO/tXtF3WRTlA==",
|
||||
"dev": true,
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/msedge-tts": {
|
||||
"version": "2.0.9",
|
||||
"resolved": "https://registry.npmjs.org/msedge-tts/-/msedge-tts-2.0.9.tgz",
|
||||
"integrity": "sha512-Qw9pphrz/GTGee+XbXrmwXkDWsR6BVdKyDJKDZBC1RnmBMrZJ8VfMTC44j0v9dIptmLMkmLOuBtQX6LHuodcDw==",
|
||||
"hasInstallScript": true,
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"axios": "^1.11.0",
|
||||
"buffer": "^6.0.3",
|
||||
"isomorphic-ws": "^5.0.0",
|
||||
"stream-browserify": "^3.0.0",
|
||||
"ws": "^8.21.3"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">=16.0.0"
|
||||
}
|
||||
},
|
||||
"node_modules/nanoid": {
|
||||
"version": "3.3.15",
|
||||
"resolved": "https://registry.npmjs.org/nanoid/-/nanoid-3.3.15.tgz",
|
||||
@@ -1855,6 +2213,15 @@
|
||||
"node": ">=0.4.0"
|
||||
}
|
||||
},
|
||||
"node_modules/proxy-from-env": {
|
||||
"version": "2.1.0",
|
||||
"resolved": "https://registry.npmjs.org/proxy-from-env/-/proxy-from-env-2.1.0.tgz",
|
||||
"integrity": "sha512-cJ+oHTW1VAEa8cJslgmUZrc+sjRKgAKl3Zyse6+PV38hZe/V6Z14TbCuXcan9F9ghlz4QrFr2c92TNF82UkYHA==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">=10"
|
||||
}
|
||||
},
|
||||
"node_modules/pump": {
|
||||
"version": "3.0.4",
|
||||
"resolved": "https://registry.npmjs.org/pump/-/pump-3.0.4.tgz",
|
||||
@@ -1879,6 +2246,20 @@
|
||||
"url": "https://github.com/sponsors/sindresorhus"
|
||||
}
|
||||
},
|
||||
"node_modules/readable-stream": {
|
||||
"version": "3.6.2",
|
||||
"resolved": "https://registry.npmjs.org/readable-stream/-/readable-stream-3.6.2.tgz",
|
||||
"integrity": "sha512-9u/sniCrY3D5WdsERHzHE4G2YCXqoG5FTHUiCC4SIbr6XcLZBY05ya9EKjYek9O5xOAwjGq+1JdGBAS7Q9ScoA==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"inherits": "^2.0.3",
|
||||
"string_decoder": "^1.1.1",
|
||||
"util-deprecate": "^1.0.1"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">= 6"
|
||||
}
|
||||
},
|
||||
"node_modules/resolve-alpn": {
|
||||
"version": "1.2.1",
|
||||
"resolved": "https://registry.npmjs.org/resolve-alpn/-/resolve-alpn-1.2.1.tgz",
|
||||
@@ -1963,6 +2344,26 @@
|
||||
"fsevents": "~2.3.2"
|
||||
}
|
||||
},
|
||||
"node_modules/safe-buffer": {
|
||||
"version": "5.2.1",
|
||||
"resolved": "https://registry.npmjs.org/safe-buffer/-/safe-buffer-5.2.1.tgz",
|
||||
"integrity": "sha512-rp3So07KcdmmKbGvgaNxQSJr7bGVSVk5S9Eq1F+ppbRo70+YeaDxkw5Dd8NPN+GD6bjnYm2VuPuCXmpuYvmCXQ==",
|
||||
"funding": [
|
||||
{
|
||||
"type": "github",
|
||||
"url": "https://github.com/sponsors/feross"
|
||||
},
|
||||
{
|
||||
"type": "patreon",
|
||||
"url": "https://www.patreon.com/feross"
|
||||
},
|
||||
{
|
||||
"type": "consulting",
|
||||
"url": "https://feross.org/support"
|
||||
}
|
||||
],
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/semver": {
|
||||
"version": "6.3.1",
|
||||
"resolved": "https://registry.npmjs.org/semver/-/semver-6.3.1.tgz",
|
||||
@@ -2037,6 +2438,25 @@
|
||||
"dev": true,
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/stream-browserify": {
|
||||
"version": "3.0.0",
|
||||
"resolved": "https://registry.npmjs.org/stream-browserify/-/stream-browserify-3.0.0.tgz",
|
||||
"integrity": "sha512-H73RAHsVBapbim0tU2JwwOiXUj+fikfiaoYAKHF3VJfA0pe2BCzkhAHBlLG6REzE+2WNZcxOXjK7lkso+9euLA==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"inherits": "~2.0.4",
|
||||
"readable-stream": "^3.5.0"
|
||||
}
|
||||
},
|
||||
"node_modules/string_decoder": {
|
||||
"version": "1.3.0",
|
||||
"resolved": "https://registry.npmjs.org/string_decoder/-/string_decoder-1.3.0.tgz",
|
||||
"integrity": "sha512-hkRX8U1WjJFd8LsDJ2yQ/wWWxaopEsABU1XfkM8A+j0+85JAGppt16cr1Whg6KIbb4okU6Mql6BOj+uup/wKeA==",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"safe-buffer": "~5.2.0"
|
||||
}
|
||||
},
|
||||
"node_modules/sumchecker": {
|
||||
"version": "3.0.1",
|
||||
"resolved": "https://registry.npmjs.org/sumchecker/-/sumchecker-3.0.1.tgz",
|
||||
@@ -2158,6 +2578,12 @@
|
||||
"node": ">= 4.0.0"
|
||||
}
|
||||
},
|
||||
"node_modules/util-deprecate": {
|
||||
"version": "1.0.2",
|
||||
"resolved": "https://registry.npmjs.org/util-deprecate/-/util-deprecate-1.0.2.tgz",
|
||||
"integrity": "sha512-EPD5q1uXyFxJpCrLnCc1nHnq3gOa6DZBocAIiI2TaSCA7VCJ1UJDMagCzIkXNsUYfD1daK//LTEQ8xiIbrHtcw==",
|
||||
"license": "MIT"
|
||||
},
|
||||
"node_modules/vite": {
|
||||
"version": "5.4.21",
|
||||
"resolved": "https://registry.npmjs.org/vite/-/vite-5.4.21.tgz",
|
||||
@@ -2761,6 +3187,27 @@
|
||||
"dev": true,
|
||||
"license": "ISC"
|
||||
},
|
||||
"node_modules/ws": {
|
||||
"version": "8.22.0",
|
||||
"resolved": "https://registry.npmjs.org/ws/-/ws-8.22.0.tgz",
|
||||
"integrity": "sha512-Ydggc987+RO0AnWtZ/7Wq9FtNvcrL1b/RO0ud9mWjUPgDrsAAwQSF51sm2hm1XofbU/4jkpGEsLFsZZxU+1DOg==",
|
||||
"license": "MIT",
|
||||
"engines": {
|
||||
"node": ">=10.0.0"
|
||||
},
|
||||
"peerDependencies": {
|
||||
"bufferutil": "^4.0.1",
|
||||
"utf-8-validate": ">=5.0.2"
|
||||
},
|
||||
"peerDependenciesMeta": {
|
||||
"bufferutil": {
|
||||
"optional": true
|
||||
},
|
||||
"utf-8-validate": {
|
||||
"optional": true
|
||||
}
|
||||
}
|
||||
},
|
||||
"node_modules/yaml": {
|
||||
"version": "2.9.0",
|
||||
"resolved": "https://registry.npmjs.org/yaml/-/yaml-2.9.0.tgz",
|
||||
|
||||
@@ -25,6 +25,7 @@
|
||||
"vitest": "^2.1.9"
|
||||
},
|
||||
"dependencies": {
|
||||
"msedge-tts": "^2.0.9",
|
||||
"yaml": "^2.9.0"
|
||||
}
|
||||
}
|
||||
|
||||
@@ -7,42 +7,74 @@ name: 美学纲领与交互范式
|
||||
id: aesthetics-interaction
|
||||
artifact: 设计.美学纲领与交互范式
|
||||
declaration: >
|
||||
提取用户最想反复获得的核心体验，以及让它成立的背景与规则内核；同时钉清参与和停笔边界。
|
||||
提取用户最想反复获得的核心体验与让它成立的承重内核，写成给后续能力读的美学纲领；
|
||||
同时钉清交互范式：人称视角、模型可代写什么、用户保留什么、后果由谁裁定、何时停笔。
|
||||
when: |
|
||||
用户要的是叙事、角色扮演或持续互动，但还不能明确说出：最想反复获得什么体验，以及什么承重前提让它成立。
|
||||
用户要的是叙事、角色扮演或持续互动，但还不能明确说出：最想反复获得什么体验、自己站在什么位置、轮转到哪必须停等用户。
|
||||
用户主动改变核心体验、代入站位或关键权限时也应重写。
|
||||
when_not: |
|
||||
本局不依赖体验契约的纯转换、资料整理或机械格式任务，可不使用。
|
||||
只改遣词和具体写法 → 叙事指南；只改推进与输入用法 → 故事推进；只改舞台详情或硬法则 → 对应能力。
|
||||
boundary: |
|
||||
本能力只交核心体验、体验成立的背景与规则内核、参与边界。
|
||||
背景与规则只提取承重内核，不扩写百科；舞台细节归舞台骨架，可执行硬法则归实现机制。
|
||||
本能力只交三块：美学纲领（体验是什么、怎样发生、呈现尺度）、背景与规则的承重内核、交互范式（视角、描写权限、后果控制、停笔）。
|
||||
背景与规则只提取承重内核，不扩写百科；舞台细节归舞台骨架，可执行硬法则归实现机制；具体遣词句式归叙事指南。
|
||||
feeds: design_only
|
||||
```
|
||||
|
||||
## opening
|
||||
|
||||
```opening
|
||||
先说你最想在这局里反复获得什么体验，以及你希望自己站在什么位置。可以直接给一个场景、关系或感觉，不必先整理设定。
|
||||
先用几句话抓住你要的世界和体验核心（想到什么写什么，不必整齐）：
|
||||
|
||||
1. 原型世界是什么？变在哪里？
|
||||
例：丧尸世界，但只有我不会被感染；现代都市，但灵异是真的；也可以只说「现代都市」，变点稍后补。
|
||||
|
||||
2. 你代入吗？希望自己站在什么位置？
|
||||
· 完全代入（我就是那个人，TA 的遭遇≈我的遭遇）
|
||||
· 部分代入 / 跟着体验（有距离）
|
||||
· 操控或旁观（角色是棋子/观察对象）
|
||||
· 其它（请写明）
|
||||
|
||||
3. 你最想在这局里反复获得什么体验？
|
||||
一句话即可——权力、求生紧张、关系拉扯、解谜、荣耀、禁忌刺激……用你自己的词。
|
||||
```
|
||||
|
||||
## task
|
||||
|
||||
```task
|
||||
从用户材料中提取一份体验契约，写入 `设计.美学纲领与交互范式`。产物只回答三件事：用户最想反复获得什么核心体验；什么最小背景或规则内核让它成立；模型与用户各自能决定什么、何时必须停笔。
|
||||
从用户材料中提取一份体验契约，写入 `设计.美学纲领与交互范式`。它是后续所有能力取舍的依据，也会挂给主世界层与转述层直接阅读。产物回答三件事：
|
||||
|
||||
保留用户自己的重点和尺度。材料不足时少写，把真正影响核心体验或背景内核的缺口交给 `self_score.gap`；产物正文不放评分、诊断、追问或开放问题。
|
||||
1. 美学纲领：用户最想反复获得什么体验，它通过什么关系、处境、感官或选择反复发生，呈现时的氛围与尺度。
|
||||
2. 背景与规则：删掉就会让这份体验不成立的最小承重前提，以及它如何约束人物、选择与后果。
|
||||
3. 交互范式：用户与所扮演角色的关系、人称与视角、模型对 user 每类行为能代写到哪一步、世界后果由谁裁定、何时必须停笔等用户。
|
||||
|
||||
保留用户自己的重点、用词和尺度。材料不足时少写或写「未定」，把真正影响体验或权限的缺口交给 `self_score.gap`；产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
|
||||
## principles
|
||||
|
||||
```principles
|
||||
1. 核心体验不是题材、剧情简介或情绪标签；要写用户如何参与、体验怎样反复发生。
|
||||
2. 背景与规则寻找最小承重内核：可以是关键关系、社会前提、世界规则或作用机制。删掉它会使核心体验不成立。
|
||||
3. 角色人文关系、描写焦点和文风信号可作为核心体验的组成材料，但不为评分硬拆成独立维度。
|
||||
4. 本步不扩写世界百科，不制定完整硬机制，不细化遣词句式，不决定故事推进方法。
|
||||
5. 参与边界写可执行权限：模型可补哪些反应与行动，哪些决定必须留给用户，何时停笔。
|
||||
6. 用户改口导致体验内核改变时整份重写，不在旧契约后追加互相冲突的补丁。
|
||||
1. 核心是用户想要的体验；其余字段都是为了让它稳定地反复发生。完善设定时回头核对原本的核心体验，不让它被设定冲淡。
|
||||
2. 核心体验不是题材、剧情简介或情绪标签：写用户如何参与、体验怎样反复发生。同一设定不同用户要的可以完全不同（角斗士：荣耀 / 自由 / 血与酒），找到这个人要的那一个，而不是补全世界观。
|
||||
3. 变造世界先抓「原型 + 变在哪里」；核心体验往往来自变点。背景与规则只写承重内核：关键关系、社会前提、世界规则或作用机制。
|
||||
4. 用户常说不清想要什么：展示优于提问，大胆优于保守；用两种站位或两种感觉的短场景让用户选或改；负面信号同样有效。用户可以说不，随时准备整份改写。
|
||||
5. 用户要的内容去道德化、常态化处理：按用户给的尺度写进「边界与尺度」，写清服务体验的用法与真正的禁忌，不替用户加或减尺度。
|
||||
6. 交互范式写可执行权限，不写口号。描写权限按行为类别逐项写：哪些模型可直接代写、哪些只写到起势就停、哪些必须留给用户；「用户输入后」写模型对一句输入是只确认、补细节还是连带推出后果。
|
||||
7. 交互范式要保护美学纲领：权限、视角和停笔点的设置要让核心体验更容易发生，两者矛盾时以美学纲领为准改范式。
|
||||
8. 本步不扩写世界百科，不制定完整硬机制，不细化遣词句式，不决定故事推进方法，不输出执行单元列表或插入 order。
|
||||
9. 用户改口导致体验内核改变时整份重写，不在旧契约后追加互相冲突的补丁。
|
||||
```
|
||||
|
||||
## probe
|
||||
|
||||
```probe
|
||||
本步最先要钉清三件：站位（代入还是旁观、镜头看哪、爽从哪来）、核心感觉（反复感受到什么）、轮转停点（何时等用户）。三件写进 `体验锚定`；开场白写入前程序会检查它们。
|
||||
|
||||
用户首句、默认问题首答或黑板「用户.需求」里已有的直接采纳，已答清的不重问。
|
||||
|
||||
缺其中一件：回复里只问这一件，给 2～3 个带短场景的建议选项，不问开放式「你还想要什么」。三件已清：再挑一个最影响体验的薄弱点（视角、某类描写权限、尺度）给选项；都够用就不再追问。
|
||||
|
||||
用户讲到人物、舞台、规则细节：只抽取能改变体验、站位或权限的那一句写进正文，其余留给对应能力，本步不追问细节。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -53,29 +85,58 @@ feeds: design_only
|
||||
"技能": "美学纲领与交互范式",
|
||||
"brief": "一句话点出用户最想反复获得的核心体验",
|
||||
"体验锚定": {
|
||||
"站位": "用户与当前角色/作品的关系",
|
||||
"站位": "用户与所扮演角色/作品的关系：代入、部分代入、旁观或操控",
|
||||
"核心感觉": "最想反复获得的体验",
|
||||
"轮转停点": "何时必须停笔等待用户"
|
||||
},
|
||||
"mount": ["world-simulator", "narrator"],
|
||||
"稳变": "stable",
|
||||
"正文": {
|
||||
"核心体验": {
|
||||
"体验主张": "用户最想反复获得什么",
|
||||
"发生方式": "它通过哪些关系、处境、感官或选择反复发生",
|
||||
"美学纲领": {
|
||||
"体验内核": "给后续能力和游玩层读的纲领正文，可多段：用户要反复感到什么、它靠什么关系或处境成立、从哪一刻开始、怎样被反复兑现",
|
||||
"发生方式": "体验通过哪些关系、处境、感官或选择反复发生",
|
||||
"角色人文": "关键人物关系中的权力、尊严、亲疏或人性张力；无则省略",
|
||||
"描写信号": "应把注意力放在哪里、呈现什么质感；具体写法交给叙事指南"
|
||||
"呈现要点": {
|
||||
"氛围基调": "整体质感与情绪温度",
|
||||
"描写焦点": "注意力放在哪里：哪些瞬间、反应、细节值得放慢写；具体遣词交给叙事指南",
|
||||
"边界与尺度": "色情/暴力/血腥等按用户给的尺度：服务体验的用法与真正的禁忌",
|
||||
"其它要点": ["反差、日常化、节奏偏好等短条；无则省略"]
|
||||
}
|
||||
},
|
||||
"背景与规则": {
|
||||
"原型与变点": "熟悉的原型世界，以及变在哪里；无变造则省略",
|
||||
"核心内核": "让体验成立的最小承重前提",
|
||||
"作用方式": "它如何约束人物、选择与后果",
|
||||
"作用方式": "它如何约束人物、选择与后果，怎样持续制造该体验",
|
||||
"非必要设定": "明确本步不展开的装饰信息；无则省略"
|
||||
},
|
||||
"参与边界": {
|
||||
"用户站位": "代入、部分代入、旁观、操控或本局自定义描述",
|
||||
"模型可代演": "可补写的行动、反应和既有习惯",
|
||||
"用户保留权": "不得替用户决定的事项",
|
||||
"世界反馈": "模型可直接裁定和推进的外部后果",
|
||||
"交互范式": {
|
||||
"前置配置": {
|
||||
"用户与user关系": "完全代入 / 部分代入 / 旁观 / 操控，及其含义",
|
||||
"人称": "如：第二人称称「你」",
|
||||
"元叙事": "如：不跳出故事直接问卷，决策用情节呈现"
|
||||
},
|
||||
"叙事视角": {
|
||||
"主要视角": "镜头跟谁、看什么",
|
||||
"视角切换": "禁止，或允许的例外",
|
||||
"可写内心": "可以写谁的内心；如仅 user 或仅 NPC 外显",
|
||||
"不呈现的信息": "user 不该知道、正文不能透露的内容"
|
||||
},
|
||||
"描写权限": {
|
||||
"思考与抉择": "关键决定是否只能由用户做，模型写到哪里停",
|
||||
"言语": "模型能否替 user 说话、说到什么程度",
|
||||
"主动行动": "模型能否替 user 发起行动",
|
||||
"情绪与身体反应": "模型可补写的生理与情绪反应",
|
||||
"习惯与既有技能": "模型可按既有设定自动执行的部分",
|
||||
"外在状态": "外貌、伤势、衣着等可否直接描写",
|
||||
"用户输入后": "对用户一句输入：只确认 / 补细节 / 连带推出后果"
|
||||
},
|
||||
"后果与叙事控制": {
|
||||
"行为后果判定": "谁裁定 user 行为的成败与代价",
|
||||
"世界对user的影响": "NPC 与世界如何主动作用于 user",
|
||||
"时间与场景": "模型可否推进时间、切换场景，跨度多大",
|
||||
"引入新内容": "模型可否主动引入新人物、事件、威胁",
|
||||
"设计困境": "是否主动给 user 制造两难；怎样的两难服务体验"
|
||||
},
|
||||
"停笔时机": "必须等用户输入的明确时机"
|
||||
}
|
||||
}
|
||||
@@ -83,19 +144,24 @@ feeds: design_only
|
||||
```
|
||||
|
||||
硬规则：
|
||||
1. 输出合法 JSON；保留 `体验锚定` 仅为兼容，不做计数门槛。
|
||||
2. `正文` 只含 `核心体验`、`背景与规则`、`参与边界`；未知字段省略，不用“完备度”或诊断对象占位。
|
||||
3. 不在产物正文写 `自评`、`追问`、`开放问题`、评分依据或用户回复。
|
||||
4. `mount` 固定包含 `world-simulator` 与 `narrator`；后续投影器分别裁剪字段。
|
||||
1. 输出合法 JSON。`体验锚定` 三项必填短句：开场白写入前程序按它判定体验是否已锚定，缺项会拒绝写入开场白。
|
||||
2. `正文` 只含 `美学纲领`、`背景与规则`、`交互范式` 三块，每块是对象，不写成一段散文。未知字段写「未定」或省略，不编造百科。
|
||||
3. 不在产物里写 `自评`、`追问`、`开放问题`、完备度、评分依据或用户回复；缺口交给 `self_score.gap`。
|
||||
4. `mount` 固定包含 `world-simulator` 与 `narrator`；不写数字 `order`，不输出执行单元列表。
|
||||
|
||||
## score
|
||||
|
||||
```score
|
||||
核心体验：
|
||||
- 5：准确提取用户最想反复获得的体验，写清用户站位、体验如何发生及何时停笔；后续能力可据此取舍。
|
||||
- 3：大方向正确，但仍偏题材或情绪概括，或体验如何反复发生尚不明确。
|
||||
美学纲领：
|
||||
- 5：准确提取用户最想反复获得的体验，写清它怎样反复发生、从哪一刻成立，呈现焦点与尺度贴合用户原话；后续能力可据此取舍。
|
||||
- 3：大方向正确，但仍偏题材或情绪概括，或体验如何反复发生、尺度如何把握尚不明确。
|
||||
- 1：误读用户重点，只有剧情简介、设定标签或泛化氛围。
|
||||
|
||||
交互范式：
|
||||
- 5：站位、人称视角、逐类描写权限、后果裁定与停笔时机都可执行，且在保护核心体验。
|
||||
- 3：有站位与停笔，但描写权限笼统，或某类行为（言语、行动、反应）谁来写仍不清。
|
||||
- 1：只有口号式描述，或权限设置与核心体验矛盾。
|
||||
|
||||
背景与规则：
|
||||
- 5：找到让核心体验成立的最小内核——关键关系、背景前提或作用规则，并说明它怎样持续制造该体验。
|
||||
- 3：找到相关前提，但内核过宽、因果不清，或混有较多装饰设定。
|
||||
@@ -105,9 +171,23 @@ feeds: design_only
|
||||
## checklist
|
||||
|
||||
```checklist
|
||||
- [ ] 核心体验写的是用户反复获得什么，而非题材或剧情简介？
|
||||
- [ ] `体验锚定` 三项（站位、核心感觉、轮转停点）都已写出？
|
||||
- [ ] 美学纲领写的是用户反复获得什么、怎样发生，而非题材或剧情简介？体验内核能直接给游玩层读？
|
||||
- [ ] 背景与规则是删掉就会破坏体验的最小内核？
|
||||
- [ ] 参与边界明确模型可代演、用户保留权和停笔时机？
|
||||
- [ ] 交互范式逐类写清描写权限、后果裁定和停笔时机，并且在保护美学纲领？
|
||||
- [ ] 未扩写舞台百科、硬机制、具体文风或推进方案？
|
||||
- [ ] 产物正文不含评分、诊断、追问和开放问题？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：
|
||||
- brief 点题「垄断生存资源的绝对掌控感」；美学纲领的体验内核写清掌控感从哪一刻成立、靠什么反复兑现，边界与尺度照用户原话；交互范式写明 user 的抉择只由用户做、模型可补身体反应与习惯动作、NPC 的屈服与反抗由模型裁定、每次 NPC 给出回应后停笔。
|
||||
|
||||
坏：
|
||||
- 只有一段散文纲领，三块没有拆成对象。
|
||||
- 百科式背景堆砌，看不出体验靠哪条前提成立。
|
||||
- 交互范式只写「尊重玩家选择」，没说言语、行动、反应各由谁写。
|
||||
- 在正文里写评分、追问或「待确认」清单。
|
||||
```
|
||||
|
||||
@@ -58,7 +58,7 @@ modules:
|
||||
mount: [world-simulator]
|
||||
declaration: >-
|
||||
钉清 @玩家 扮演的那个人：名字、背景、特殊设定（身份/金手指等）；
|
||||
自动写入时加性格。代入有名片时排本步；可 设计.主角设定#短码 与开场白同槽绑定；禁止把名字/简介写进开场白正文。
|
||||
自动写入时加性格。代入有名片时排本步，写各开场共通的部分；各开场不同的玩家角色写在开场白的 meta.用户角色，由程序同步为同短码主角设定；禁止把名字/简介写进开场白正文。
|
||||
artifact: 设计.主角设定
|
||||
|
||||
# —— 世界与内容 ——
|
||||
@@ -141,7 +141,7 @@ modules:
|
||||
mount: [world-simulator, narrator]
|
||||
declaration: >-
|
||||
从偏好库选取约束，只选定一种用户输入用法，
|
||||
并写清扩写/续写定义、比例与 OOC 处理；不写遣词
|
||||
并写清每一轮里扩写（演本轮材料）与续写（材料后往下写）的定义、比例与 OOC 处理；不写遣词
|
||||
artifact: 设计.故事推进
|
||||
|
||||
# —— 机制与数据 ——
|
||||
@@ -243,7 +243,7 @@ modules:
|
||||
declaration: >-
|
||||
普通楼层创作终节点，且可增殖：每条开场各占一个节点（设计.开场白#场景短码）。
|
||||
默认 1 条；还要入口就再增殖。全部保存。游玩时左右切换，选中的一条作为 0 层
|
||||
artifact: 设计.开场白与开场变量
|
||||
artifact: 设计.开场白
|
||||
|
||||
- id: zero-layer-opening-setup
|
||||
name: 0层开场白与初态
|
||||
|
||||
@@ -35,6 +35,8 @@ feeds: gm
|
||||
|
||||
有生成规则时：认定规则的中文名，按规则的字段、池或提示、填写工作流生成这一条；用户已写明的专名与结果只扩写不改写，未写字段从已有材料往下补满。无规则且只是一条字段很少的具名内容时：直接写用户给出的字段。
|
||||
|
||||
规则的一次生成为同批时：一次 insert 写齐本批，`记录` 写成数组，每项一条记录并带上规则里的姓名或名称字段，程序按条拆成多份。先定各条如何互指、对立、同场，再填各条字段。
|
||||
|
||||
材料足够便直接成稿。缺的是合同本身时，先回到生成规则，不在这里补合同。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
|
||||
@@ -44,10 +46,13 @@ feeds: gm
|
||||
1. 有规则：字段名、取值方式、工作流阶段与依赖都以规则为准；不增删字段，不改池。
|
||||
2. 池选取按合理度选择，不是程序抽签；提示约束按那句话的范围写开。
|
||||
3. 按填写工作流推进：先骨架后血肉；阶段依赖与自检能核对的，生成后要能过。
|
||||
4. 一份产物一条记录。多条各自成稿，用中文名称区分。
|
||||
5. 无规则只写用户点名的那一条，字段保持很少；不借机发明完整人设合同。
|
||||
6. 用户已写明的专名、关系、结果保持原样，只做厚度扩写；未写字段从这些材料延伸。
|
||||
7. 给人读的键和专名用中文。材料里的英文名音译成中文，不把英文原名留在字段值里。外壳的 schema、brief、mount 保持既有键。
|
||||
4. 一份产物一条记录。单条规则一次写一条；多条各自成稿，用中文名称区分。
|
||||
5. 同批规则一次写齐本批：先对齐关系，再填字段；各条里互指的专名与关系必须一致（甲记乙为仇敌，乙也记甲，且说的是同一件事）。有关系的人在同一次里写完，不拆到不同轮。
|
||||
6. 句子级字段写到规则 `地板` 规定的可感知程度；词条、档位、数字可以短。句子级字段只丢一个词不算填完。
|
||||
7. 分级字段只填档位短值（如「金丹」）；表象留在规则的分级表现里，不复制进每条记录。
|
||||
8. 无规则只写用户点名的那一条，字段保持很少；不借机发明完整人设合同。
|
||||
9. 用户已写明的专名、关系、结果保持原样，只做厚度扩写；未写字段从这些材料延伸。
|
||||
10. 给人读的键和专名用中文。材料里的英文名音译成中文，不把英文原名留在字段值里。外壳的 schema、brief、mount 保持既有键。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -70,7 +75,7 @@ feeds: gm
|
||||
```
|
||||
|
||||
硬规则：
|
||||
1. 输出合法 JSON。`记录` 是单个对象。`名称` 用中文，并与这条记录的主名称一致。
|
||||
1. 输出合法 JSON。`记录` 是单个对象；规则为同批时写成至少 2 项的对象数组，每项带姓名或名称。`名称` 用中文，并与这条记录的主名称一致。
|
||||
2. 有规则时 `记录` 的键必须来自该规则字段，不得增删。
|
||||
3. 无规则时省略 `规则`，`记录` 只含用户给出的少量简单字段。
|
||||
4. 字段名和专名用中文；英文专名音译为中文。
|
||||
@@ -81,9 +86,9 @@ feeds: gm
|
||||
|
||||
```score
|
||||
工作流符合度：
|
||||
- 5：严格按规则填写工作流推进；字段、池选取、依赖与自检都落实；无规则路径则字段简单够用且不越权发明合同。
|
||||
- 3：大体按工作流，但有一处阶段跳跃、依赖未咬合，或某字段明显偏薄。
|
||||
- 1：忽略工作流；乱增字段；或把用户已定内容改掉。
|
||||
- 5：严格按规则填写工作流推进；字段、池选取、依赖与自检都落实；句子级字段达到地板；同批各条互指一致；无规则路径则字段简单够用且不越权发明合同。
|
||||
- 3：大体按工作流，但有一处阶段跳跃、依赖未咬合，或某句子级字段明显低于地板。
|
||||
- 1：忽略工作流；乱增字段；句子级只填一个词；同批互指对不上；或把用户已定内容改掉。
|
||||
```
|
||||
|
||||
## checklist
|
||||
@@ -91,7 +96,18 @@ feeds: gm
|
||||
```checklist
|
||||
- [ ] 有规则则按字段与填写工作流执行；无规则则只是一条字段很少的具名内容？
|
||||
- [ ] 一份产物只有一条记录，名称是中文？
|
||||
- [ ] 同批规则已一次写齐，各条互指的专名与关系一致？
|
||||
- [ ] 句子级字段达到地板？分级字段只填档位短值？
|
||||
- [ ] 记录键值合法，用户已定专名未改写，英文专名已音译？
|
||||
- [ ] 未改合同？
|
||||
- [ ] 产物正文不含评分、诊断、追问？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：规则一次生成=同批、关系=开局仇敌二人 → 一次写齐 2 条，甲的仇敌字段写乙与夺位之事，乙的写甲与同一件事，两条对得上。
|
||||
坏：记录里加了规则没有的字段（规则无「性别」却写进去）。
|
||||
坏：规则要求句子级的「气质」只填「阴郁」，没达到地板。
|
||||
坏：同批的仇敌二人分两轮各写各的，互指的名字和事由对不上。
|
||||
```
|
||||
|
||||
@@ -31,7 +31,7 @@ feeds: gm,auditor
|
||||
## task
|
||||
|
||||
```task
|
||||
为单一对象族写出一份生成合同，写入 `设计.生成规则#中文规则名`。先写名称与用途，再按用途推出字段与每项填充要求；需要封闭选取的字段绑池，更适合一句话约束的字段写提示；最后写填写工作流（含依赖、自检、反模式），供具体实例与游玩期动态生成共用。
|
||||
为单一对象族写出一份生成合同，写入 `设计.生成规则#中文规则名`。先写名称与用途，再按用途推出字段与每项填充要求；描写类字段写明地板，标签看不懂的分级另写分级表现；需要封闭选取的字段绑池，更适合一句话约束的字段写提示；最后写填写工作流（含依赖、自检、反模式），供具体实例与游玩期动态生成共用。
|
||||
|
||||
每个产物只含一条规则。材料足够便直接成稿；真正会改合同的缺口（生命周期、必填字段、单条/同批关系）交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -39,15 +39,20 @@ feeds: gm,auditor
|
||||
## principles
|
||||
|
||||
```principles
|
||||
1. 用途决定字段：每个字段都要能回答「对这个用途是否必要」；女 NPC 不写性别；药水要写清效果参数，不要只写外观。
|
||||
1. 用途决定字段：每个字段都要能回答「对这个用途是否必要」；女 NPC 不写性别；药水要写清效果参数，不要只写外观；怪物这类要接战斗的，等级、攻击骰等玩法接口不能漏。
|
||||
2. 字段不多不少、互不重叠：资产与存款这类同义字段只留一个。
|
||||
3. 取值分三类：自由生成；从池中选取（填写方按合理度选，不是伪随机程序）；提示约束（如「现代都市中正常的职业」）。默认优先提示约束；只有需要封闭一致或分阶段选型时才建池。
|
||||
4. 填写工作流按重要性推进：先锚定会约束后续的选项，再填血肉；阶段、依赖、自检、反模式都服务于「怎么填对」，写短写准。
|
||||
5. 生命周期三选一：`仅游玩期`（创作期不预填）、`只预生成`（预填完即可丢掉规则）、`预生成并游玩期`（既写种子，开场后仍可再生成）。选哪一档不改变「要不要先写这份合同」。
|
||||
6. `只预生成` 为中间层；另两类为最终层，挂主世界层与旁观。
|
||||
7. 上游舞台/机制只短引用；不在本步发明共通前提百科。
|
||||
8. 不输出最终实例记录。
|
||||
9. 给人读的键和专名用中文。材料里的英文名音译成中文。外壳的 schema、brief、mount 保持既有键。
|
||||
3. 取值分三类：自由生成；从池中选取（填写方按合理度选，不是伪随机程序）；提示约束（如「现代都市中正常的职业」）。默认优先提示约束；只有能数完、必须全员一致的封闭集合（阶层、强度档、武器种类）或分阶段选型时才建池。
|
||||
4. 姓名、绰号、能力概念、外貌、气质写不完：用自由或提示约束，不建池，也不在填充要求或提示里列一串可选值冒充穷举。要反复生成又怕雷同时，`提示` 可写 2～6 个「方向名：一句概括」划多样的范围；方向不挂作品名、不堆例句。
|
||||
5. 描写类字段（外貌、能力现象、限制、互动手感等）写 `地板`：最低要写到读者能看见、听见或碰到什么，对手能抓住什么空隙。词条、档位、数字、骰子可以短，不写地板。规则里出现的示例必须达到地板，不用「气质：阴郁」「能力：时空斩」这种单标签当示例。
|
||||
6. 标签本身看不懂的分级（境界、品阶、强度档、好感档）另写 `分级表现`：每档一句概括 + 一段可感知表象，相邻档必须能分清。实例字段只填档位短值，叙事按此表展开。
|
||||
7. 外貌、服装、仪式等几面拼成一格的字段，在该字段 `填充要求` 里写配方：用哪几面、每面写什么、怎么拼成一段。字段仍是一个扁平值，不拆嵌套对象，也不为每一面各建池。
|
||||
8. 填写工作流按重要性推进：先锚定会约束后续的选项，再填血肉；阶段、依赖、自检、反模式都服务于「怎么填对」，写短写准。会反复生成的规则，自检清单要含去重：与已有记录不撞名、不撞核心特征。
|
||||
9. 生命周期三选一：`仅游玩期`（创作期不预填）、`只预生成`（预填完即可丢掉规则）、`预生成并游玩期`（既写种子，开场后仍可再生成）。选哪一档不改变「要不要先写这份合同」。
|
||||
10. `只预生成` 为中间层；另两类为最终层，挂主世界层与旁观。
|
||||
11. `数量` 与 `一次生成` 是两件事：`数量` 是总共几条；`一次生成` 看记录之间有没有必须一起钉死的关系（互指、对立、同场），不看总数。有 → 同批，`关系` 只写谁对谁；彼此独立 → 单条。「总共 5 个」不等于同批。
|
||||
12. 多张名片共享的台上格局、世界状况、既成法则归舞台骨架或实现机制：已有则在 `用途` 或 `填充要求` 里短引用；没有则只写填写时必须守的最短条款，缺口交 `self_score.gap`。不在规则里重抄「共通前提」长文，也不借 `关系` 写前史与社会律。
|
||||
13. 不输出最终实例记录。
|
||||
14. 给人读的键和专名用中文。材料里的英文名音译成中文。外壳的 schema、brief、mount 保持既有键。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -78,10 +83,11 @@ feeds: gm,auditor
|
||||
"字段": [
|
||||
{
|
||||
"名": "顶层字段名",
|
||||
"填充要求": "填什么、填到多厚、为何本用途需要它",
|
||||
"填充要求": "填什么、填到多厚、为何本用途需要它；几面拼一格的写配方（哪几面、各写什么）",
|
||||
"取值方式": "自由|池选取|提示约束",
|
||||
"绑定池": "池选取时填写池名称；其它省略",
|
||||
"提示": "提示约束时写可执行的一句范围；其它省略"
|
||||
"提示": "提示约束时写可执行的一句范围，或 2～6 个「方向名：一句概括」；其它省略",
|
||||
"地板": "描写类字段写：最低填到什么可感知程度；词条、档位、数字省略"
|
||||
}
|
||||
],
|
||||
"池": [
|
||||
@@ -91,6 +97,18 @@ feeds: gm,auditor
|
||||
"条目": ["封闭可选项；只列必须一致的有限集合"]
|
||||
}
|
||||
],
|
||||
"分级表现": [
|
||||
{
|
||||
"字段": "标签看不懂的分级字段名；无此类分级则省略整个键",
|
||||
"档位": [
|
||||
{
|
||||
"值": "档位短值，与实例所填一致",
|
||||
"概括": "一句：这档能做到什么量级",
|
||||
"表象": "看见、听见、碰到什么；须能和上下档分清"
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
"填写工作流": {
|
||||
"工作流": {
|
||||
"阶段1 骨架锚定": "先选哪些池/字段、定下什么内核",
|
||||
@@ -119,29 +137,30 @@ feeds: gm,auditor
|
||||
硬规则：
|
||||
1. 输出合法 JSON；每个产物只描述一条规则，禁止 `rules` 数组堆多条。
|
||||
2. `字段` 由用途推出；可省略的字段不写；重叠字段合并。
|
||||
3. `池` 可 `[]`；池选取字段必须能对应到某个池；提示约束字段必须有 `提示`。
|
||||
3. `池` 可 `[]`；池选取字段必须能对应到某个池；提示约束字段必须有 `提示`。姓名、绰号、能力概念、外貌、气质不绑池。
|
||||
4. `填写工作流` 必含 `工作流`；`依赖关系` / `自检清单` / `反模式` 为子字段，无则 `[]`，不要拆成正文顶层并列大段。
|
||||
5. `一次生成` 为同批时关系必填且本批至少 2 条；为单条时一次只交 1 条。
|
||||
6. 字段值为扁平可填内容；不要嵌套对象当字段值契约。
|
||||
7. 不在产物正文写自评、追问、开放问题或最终记录。
|
||||
8. `只预生成` 时程序记为中间层；`仅游玩期` / `预生成并游玩期` 记为最终层。
|
||||
9. 字段名和专名用中文；英文专名音译为中文。
|
||||
6. 字段值为扁平可填内容；不要嵌套对象当字段值契约。几面拼一格的配方写在 `填充要求`。
|
||||
7. 描写类字段必须有 `地板`；规则里任何示例都要达到地板。标签看不懂的分级必须在 `分级表现` 写齐每档；没有此类分级则省略 `分级表现`。
|
||||
8. 不在产物正文写自评、追问、开放问题或最终记录。
|
||||
9. `只预生成` 时程序记为中间层；`仅游玩期` / `预生成并游玩期` 记为最终层。
|
||||
10. 字段名和专名用中文；英文专名音译为中文。
|
||||
|
||||
## score
|
||||
|
||||
```score
|
||||
字段妥当：
|
||||
- 5：字段由用途推出，不多不少、不重叠；该有的玩法/效果接口齐全，该省略的已省略。
|
||||
- 3：大体可用，但有 1 个冗余、重叠或明显缺失字段。
|
||||
- 1：通用名片乱塞，或关键效果/接口字段缺失。
|
||||
- 5：字段由用途推出，不多不少、不重叠；该有的玩法/效果接口齐全，该省略的已省略；描写类字段有达到可感知的地板，难懂分级有分级表现。
|
||||
- 3：大体可用，但有 1 个冗余、重叠或明显缺失字段，或个别描写字段地板含糊、某档分级只有阶名。
|
||||
- 1：通用名片乱塞，或关键效果/接口字段缺失；描写字段与示例都是单标签。
|
||||
|
||||
池与提示：
|
||||
- 5：该用池的用了，不该用的改成一句提示约束；池精炼可选取；提示具体可执行。
|
||||
- 5：只有封闭集合用池，不可穷举项用提示或自由；提示具体可执行，方向式提示以「方向名：一句概括」为主、不挂作品名。
|
||||
- 3：大体合理，但个别字段误用/漏用池，或池过泛、提示含糊。
|
||||
- 1：大段假穷举，或本该一句话提示却做成巨型词表。
|
||||
- 1：为姓名、气质、能力等建词池，大段假穷举，或本该一句话提示却做成巨型词表。
|
||||
|
||||
填写工作流：
|
||||
- 5：阶段简洁有力，依赖与自检/反模式能引导按重要性推进，不臃肿。
|
||||
- 5：阶段简洁有力，依赖与自检/反模式能引导按重要性推进，反复生成时含去重，不臃肿。
|
||||
- 3：有阶段但顺序或依赖不清，或清单偏长却抓不住关键咬合。
|
||||
- 1：无工作流，或与字段脱节，无法指导动态生成。
|
||||
```
|
||||
@@ -151,8 +170,30 @@ feeds: gm,auditor
|
||||
```checklist
|
||||
- [ ] 名称与用途已写清？
|
||||
- [ ] 字段由用途推出，无冗余重叠、无关键缺失？
|
||||
- [ ] 池与提示分工正确？
|
||||
- [ ] 填写工作流含子字段：工作流 / 依赖关系 / 自检清单 / 反模式？
|
||||
- [ ] 生命周期与一次生成已钉？
|
||||
- [ ] 池与提示分工正确？姓名、绰号、能力、外貌、气质没有建池？
|
||||
- [ ] 描写类字段有地板，示例不是单标签？难懂分级有分级表现？
|
||||
- [ ] 几面拼一格的字段在填充要求里写了配方？
|
||||
- [ ] 填写工作流含子字段：工作流 / 依赖关系 / 自检清单 / 反模式？反复生成时自检含去重？
|
||||
- [ ] 生命周期已钉？一次生成按关系定，而非按总数？
|
||||
- [ ] 共享格局只引用舞台骨架，没有重抄共通前提？
|
||||
- [ ] 未输出最终实例、百科长文、评分或追问？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：
|
||||
- 武功典籍：字段为名字 / 效果 / 阶层。阶层池选取封闭档位，另写 `分级表现`，每阶一句概括 + 可感知表象；名字与效果用提示约束。
|
||||
- 女主规则：不设「性别」；池为 `[]`；内在冲突自由写，地板是「能在一场戏里演出来的两难」。
|
||||
- D&D 怪物：名称、等级、攻击骰齐全；气质用提示约束，地板写到一个可观察的习性。
|
||||
- 外貌：`填充要求` 写配方「轮廓一句 + 质感一句 + 一个旁人会记住的细节」，拼成扁平的一段「外貌」。
|
||||
- 互为仇敌的两名开局 NPC → 一次生成=同批，关系写谁夺了谁的什么；可互换的酒馆路人 → 单条；总共要 5 个但彼此无关 → 数量 5、一次生成单条。
|
||||
|
||||
坏：
|
||||
- 为气质、绰号、能力各建一个 8～20 词的池，填出来全是标签。
|
||||
- 提示里的方向每条挂作品名或例句，变成假选项库。
|
||||
- 示例写「气质：阴郁」「能力：时空斩」，地板被示例教坏。
|
||||
- 阶层只写「A」「金丹」，没有分级表现，叙事时无从下笔。
|
||||
- 把社会格局、前史写成「共通前提」长文塞进规则，而不是引用舞台骨架。
|
||||
- 女主规则塞「性别」；怪物规则只有名字和性格，没有等级、攻击骰。
|
||||
```
|
||||
|
||||
@@ -11,8 +11,7 @@ declaration: >
|
||||
when: |
|
||||
本局存在拿掉后会改变故事骨架的既成法则，且后续需要据此推演环境变化、事件后果或人物反应时使用。
|
||||
when_not: |
|
||||
只有熟悉常识，没有特殊因果法则 → 不使用。
|
||||
地点、势力与社会格局 → 舞台骨架；多份人物或要先审字段的名片 → 生成规则，单条字段很少的具名内容才归具体实例；字段合同 → 生成规则。
|
||||
只有熟悉常识，没有特殊因果法则 → 不使用。 「特殊」只是身份或开场处境、能被舞台骨架、具体实例或开场白覆盖（如现实世界里模拟一名普通大学生）→ 不使用。 地点、势力与社会格局 → 舞台骨架；多份人物或要先审字段的名片 → 生成规则，单条字段很少的具名内容才归具体实例；字段合同 → 生成规则。
|
||||
boundary: |
|
||||
本能力交稳定因果，不预写唯一剧情，也不把所有人物钉成同一种反应。
|
||||
不解释起源，不扩写世界百科，不制定推进节奏、变量表或检定公式。
|
||||
@@ -39,12 +38,16 @@ feeds: gm
|
||||
|
||||
```principles
|
||||
1. 机制是稳定因果：写世界承认什么、何时生效、会带来什么后果。
|
||||
2. 内部使用移除检验筛选承重法则；只能回答“更丰富”或“以后也许有用”的候选不纳入。缺失后果不输出。
|
||||
3. 因果影响写行动可能性、代价、资源与关系怎样改变；推演口径写推演情节和人物反应时必须考虑什么。
|
||||
4. 人物反应保持开放：同一机制下，人物可因认知、利益、身份和性格不同而隐藏、合作、利用、抵抗或误判。
|
||||
5. 多条机制可一次生成。分别写清后，再说明它们是前提、叠加、制衡还是反馈循环；没有关系则不硬连。
|
||||
6. 机制不等于题材标签。“灵气复苏”“全民直播求生”必须拆成可推演的生效方式与因果链。
|
||||
7. 一句话够用便停止；不补起源史、体系百科、人物卡、剧情大纲或推进节奏。
|
||||
2. 法则可以是超常前提、约束性公理，也可以是社会性或规范性的既成事实，如“这个世界没有花里胡哨的魔法，只有繁衍到极致的斗气”“这个西幻世界里男尊女卑是常态”“这里以三国杀高手为尊”。同一设定有很多面向，只写后续推演必须承认的那一面。
|
||||
3. 身份特殊、开场处境特殊不是法则：能由舞台骨架、具体实例或开场白承担的（如现实世界里的普通大学生），不写成机制。
|
||||
4. 内部使用移除检验筛选承重法则；只能回答“更丰富”或“以后也许有用”的候选不纳入。缺失后果不输出。
|
||||
5. 内部按基准世界做最小充分继承，校验过程不写入产物：现实地球沿用现实物理、心理与社会规律；变造地球只在用户已表达或体验依赖处承认变造；具体作品世界优先最近一层作品设定，原作未定义处再参考其底层世界；类型世界只继承与当前体验有关的类型约定；从头创造以显式设定为准，未定义处用最少常识；混合或二创只处理与当前法则有关的继承与变造。
|
||||
候选法则超出基准世界已知可能性时，只有用户材料已明示或体验确实依赖才写入，不暗中变造；拿不准是否在用户设想中，就不写进产物，把这处变造交给 `self_score.gap`。
|
||||
6. 因果影响写行动可能性、代价、资源与关系怎样改变；推演口径写推演情节和人物反应时必须考虑什么。
|
||||
7. 人物反应保持开放：同一机制下，人物可因认知、利益、身份和性格不同而隐藏、合作、利用、抵抗或误判。
|
||||
8. 多条机制可一次生成。分别写清后，再说明它们是前提、叠加、制衡还是反馈循环；没有关系则不硬连。
|
||||
9. 机制不等于题材标签。“灵气复苏”“全民直播求生”必须拆成可推演的生效方式与因果链；题材标签也不能代替“这条法则缺了故事会换”。能指明它支撑美学纲领里哪段体验时，写进 `服务体验`。
|
||||
10. 一句话够用便停止；不补起源史、体系百科、人物卡、剧情大纲或推进节奏。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -60,7 +63,8 @@ feeds: gm
|
||||
"既成机制": [
|
||||
{
|
||||
"名称": "简短稳定的人话名",
|
||||
"法则内容": "世界中稳定成立什么",
|
||||
"法则内容": "世界中稳定成立什么；可以是超常前提、公理，也可以是社会常态或规范",
|
||||
"服务体验": "一句话说明它支撑美学纲领里的哪段体验；无明确对应则省略",
|
||||
"生效方式": [
|
||||
"对谁、在什么条件下、以什么方式起作用"
|
||||
],
|
||||
@@ -89,7 +93,7 @@ feeds: gm
|
||||
硬规则：
|
||||
1. 输出合法 JSON；`既成机制` 可含多条承重机制。
|
||||
2. `缺失后果`、移除检验和被淘汰候选只用于内部推理，不得进入正文。
|
||||
3. 每条机制必须写 `法则内容 / 生效方式 / 因果影响 / 推演口径`。
|
||||
3. 每条机制必须写 `法则内容 / 生效方式 / 因果影响 / 推演口径`；`服务体验` 与 `边界或例外` 可选。
|
||||
4. `共同作用` 只写真实存在的机制关系；无则 `[]`。
|
||||
5. 推演口径提供因果依据，不指定唯一剧情，也不规定所有人物作同一种反应。
|
||||
6. 不在产物正文写自评、追问、开放问题、起源百科或检定公式。
|
||||
@@ -101,7 +105,7 @@ feeds: gm
|
||||
骨架必要度：
|
||||
- 5：每条都是拿掉后会改变故事骨架的承重法则，没有常识、装饰或下游细节。
|
||||
- 3：主体承重，但夹有一条作用较弱或可交给其它能力的内容。
|
||||
- 1：只有题材标签或装饰设定，没有稳定承重法则。
|
||||
- 1：只有题材标签或装饰设定，或把身份、开场处境、暗中加入的变造写成法则。
|
||||
|
||||
因果清晰度：
|
||||
- 5：生效条件与影响链清楚，能据此推演环境、事件和关系怎样变化。
|
||||
@@ -121,6 +125,7 @@ feeds: gm
|
||||
- [ ] 法则内容、生效方式、因果影响与推演口径均可执行？
|
||||
- [ ] 多条机制的共同作用已写清，或明确无须关联？
|
||||
- [ ] 能支持不同人物作出不同但合理的反应？
|
||||
- [ ] 未把身份或开场处境写成法则？超出基准世界的变造都来自用户材料或体验依赖？
|
||||
- [ ] 未写成题材标签、世界百科、剧情大纲或裁定表？
|
||||
- [ ] 产物正文不含评分、诊断、追问？
|
||||
```
|
||||
@@ -136,4 +141,14 @@ feeds: gm
|
||||
- 可拆为“强制进入求生场”“关键行为公开直播”“观众反馈可进入场内（若本局采用）”等多条机制。
|
||||
- 公开评价与即时生存会共同改变选择代价；人物是否表演、隐藏或利用镜头，应由其是否在乎声誉、是否相信能返回现实等条件推出。
|
||||
- 观众若只能观看，就不能推演出打赏资源；边界会直接改变因果链。
|
||||
|
||||
好：
|
||||
- 社会常态类法则“这个西幻世界里男尊女卑是常态”：生效方式写继承、婚配、对外交涉由谁做主；因果影响写女性角色行动的额外代价与绕行渠道；推演口径写人物可能顺从、利用或反抗，不规定人人同一态度。
|
||||
- “这个世界没有花里胡哨的魔法，只有繁衍到极致的斗气”：只写斗气是唯一超常力量、怎样决定地位与胜负，不展开境界百科。
|
||||
- 公理“黑暗森林”：一句话钉“暴露坐标即招来打击，文明默认隐藏”；后续情节不能靠对外广播求援收场，不写宇宙社会学论文。
|
||||
|
||||
坏：
|
||||
- 现实世界模拟普通大学生，把“主角是普通大学生”写成机制：这是身份与开场处境，交给具体实例或开场白；没有特殊法则就不使用本能力。
|
||||
- 二创里暗中加入原作没有的力量体系（如在无超能力的原作里写入“觉醒异能”）：用户未设想的变造不写入，拿不准时交给 `self_score.gap`。
|
||||
- 只写“灵气复苏”“末世”等题材标签，没有生效方式与因果链：主世界层无法据此推演任何后果。
|
||||
```
|
||||
|
||||
@@ -7,18 +7,18 @@ name: 叙事指南
|
||||
id: narrative-guide
|
||||
artifact: 设计.叙事指南
|
||||
declaration: >
|
||||
从文风库选取本局写法，落成可执行的遣词要求、笔墨落点、少量禁忌与可选写法档；不钉故事怎么往前。
|
||||
继承美学纲领与交互范式，从文风库选取本局写法，落成可执行的遣词要求、视角执行、段落形态、笔墨落点、少量禁忌与可选写法档；不钉故事怎么往前。
|
||||
when: |
|
||||
已有美学纲领（或等价），但游玩正文仍不知道怎么写：遣词什么要求、笔墨落谁身上、场合是否要换写法档、有哪些绝对不能写。
|
||||
用户主动谈文风、措辞、写法档时也使用。
|
||||
已有美学纲领（或等价），但游玩正文仍不知道怎么写：遣词什么要求、视角怎么落笔、笔墨落谁身上、场合是否要换写法档、有哪些绝对不能写。
|
||||
用户主动谈文风、措辞、叙述声音、写法档时也使用。
|
||||
when_not: |
|
||||
本技能只给对话落盘。工序编排不排本步（编排池是「叙事指南与故事推进」）。
|
||||
美学未钉 → 先「美学纲领与交互范式」。
|
||||
用户明确说本局不用文风细则，且美学里已有可执行遣词方向 → 可不使用。
|
||||
只谈推进、输入用法 → 「故事推进」。版式 → 「正文组成」。字数与可见区域 → 「回复呈现」。
|
||||
只谈推进、输入用法 → 「故事推进」。单轮字数与篇幅 → 「故事推进」；可见区域 → 「回复呈现」。版式 → 「正文组成」。
|
||||
boundary: |
|
||||
本能力只交怎么写：选用文风、遣词要求、笔墨落点、少量绝对禁忌、可选写法档。
|
||||
推进与输入用法归故事推进；体验契约归美学纲领。
|
||||
本能力只交怎么写：选用文风、遣词要求（含叙述声音、视角执行、段落形态）、笔墨落点与代入取向、少量绝对禁忌、可选写法档。
|
||||
推进、输入用法、单轮字数与篇幅归故事推进；体验契约、人称与叙事视角的定义归美学纲领与交互范式。
|
||||
feeds: narrator,gm
|
||||
```
|
||||
|
||||
@@ -31,9 +31,9 @@ feeds: narrator,gm
|
||||
## task
|
||||
|
||||
```task
|
||||
把本局写法整理为叙事指南，写入 `设计.叙事指南`。先用 `read_library_entry` 从绑定的文风库读取最相关条目（默认 1～2；确需第三种时须写 reason 并标明适用场合），再按本局体验本局化，写成可执行的风格推荐与要求。
|
||||
把本局写法整理为叙事指南，写入 `设计.叙事指南`。先继承上游「美学纲领与交互范式」：把 `美学纲领.体验内核`、`呈现要点.描写焦点` 与 `呈现要点.边界与尺度` 改写成做法向句子；`交互范式` 已定的人称与叙事视角不重定义，只在 `遣词.视角` 写怎么执行。再用 `read_library_entry` 从绑定的文风库读取最相关条目（默认 1～2；确需第三种时须写 reason 并标明适用场合），按本局体验本局化，写成可执行的风格推荐与要求。
|
||||
|
||||
产物回答：选了哪些文风、各自用在什么场合；遣词与笔墨要怎么做；只有少数绝对禁忌。除禁忌外，少写“不要怎样”，多写“要怎样”。不写示范段落，不写推进与输入用法。
|
||||
产物回答：选了哪些文风、各自用在什么场合；遣词、视角执行、段落形态与笔墨要怎么做；叙述声音与代入取向（推得出才写）；只有少数绝对禁忌。除禁忌外，少写“不要怎样”，多写“要怎样”。不写示范段落，不写推进、输入用法或单轮字数。
|
||||
|
||||
材料足够便直接成稿；真正会改写法的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -41,12 +41,16 @@ feeds: narrator,gm
|
||||
## principles
|
||||
|
||||
```principles
|
||||
1. 文风库是参考，不是照抄：读取后改成本局可执行要求，并记下适用场合与微调。
|
||||
2. 优先 1～2 种文风；确需第三种时，每种都必须有清晰切换场合，禁止无差别堆叠。
|
||||
3. 写法用“风格推荐 + 要求”表达：对象、节奏、注意力落点、句段形态都写成要做什么。示范段落不写。
|
||||
4. 绝对禁忌只留真正碰不得的少数几条；其余偏好都改成正向做法。
|
||||
5. 写法档是多文风/多调性容器：有切换才写，无切换整项省略。场合名要能指导何时切入，不是具体剧情桥段清单。
|
||||
6. 不写故事推进、输入用法、字数区间或版式壳。
|
||||
1. 美学写要什么感受，本步写怎么遣词、笔墨落哪。已在美学写清的感受不重问，直接改写成做法：体验内核与描写焦点落成详写对象和写法要求；边界与尺度按美学给的用法写进写法要求或写法档，只把其中真正碰不得的收进禁忌。
|
||||
2. 文风库是参考，不是照抄：读取后改成本局可执行要求，并记下适用场合与微调。
|
||||
3. 优先 1～2 种文风；确需第三种时，每种都必须有清晰切换场合，禁止无差别堆叠。
|
||||
4. 写法用“风格推荐 + 要求”表达：对象、节奏、注意力落点、句段形态都写成要做什么。示范段落不写。
|
||||
5. 段落形态默认值：用户未提段落形态时，`遣词.写法要求` 写入「完整长自然段叙述，不要单句成段」；用户另定则按用户写。段落形态归本步，单轮字数与篇幅归故事推进。
|
||||
6. 视角只写执行：人称、主要视角、可写内心、不呈现的信息以交互范式为准，本步写落笔做法——镜头距离（贴身 / 中景 / 远观，何时拉近）；内心写到哪（只写交互范式允许写内心的一方，其余人物用外显言行与身体反应表现）；如何不漂人称（全文固定一种称呼，回忆、转述与内心独白也不换人称）。交互范式未定视角时按美学站位取最贴的写法，并在 gap 写明。
|
||||
7. `叙述者` 写叙述声音：冷眼旁观、调侃、贴身低语等，以及它对人物的态度。`代入取向` 写笔墨替谁看、读者站在谁那边。两项都从美学站位与体验推出；推不出或与视角执行重复时省略。
|
||||
8. 绝对禁忌只留真正碰不得的少数几条；其余偏好都改成正向做法。
|
||||
9. 写法档是多文风/多调性容器：有切换才写，无切换整项省略。场合名要能指导何时切入，不是具体剧情桥段清单。用户报出「当众折辱」「独处一室」这类具体场合时，判断它要哪种写法，升一档写进对应写法档的适用场合与写法要求（感官密度、节奏、对话长短、详略），场合本身不单列；每档写清切入与切回。
|
||||
10. 不写故事推进、输入用法、单轮字数与篇幅（归故事推进）或版式壳。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -68,19 +72,34 @@ feeds: narrator,gm
|
||||
],
|
||||
"遣词": {
|
||||
"整体取向": "本局叙述应呈现的风格方向",
|
||||
"叙述者": "可选：叙述声音，如冷眼旁观 / 调侃 / 贴身低语，及它对人物的态度；无则省略",
|
||||
"风格推荐": ["优先采用的节奏、用词层级、句段形态"],
|
||||
"写法要求": ["要写到什么密度、如何处理对话与叙述、注意力放哪"]
|
||||
"写法要求": [
|
||||
"完整长自然段叙述，不要单句成段（用户另定段落形态则改写）",
|
||||
"要写到什么密度、如何处理对话与叙述、注意力放哪"
|
||||
],
|
||||
"视角": {
|
||||
"镜头距离": "如：贴身跟随 user，冲突时拉近到手与眼神，转场时拉远一句带过",
|
||||
"内心": "如：只写 user 的内心；NPC 只写外显言行与身体反应",
|
||||
"人称": "如：全文第二人称「你」，回忆与内心独白不换人称"
|
||||
}
|
||||
},
|
||||
"笔墨焦点": {
|
||||
"代入取向": "可选：笔墨替谁看、读者站在谁那边；无则省略",
|
||||
"详写": ["应花笔墨的对象与感受"],
|
||||
"可略": ["可一带而过的部分"]
|
||||
},
|
||||
"禁忌": ["仅少数绝对不能写的内容；无则省略"],
|
||||
"写法档": [
|
||||
{
|
||||
"名字": "档名，如展示情色 / 网文爽文",
|
||||
"适用场合": "何时切到这一档",
|
||||
"写法要求": "这一档要怎么写、笔墨落哪、如何切回"
|
||||
"名字": "展示情色",
|
||||
"适用场合": "切到身体与权力交换时；用户报的「独处一室」这类场合升到这一档",
|
||||
"写法要求": "感官词加重、节奏放慢、对话变短；切回日常则收"
|
||||
},
|
||||
{
|
||||
"名字": "网文爽文",
|
||||
"适用场合": "打脸、反转时；用户报的「当众折辱」之后的反击升到这一档",
|
||||
"写法要求": "可复述对方原话、结果写满、过程可略；反转落定后切回默认文风"
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -91,41 +110,53 @@ feeds: narrator,gm
|
||||
1. 输出合法 JSON；`选用文风` 默认 1～2 项，最多 3 项，每项必须有适用场合与本局要求。
|
||||
2. 不写示范段落、示例正文或仿写样章。
|
||||
3. 除 `禁忌` 外，正文以正向要求为主；空话氛围词改成可执行做法。
|
||||
4. `写法档` 无切换时整项省略；有则每档写清场合与写法要求。
|
||||
5. 不在产物正文写自评、追问、开放问题或推进/输入用法。
|
||||
6. `mount` 固定包含 `narrator` 与 `world-simulator`。
|
||||
4. 用户未定段落形态时，`遣词.写法要求` 必含「完整长自然段叙述，不要单句成段」。
|
||||
5. `遣词.视角` 只写交互范式人称与视角的执行，不改人称、不扩大可写内心的范围。`叙述者`、`代入取向` 可选，无则省略。
|
||||
6. `写法档` 无切换时整项省略；有则每档写清场合与写法要求；具体场合升档写入，不单列桥段。
|
||||
7. 不在产物正文写自评、追问、开放问题，或推进、输入用法、单轮字数。
|
||||
8. `mount` 固定包含 `narrator` 与 `world-simulator`。
|
||||
|
||||
## score
|
||||
|
||||
```score
|
||||
遣词可执行：
|
||||
- 5：风格推荐与要求具体到能直接落笔，无需再猜味道。
|
||||
- 3：大方向清楚，但仍有几处停留在氛围概括。
|
||||
- 5：风格推荐与要求具体到能直接落笔，无需再猜味道；视角执行与段落形态写清。
|
||||
- 3：大方向清楚，但仍有几处停留在氛围概括，或视角执行含糊。
|
||||
- 1：只有题材或情绪标签，无法指导遣词。
|
||||
|
||||
笔墨与禁忌：
|
||||
- 5：详略落点清楚；绝对禁忌少而可执行；其余多为正向要求。
|
||||
- 5：详略落点清楚，贴合美学的描写焦点与边界尺度；绝对禁忌少而可执行；其余多为正向要求。
|
||||
- 3：落点大体有，但禁忌偏多或详略仍含糊。
|
||||
- 1：几乎不谈笔墨，或堆满否定句却无正向做法。
|
||||
|
||||
文风选用：
|
||||
- 5：1～2（偶发 3）条贴合本局；多档时切换场合清楚，无重复堆叠。
|
||||
- 5：1～2（偶发 3）条贴合本局；多档时切换场合清楚，具体场合已升档写入，无重复堆叠。
|
||||
- 3：选用相关，但场合区分弱，或有一条不够贴。
|
||||
- 1：未读库硬编，或多种文风无差别并列。
|
||||
|
||||
本局化：
|
||||
- 5：库内容已改成本局说法与要求，能看出针对本局体验的取舍。
|
||||
- 3：有微调，但仍有整段近乎原文搬用。
|
||||
- 5：库内容与美学感受已改成本局说法与要求，能看出针对本局体验的取舍。
|
||||
- 3：有微调，但仍有整段近乎原文搬用，或重问了美学已写清的感受。
|
||||
- 1：基本照抄库条目，或与本局体验脱节。
|
||||
```
|
||||
|
||||
## checklist
|
||||
|
||||
```checklist
|
||||
- [ ] 已继承美学的体验、描写焦点与边界尺度，改写成做法？
|
||||
- [ ] 已按需读取文风库并本局化？
|
||||
- [ ] 选用文风 1～2（最多 3），各有适用场合？
|
||||
- [ ] 遣词是风格推荐与要求，不是示范段落？
|
||||
- [ ] 遣词是风格推荐与要求，不是示范段落？用户未定段落形态时已写长自然段默认？
|
||||
- [ ] 视角只写执行，未改交互范式的人称与可写内心？
|
||||
- [ ] 具体场合已升档写入写法档，未单列桥段？
|
||||
- [ ] 除少数禁忌外，多为“要怎样”？
|
||||
- [ ] 未写入推进、输入用法、字数或版式？
|
||||
- [ ] 未写入推进、输入用法、单轮字数或版式？
|
||||
- [ ] 产物正文不含评分、诊断、追问？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：美学的「掌控感」落成详写对方的迟疑与屈从、user 的动作从容简短；视角写「第二人称贴身镜头，只写你的内心，NPC 只写外显」；写法要求含长自然段默认；用户提的「当众折辱」升进「网文爽文」档，写清切入与切回。
|
||||
坏：重新定义人称或扩大可写内心；把「当众折辱」「独处一室」列成桥段清单；写示范段落；在本步写单轮字数或推进方式。
|
||||
```
|
||||
|
||||
@@ -5,10 +5,10 @@
|
||||
```meta
|
||||
name: 开场白与开场变量
|
||||
id: opening-setup
|
||||
artifact: 设计.开场白与开场变量
|
||||
artifact: 设计.开场白
|
||||
declaration: >
|
||||
普通楼层创作终节点，且可增殖：每条开场是一个节点，落成 `设计.开场白#场景短码`。
|
||||
默认写 1 条；用户要多个入口时再增殖。每条自带 present.v1、同真相初态与可选用户角色。
|
||||
默认写 1 条；用户要多个入口时再增殖。每条自带 present.v1、同真相初态与这条开场的最少玩家角色事实。
|
||||
全部保存。游玩时左右切换，选中的一条作为 0 层。
|
||||
when: |
|
||||
回复呈现、正文组成和其它本局内容能力均已完成，且程序已依次编译信息可见范围、游玩执行方式、上下文排序与运行配置后，需要可开玩的第一屏时使用。
|
||||
@@ -34,22 +34,25 @@ feeds: gm,narrator
|
||||
## task
|
||||
|
||||
```task
|
||||
读取程序已组装的 `设计.worker集` 和全部最终层产物，写入 `设计.开场白与开场变量`。这是普通楼层创作终节点；不得在本步首次整理执行单元或上下文顺序。
|
||||
读取程序已组装的 `设计.worker集` 和全部最终层产物，把每一张开场写成 `设计.开场白#场景短码`。这是普通楼层创作终节点；不得在本步首次整理执行单元或上下文顺序。
|
||||
|
||||
默认只生成 1 条开场。用户要多个入口时，每条再增殖一份产物，对话落盘写入 `设计.开场白#场景短码`（如 dorm、gate）；改某一条则覆盖同一完整 tag。不要把多条入口收成聊天里的选择题。每条候选各自携带：
|
||||
- 稳定 id 与一句区分说明
|
||||
- 可选用户角色名片（不同候选可不同）
|
||||
- 与该条叙述同真相的变量初值
|
||||
- 若本局有拓扑：当前位置/焦点与已揭示范围
|
||||
position 必须带场景短码，例如 `设计.开场白#opening-cycle2`。改某一条就再 insert 同一个完整 tag。裸 `设计.开场白` 和整包 `设计.开场白与开场变量` 会和带短码的卡抢同一条写入，不要使用。
|
||||
|
||||
默认只生成 1 条，仍然要带场景短码。用户要多个入口时，每条再 insert 一个新的短码。不要把多条入口收成聊天里的选择题，也不要把多条塞进同一次 insert。游玩时按这些短码左右切换，选中的一条作为 0 层，并整套换该条正文、初态与角色。
|
||||
|
||||
每一张卡携带：
|
||||
- 符合正文组成数据契约的 `present.v1`（只填模型生成区）
|
||||
|
||||
全部候选都保存。每条正文只写在该候选的 `正文` 里，不要再抄一份全文或初值。`默认候选id` 指向第一次进入 0 层的那条；程序按这个 id 读取它的正文和初值。游玩时左右切换，选中的一条作为 0 层，并整套换该条正文、初态与角色。
|
||||
- 可选一句区分说明
|
||||
- 这一条开场的玩家角色 `meta.用户角色`（名字 + 简介）：不同开场的玩家可以不同，写在各自卡里；程序会同步成同短码的主角设定，游玩切到这条时替换 `@玩家`
|
||||
- 与该条叙述同真相的变量初值：这张卡有要开场定下的变量才写 `开场变量`；没有就不写，卡仍然是 `设计.开场白#场景短码`
|
||||
- 若本局有拓扑：当前位置/焦点与已揭示范围
|
||||
|
||||
硬规格：
|
||||
1. 每条可见正文至少出现一次字面 `@玩家`；不写死具体姓名顶名字位。
|
||||
1. 可见正文里凡指称玩家角色的名字位都写字面 `@玩家`，每条至少出现一次。「你」可以和 `@玩家` 并用（如「你（@玩家）站在闸机前」），但全文只有「你」、零个 `@玩家` 不合格；不写死具体姓名，也不用「用户」「主角」顶名字位。
|
||||
2. `present.v1` 的区域只覆盖正文组成声明的模型生成 blocks；程序投影区留空给程序灌。
|
||||
3. 不输出 `名称.随机数` 的具体数组；随机行由程序在第一次正式回复前或回复时补齐。
|
||||
4. 开场宜短，愿点进游玩即可，不抄故事推进的单轮篇幅预算当第一章。
|
||||
5. 人物、地点、势力只引用已有产物里的；第一拍需要的路人只写身份，不在开场新造有名有姓的设定。
|
||||
|
||||
材料足够便直接成稿；会改变第一拍成立的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -57,12 +60,13 @@ feeds: gm,narrator
|
||||
## principles
|
||||
|
||||
```principles
|
||||
1. 开场为主，初态配套。叙述与该条变量、拓扑初态必须同真相。
|
||||
2. 默认一条。多条是不同入口，每条一个可增殖节点，不是同文换皮，也不是先让用户选唯一开场。
|
||||
3. 用户角色可按候选不同；没有名片时仍用 `@玩家` 占位，进游玩再替换。
|
||||
4. 只写模型生成区。状态芯片、地图等权威显示由程序投影，不在开场正文里重编一份。
|
||||
5. 不剧透隐藏档，不写成第一章，不在本步改前端模板。
|
||||
6. 本步结束后创作收口；不要再排后续内容能力。
|
||||
1. 开场为主，初态配套。叙述与该条变量、拓扑初态必须同真相；写完发现两者对不上时以叙述为准，改初值。
|
||||
2. 默认一条。多条是不同入口，每条一个 `设计.开场白#场景短码`，不是同文换皮，也不是先让用户选唯一开场。没有变量的卡省略 `开场变量`，短码照旧。
|
||||
3. 玩家角色最少化：把这个角色交还给用户扮演。`用户角色.简介` 只写这条开场成立所必需的客观事实——身份、处境、与在场人物的已知关系，一到三句；不写性格、心理、动机、口癖、外貌细节，也不替玩家定态度或下一步。共通的主角设定已有的事实不重抄，只写这条开场独有或与它不同的部分。
|
||||
4. 用户没给名字时不编名字，省略 `用户角色`，正文照用 `@玩家`；这条开场必需的身份事实写进叙述里让玩家看见。
|
||||
5. 只写模型生成区。状态芯片、地图等权威显示由程序投影，不在开场正文里重编一份。
|
||||
6. 文风、视角和人称跟叙事指南与交互范式；不剧透隐藏档，不写成第一章，不在本步改前端模板。
|
||||
7. 本步结束后创作收口；不要再排后续内容能力。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -74,62 +78,48 @@ feeds: gm,narrator
|
||||
## output
|
||||
|
||||
```output
|
||||
position: 设计.开场白#场景短码
|
||||
content:
|
||||
{
|
||||
"schema": "context-fragment.v1",
|
||||
"技能": "开场白与开场变量",
|
||||
"brief": "一句话：默认开场场景 + 有几条候选",
|
||||
"mount": ["world-simulator", "narrator"],
|
||||
"稳变": "stable",
|
||||
"正文": {
|
||||
"默认候选id": "opening-1",
|
||||
"开场白候选": [
|
||||
"schema": "present.v1",
|
||||
"shell": "与正文组成一致",
|
||||
"blocks": {
|
||||
"body": "清晨，@玩家推开门。……",
|
||||
"其余模型生成区": "read_module 回包 reply_contract.model_blocks 里的每一项都要出现，类型按契约"
|
||||
},
|
||||
"meta": {
|
||||
"区分": "多条时写这一入口和其它条差在哪；仅一条可省略",
|
||||
"用户角色": {
|
||||
"名字": "用户给过的名字；没给就省略整个用户角色",
|
||||
"简介": "一到三句客观事实：这条开场里玩家是谁、处境如何、与在场人物的已知关系"
|
||||
},
|
||||
"开场变量": [
|
||||
{
|
||||
"id": "opening-1",
|
||||
"区分": "这条入口与其它条差在哪；仅一条时可写「默认」",
|
||||
"meta": {
|
||||
"用户角色": {
|
||||
"名字": "可选；无则省略整段",
|
||||
"简介": "可选"
|
||||
},
|
||||
"开场变量": [
|
||||
{
|
||||
"名": "与变量设计真值名一致",
|
||||
"值": "初值",
|
||||
"依据": "开场事实"
|
||||
}
|
||||
],
|
||||
"拓扑初态": [
|
||||
{
|
||||
"topology_id": "有拓扑时填写",
|
||||
"当前位置": "节点 id",
|
||||
"已揭示": ["节点或连接 id"]
|
||||
}
|
||||
]
|
||||
},
|
||||
"正文": {
|
||||
"schema": "present.v1",
|
||||
"blocks": {
|
||||
"body": "清晨，@玩家推开门。……"
|
||||
},
|
||||
"meta": {
|
||||
"suggested_actions": ["若正文组成启用文字选项则给短建议；否则 []"]
|
||||
}
|
||||
}
|
||||
"名": "与变量设计真值名一致",
|
||||
"值": "初值",
|
||||
"依据": "开场事实"
|
||||
}
|
||||
],
|
||||
"用户可行动空间": "开场结束后玩家显然能尝试什么"
|
||||
"拓扑初态": [
|
||||
{
|
||||
"topology_id": "有拓扑时填写",
|
||||
"当前位置": "节点 id",
|
||||
"已揭示": ["节点或连接 id"]
|
||||
}
|
||||
],
|
||||
"suggested_actions": ["若正文组成启用文字选项则给短建议；否则 []"]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
硬规则：
|
||||
1. 输出合法 JSON；`开场白候选` 默认 1 条，仅用户明确要求时 2～3 条。
|
||||
2. `默认候选id` 必须指向候选之一。正文和初值只写在该候选上，不另抄全文或开场变量。
|
||||
3. 每条 `正文` 必须是 `present.v1`，只含正文组成允许模型生成的 blocks；每条可见正文须含字面 `@玩家`。
|
||||
4. 不同候选可有不同用户角色与初态；切换时整套替换，不混用。
|
||||
1. 每次 insert 只写一张卡。position 是 `设计.开场白#场景短码`，content 是这一张的 `present.v1`。
|
||||
2. 这张卡有初值才写 `meta.开场变量`；没有变量就省略该字段，position 仍带场景短码。
|
||||
3. `present.v1` 只含正文组成允许模型生成的 blocks；可见正文须含字面 `@玩家`。
|
||||
4. 不同卡可有不同用户角色与初态；切换时整套替换，不混用。
|
||||
5. 不输出随机数具体值，不重写前端，不把程序投影区填成模型编造状态。
|
||||
6. 产物正文不含自评、追问或开放问题。
|
||||
7. `mount` 固定为 `["world-simulator", "narrator"]`。
|
||||
7. 玩家角色只写在 `meta.用户角色`，名字不进 blocks 正文；程序据此写入同短码的 `设计.主角设定#场景短码`，不要另 insert 一份。
|
||||
```
|
||||
|
||||
## score
|
||||
|
||||
@@ -150,9 +140,9 @@ feeds: gm,narrator
|
||||
- 1：叙述与初值冲突，或多条共用一套却声称不同入口。
|
||||
|
||||
角色占位：
|
||||
- 5：每条含字面 `@玩家`；有名片时不把姓名写进正文名字位。
|
||||
- 3：占位大体正确，有一条名片与正文关系略乱。
|
||||
- 1：写死姓名，或全文没有 `@玩家`。
|
||||
- 5：每条含字面 `@玩家`；`用户角色.简介` 只有这条开场必需的客观事实，玩家性格与决定留给用户。
|
||||
- 3：占位正确，但简介多写了性格、动机或与共通主角设定重复的内容。
|
||||
- 1：写死姓名、全文只有「你」没有 `@玩家`，或简介替玩家定了性格和态度。
|
||||
|
||||
收口可切换：
|
||||
- 5：默认条明确，全部候选可保存；多条时入口差异成立。
|
||||
@@ -163,17 +153,22 @@ feeds: gm,narrator
|
||||
## checklist
|
||||
|
||||
```checklist
|
||||
- [ ] 默认 1 条；仅用户要求时才有 2～3 条？
|
||||
- [ ] `默认候选id` 指向一条候选，且该条正文、初值没有在外面再抄一份？
|
||||
- [ ] 每条含 `@玩家`，present.v1 只含模型生成区？
|
||||
- [ ] 各候选初态自洽，未填写随机数具体值？
|
||||
- [ ] position 是 `设计.开场白#场景短码`，没有另写裸 `设计.开场白` 或 `设计.开场白与开场变量`？
|
||||
- [ ] 默认 1 条；仅用户要求时才再增殖别的短码？
|
||||
- [ ] 有初值才写 `开场变量`；没有变量就省略，卡名不变？
|
||||
- [ ] 含 `@玩家`，present.v1 只含模型生成区？
|
||||
- [ ] `用户角色.简介` 只有必需客观事实，没有性格、心理和替玩家做的决定？
|
||||
- [ ] 初态与这条叙述同真相，未填写随机数具体值？
|
||||
- [ ] 未重开前端、篇幅或区域清单？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
- 单条：堂屋债契，body 含 `@玩家`，开场变量写债额初值；状态栏由程序投影。
|
||||
- 用户要两条入口：一条放贷人名片、一条社畜穿越名片；各带自己的初态，默认指定其中一条开玩。
|
||||
- 不合适：开场里填写 `敏捷.随机数=[…]`；或把地图权威数据写成模型散文却不给拓扑初态。
|
||||
- 单条：`设计.开场白#debt-hall`，堂屋债契，body 含 `@玩家`，开场变量写债额初值；状态栏由程序投影。
|
||||
- 用户要两条入口：`设计.开场白#lender` 与 `设计.开场白#office`，各带自己的初态。没有初值的那条省略开场变量。
|
||||
- 单条带玩家角色：`meta.用户角色` 为 `{"名字":"沈聿","简介":"本地放贷人，持有这家的债契；与这对母女相识但不熟。"}`，body 写「清晨，@玩家推开堂屋门，债契还摊在桌上。」
|
||||
- 不合适：再写一份裸 `设计.开场白` 或 `设计.开场白与开场变量`；开场里填写 `敏捷.随机数=[…]`。
|
||||
- 不合适：全文只有「你推开门」，零个 `@玩家`。
|
||||
- 不合适：简介写「沉稳果决，内心渴望被认可，决定今天收回债款」——性格、心理和决定都该留给玩家。
|
||||
```
|
||||
|
||||
@@ -37,11 +37,14 @@ feeds: gm
|
||||
|
||||
```principles
|
||||
1. 只设计用户已明确给出、核心体验依赖或运行确实需要的内容。
|
||||
2. “首富之子”“唯我独法”等若是核心体验成立条件，属于必要设定，不算过度设计。
|
||||
3. 特殊能力写“能做什么 + 限制”；世界为何如此由实现机制处理。
|
||||
4. `称呼` 可写固定名字，也可写 `@玩家` 表示刻意留白；不要因无名阻塞成卡。
|
||||
5. 行为依据随推进权增减：用户控制越多越简；LLM 代演越多越具体。
|
||||
6. 不写 NPC 名片、世界百科、文风规则或开场事件。
|
||||
2. 一张脸测试：逐句问「删掉这句，玩家还知道自己是谁吗？」删了仍知道的句子不落盘。
|
||||
3. “首富之子”“唯我独法”等若是核心体验成立条件，属于必要设定，不算过度设计。
|
||||
4. 特殊能力写“能做什么 + 限制”；世界为何如此由实现机制处理。
|
||||
5. `称呼` 可写固定名字，也可写 `@玩家` 表示刻意留白；不要因无名阻塞成卡。
|
||||
6. 身份与背景一两句；用户没给就写「未定」或省略，不代编来历。
|
||||
7. 行为依据随推进权增减：用户控制越多越简；LLM 代演越多越具体。
|
||||
8. 本能力只写各开场共通的玩家角色。不同开场白里玩家身份、处境不同的部分，由开场白写在自己的 `meta.用户角色`，程序会同步成 `设计.主角设定#场景短码`；本能力不为各开场另写 `设计.主角设定#短码`。
|
||||
9. 不写 NPC 名片、世界百科、文风规则或开场事件。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -101,3 +104,12 @@ feeds: gm
|
||||
- [ ] 称呼允许固定名字或 `@玩家` 留白？
|
||||
- [ ] 未写 NPC、世界百科、文风或开场事件？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
- 好：称呼「林晚」；身份与背景一句穿来当日的处境；特殊设定「能看见别人对她的好感数字」，限制「别人看不见」。
|
||||
- 坏：把侯府地理、朝堂势力写进背景。
|
||||
- 坏：把开场羞辱戏全文写进名片。
|
||||
- 坏：为宿舍开场和天台开场各写一张 `设计.主角设定#短码`；各开场不同的处境应写进对应开场白的 `meta.用户角色`。
|
||||
```
|
||||
|
||||
@@ -30,7 +30,7 @@ feeds: gm
|
||||
## task
|
||||
|
||||
```task
|
||||
写入 `设计.随机范围整理`。每项是 `变量.当前` 的一行，行名写成「名称.随机数」，玩家默认不可见。值是剩余数字组成的数组，下一个可用的是第一个。
|
||||
先扫实现机制、生成规则、变量设计、叙事与故事推进、美学，列出所有需要随机数才能裁定的点，再写入 `设计.随机范围整理`。每项是 `变量.当前` 的一行，行名写成「名称.随机数」，玩家默认不可见。值是剩余数字组成的数组，下一个可用的是第一个。
|
||||
|
||||
用掉第一个后，程序改写这一行，去掉那个数。改完后如果只剩 1 个，就按范围补到预备个数，不要等这一行空了。模型不自己编数字。
|
||||
|
||||
@@ -41,9 +41,10 @@ feeds: gm
|
||||
|
||||
```principles
|
||||
1. 只登记上游已经成立的掷法，不新写法则。
|
||||
2. 一行一项，行名是「名称.随机数」，例如 `敏捷.随机数`。
|
||||
3. 和别的隐藏变量一样存在现有表里：能改、有版本；不对玩家默认展示。
|
||||
4. 预备个数够连用即可，少了靠「只剩 1 个就补」，不要一次放很多。
|
||||
2. 一行一项，一项只管一类用途，行名是「名称.随机数」，例如 `敏捷.随机数`。不要把「战斗」糊成一行却含命中、伤害、闪避几种掷法，拆成 `命中.随机数`、`伤害.随机数`、`闪避.随机数`。
|
||||
3. 范围写成人与程序都能读的形式，如 `1～100`、`力量*5～10`，并注明单位（整数、闭区间等）；不写「中等偏高」这类散文。
|
||||
4. 和别的隐藏变量一样存在现有表里：能改、有版本；不对玩家默认展示。
|
||||
5. 预备个数够连用即可，少了靠「只剩 1 个就补」，不要一次放很多。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -68,6 +69,7 @@ feeds: gm
|
||||
"用途": "敏捷检定",
|
||||
"范围": "1～100",
|
||||
"表达式": "无表达式时省略",
|
||||
"单位": "整数，闭区间",
|
||||
"依赖真值": [],
|
||||
"预备个数": 3,
|
||||
"来源": "上游短引用"
|
||||
|
||||
@@ -18,7 +18,8 @@ when_not: |
|
||||
boundary: |
|
||||
回复呈现决定“有什么、何时出现、来自哪里”；本能力决定“在前端哪里、如何渲染”。
|
||||
状态和拓扑数据由程序投影，本能力只写绑定与渲染器；不得让正文模型重写权威数据。
|
||||
隐藏维护区不进 DOM。随机数是变量表隐藏行，也与本能力无关。
|
||||
隐藏维护区不进 DOM；表由旁观副 LLM 维护，主回复不建填表通道。
|
||||
随机数是变量表隐藏行，也与本能力无关。
|
||||
壳参考库只提供布局与交互方法参考，不是必须照抄的固定产品模板。
|
||||
feeds: gm,narrator
|
||||
```
|
||||
@@ -35,11 +36,11 @@ feeds: gm,narrator
|
||||
为普通楼层生成一份可复用前端，写入 `设计.正文组成`。每条助手消息仍各自形成一个气泡；程序把当轮 `present.v1` 灌入同一模板。
|
||||
|
||||
执行顺序：
|
||||
1. 读取回复呈现，只处理其中实际启用的语义区域。
|
||||
1. 读取回复呈现，只处理其中实际启用的语义区域。同时读美学纲领的呈现要点与叙事指南，作为配色、字体层级和纹样的依据；不照搬参考壳的样子。
|
||||
2. 从【壳参考库】选一只与本局已启用区域最接近的壳。按该壳备注改写样本里的名称、栏目、配色和内部小节。本局没有的区从样本骨架删除并重排，不留空位。
|
||||
3. 为每个用户可见区域定义 block、位置、显示名和渲染方式。模型生成区读取对应 block；程序投影区直接绑定权威数据。
|
||||
4. 隐藏 `state_ops` 等维护区列入非渲染处理，不创建 DOM 节点。
|
||||
5. 写完整 HTML/CSS/JS，创作时一次生成并固定；游玩只灌数据，不再重写前端。美化用配色、字体层级、边框、间距，以及 CSS 或简单内联 SVG，按本次能写稳的程度完成。JS 通过 `present.onData(render)` 接收数据；地图所需的 SVG 从拓扑数据绘制，游玩模型不输出 SVG。
|
||||
5. 写完整 HTML/CSS/JS，创作时一次生成并固定；游玩只灌数据，不再重写前端。每个可见区放一个空节点，`data-block` 等于数据契约里的 id。程序按契约把当轮内容填进这些节点；JS 只做拓扑图和本地切换，通过 `present.onData(cb)` 拿当轮数据；不要用 `textContent` 写 markdown/text 正文。根容器（如 `#app`）`width: 100%`，铺满宿主气泡/iframe，禁止整壳 `max-width` + `margin: 0 auto` 收成居中卡片；若要控阅读行长，只给正文柱设 `max-width`（可 `margin-inline: auto`）。气泡随内容增高，禁止 `100vh` 和整页或侧栏滚动条，滚动只留在外层对话。美化用配色、字体层级、边框、间距，以及 CSS 或简单内联 SVG，按本次能写稳的程度完成。地图所需的 SVG 从拓扑数据绘制，游玩模型不输出 SVG。
|
||||
6. 给一份非正史 `present.v1` 预览数据，覆盖所有可见区域与空态。
|
||||
|
||||
模板总量不超过 60 KB。材料足够便直接成稿；会改变阅读结构的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
@@ -51,12 +52,14 @@ feeds: gm,narrator
|
||||
1. 语义区域必须全部有去处；未启用区域不补齐，隐藏维护区不渲染。
|
||||
2. 正文是主阅读面。辅助状态、线索、地图和文字选项按用途确定层级，不平均分割注意力。
|
||||
3. 程序投影区只消费权威 block：变量、拓扑和程序结果不要求模型重复填写。
|
||||
4. HTML 只放稳定骨架；数据由 `present.onData` 注入。缺 block 时按合同省略或显示空态。
|
||||
5. Tab、折叠、缩放等纯本地交互可以使用；不得联网、提交表单、弹窗、顶层导航或向宿主发送用户行动。
|
||||
6. Markdown 使用宿主提供的安全渲染能力；动态文字使用安全 API 或转义后写入，不能拼接未清洗 HTML。
|
||||
7. 地图/关系图由 JS 根据结构化拓扑绘制；显示坐标只属于布局，不回写世界真相。
|
||||
8. 预览只用于验收前端，不写成开场白或正史。
|
||||
9. 美化落在配色、字体层级、边框、间距和简单内联 SVG 上，写进 CSS，按本次能写稳的程度完成。
|
||||
4. HTML 只放稳定骨架，每个可见区一个 `data-block` 空节点。程序按数据契约灌数；`present.onData` 只留给拓扑和本地交互。markdown/text 正文禁止用 `textContent` 写入（会挤掉换行并露出 `**`）；若本地必须写正文，用 `present.markdown` 再赋 `innerHTML`。缺 block 时按合同省略或显示空态。不要把气泡做成独立窗口。
|
||||
5. 外壳铺满宿主可用宽度（`width: 100%`、背景与顶栏跟宽走）。行长只限正文柱，不限整壳；缺则按宿主全宽排，不另造居中窄卡片。
|
||||
6. Tab、折叠、缩放等纯本地交互可以使用；不得联网、提交表单、弹窗、顶层导航或向宿主发送用户行动。
|
||||
7. Markdown 使用宿主提供的安全渲染能力；动态文字使用安全 API 或转义后写入，不能拼接未清洗 HTML。
|
||||
8. 地图/关系图由 JS 根据结构化拓扑绘制；显示坐标只属于布局，不回写世界真相。
|
||||
9. 预览只用于验收前端，不写成开场白或正史。示例灌数不为好看新造人名、地名、势力，用已有产物里的；文笔及格即可。
|
||||
10. 美化落在配色、字体层级、边框、间距和简单内联 SVG 上，写进 CSS，按本次能写稳的程度完成。
|
||||
11. 改稿以最新产物为底。用户只提美化时只改样式，`数据契约` 和 `区域落点` 保持不变；要换阅读结构必须用户明确提出。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -121,8 +124,8 @@ feeds: gm,narrator
|
||||
}
|
||||
],
|
||||
"frontend": {
|
||||
"html": "<div id=\"app\">完整稳定骨架</div>",
|
||||
"css": "完整样式；正文优先；含窄屏规则",
|
||||
"html": "<div id=\"app\"><article data-block=\"body\"></article><section data-block=\"本局其它可见区域 id\"></section></div>",
|
||||
"css": "完整样式；根容器 width:100%；正文优先；行长只限正文柱；含窄屏规则",
|
||||
"js": "使用 present.onData(render) 的完整渲染代码"
|
||||
},
|
||||
"示例灌数": {
|
||||
@@ -139,7 +142,7 @@ feeds: gm,narrator
|
||||
硬规则：
|
||||
1. 输出合法 JSON；`数据契约.blocks` 必须覆盖回复呈现的全部用户可见区域，id 保持一致。
|
||||
2. 程序投影区不得改成模型生成；隐藏维护区只能进入 `不渲染`，不得出现在 DOM 或预览 blocks。
|
||||
3. `frontend.html/css/js` 均须给出可执行内容；合计不超过 60 KB。JS 必须调用 `present.onData(render)`。
|
||||
3. `frontend.html/css/js` 均须给出可执行内容；合计不超过 60 KB。每个可见 block 对应一个 `data-block` 空节点，id 与契约一致。JS 可以调用 `present.onData` 画拓扑或做本地切换，但文本区由程序填充。根容器铺满宿主宽度；不要整壳 `max-width` 居中卡片；不要 `100vh`，不要整页或侧栏滚动条。
|
||||
4. HTML 不嵌入本轮数据；动态文字安全渲染。禁止网络、表单、弹窗、顶层导航和宿主行动回传。
|
||||
5. topology 渲染读取结构化节点与连接；不得要求游玩模型输出 SVG。
|
||||
6. `示例灌数` 是非正史预览，schema 固定 `present.v1`，只含用户可见 blocks。
|
||||
@@ -156,9 +159,9 @@ feeds: gm,narrator
|
||||
- 1：漏区，或擅自添加回复呈现没有的内容。
|
||||
|
||||
阅读结构：
|
||||
- 5：正文始终是主阅读面，辅助信息层级与响应式布局服务本局体验。
|
||||
- 3：总体可读，但某个并列区略抢正文或窄屏处理不足。
|
||||
- 1：区域平均堆叠，阅读焦点混乱。
|
||||
- 5：正文始终是主阅读面；外壳铺满宿主宽度；辅助层级与响应式服务本局体验。
|
||||
- 3：总体可读，但并列区略抢正文、窄屏不足，或整壳被限宽成居中卡片。
|
||||
- 1：区域平均堆叠，阅读焦点混乱，或外壳明显窄于宿主。
|
||||
|
||||
权威绑定：
|
||||
- 5：变量、拓扑和程序结果只读权威投影，模型生成区边界清楚。
|
||||
@@ -182,7 +185,7 @@ feeds: gm,narrator
|
||||
- [ ] 回复呈现的全部可见区域均有 block、落点与渲染方式？
|
||||
- [ ] 隐藏维护区未进入 DOM 或预览？
|
||||
- [ ] 程序投影的变量和拓扑没有要求模型重写？
|
||||
- [ ] HTML/CSS/JS 完整，JS 使用 present.onData(render)，总量不超 60 KB？
|
||||
- [ ] HTML/CSS/JS 完整，每个可见区有 `data-block` 空节点，总量不超 60 KB，根容器铺满宿主宽度，正文不用 textContent 灌 markdown，没有整壳居中窄卡片或独立滚动窗口？
|
||||
- [ ] 美化已用配色、字体层级、边框、间距或简单 SVG 写进 CSS，程度限于本次能写稳的部分？
|
||||
- [ ] 本地交互不联网、不回传行动，动态内容安全渲染？
|
||||
- [ ] present.v1 预览覆盖可见区域并明确非正史？
|
||||
@@ -196,5 +199,5 @@ feeds: gm,narrator
|
||||
- 正文 + 地图：JS 根据 topology block 绘制 SVG；节点名称安全转义，窄屏可切到列表降级。
|
||||
- 多个辅助区：确有必要时使用本地 Tab；切换只改变当前气泡显示，不向宿主发送消息。
|
||||
- 武侠阅读：纸色、墨色、细线框和一条简单 SVG 纹样；只保留本局启用的区。
|
||||
- 不合适：把隐藏 state_ops 做成折叠面板；只写“机械风界面”却不给代码；让模型每轮输出地图 SVG。
|
||||
- 不合适：把隐藏 state_ops 做成折叠面板；只用 textContent 写正文导致不分段、露出 `**`；只写“机械风界面”却不给代码；让模型每轮输出地图 SVG。
|
||||
```
|
||||
|
||||
@@ -39,9 +39,9 @@ feeds: gm,narrator
|
||||
- 对应的上游引用与正向内容要求。
|
||||
- 没有内容时是省略、给空态，还是仍须输出。
|
||||
|
||||
正文区域只引用 `设计.故事推进` 的单轮篇幅与扩写/续写预算，不复制数值。状态、地图等权威内容使用程序投影，正文模型不重复生成。选项若启用，只定义文字建议的组成规则，不声明点击交互。
|
||||
正文区域只引用 `设计.故事推进` 的单轮篇幅与扩写/续写预算，不复制数值。状态、地图等权威内容使用程序投影，正文模型不重复生成。选项若启用，写清数量、各项的差异维度（如冒险/稳重）和每项的组成规则，让选项彼此不同；不声明点击交互。
|
||||
|
||||
简单变量采用 `inline_tool / inline_structured` 内联维护时，回复合同包含用户不可见的变量更新语义区；程序优先把它投递为 tool call，工具不可用时改为结构化 `state_ops`。`auditor_bundle` 由副 Worker 维护，不建立主回复更新区。
|
||||
表维护由旁观副 LLM 固定承担；本步**不**建立主回复里的隐藏变量更新区（无 `state_ops` / maintenance_channel）。可见区只投影权威数据。
|
||||
|
||||
材料足够便直接成稿；真正会改变区域合同的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -54,10 +54,10 @@ feeds: gm,narrator
|
||||
3. 模型生成区写内容要求；程序投影区写权威 ref 与字段选择。程序已有的数据不让模型复述一份。
|
||||
4. 正文篇幅、扩写和续写预算只引用故事推进；文风只引用叙事指南。
|
||||
5. 状态与拓扑的真相来自运行数据。可见区域只选择投影字段，不在这里修改变量或拓扑。
|
||||
6. 简单变量的隐藏更新也是语义区域；运行期可投递为 tool call 或结构化 `state_ops`。副 Worker 路径不在主回复伪造空区域。
|
||||
6. 不在主回复建隐藏填表区；表由旁观副 LLM 维护。
|
||||
7. 选项是文本建议。是否有交互、点击后做什么，只有正文组成和宿主前端真正实现时才声明。
|
||||
8. 区域要求使用正向规格；没有特殊要求便保持短，不为完整感堆区域。
|
||||
9. 每个可见区属于一种：抬头、正文、小故事、数据卡、结构、文字选项。抬头是页首一行短内容；条目一多或当前值一长则改用数据卡。数据卡和结构要写出露出的列或字段。隐藏的变量更新不占这些区。
|
||||
9. 每个可见区属于一种：抬头、正文、小故事、数据卡、结构、文字选项。抬头是页首一行短内容；小故事是旁支短文，不推进主线，也不复述正文；条目一多或当前值一长则改用数据卡。数据卡和结构要写出露出的列或字段。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -119,6 +119,33 @@ feeds: gm,narrator
|
||||
"正向说明该区域应该提供什么"
|
||||
],
|
||||
"空值处理": "omit|empty_state|error"
|
||||
},
|
||||
{
|
||||
"id": "choices",
|
||||
"名称": "文字选项",
|
||||
"用途": "给用户几条方向不同的行动参考",
|
||||
"用户可见": true,
|
||||
"出现": {
|
||||
"方式": "every_turn",
|
||||
"条件": ""
|
||||
},
|
||||
"内容类型": "list",
|
||||
"区域种类": "文字选项",
|
||||
"来源": {
|
||||
"方式": "model_generated",
|
||||
"ref": "设计.故事推进"
|
||||
},
|
||||
"选项": {
|
||||
"数量": 2,
|
||||
"各项": [
|
||||
{ "差异维度": "冒险", "组成": "一句推进风险或探索的具体行动" },
|
||||
{ "差异维度": "稳重", "组成": "一句观察、保守或缓和局面的具体行动" }
|
||||
]
|
||||
},
|
||||
"内容要求": [
|
||||
"每项是用户角色此刻能直接采取的一句行动，按各自差异维度写，彼此不重复"
|
||||
],
|
||||
"空值处理": "omit"
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -130,8 +157,8 @@ feeds: gm,narrator
|
||||
2. 普通楼层至少含 `body`；正文只引用故事推进，不复制目标字数、占比或字数预算。
|
||||
3. 每个区域必须声明出现方式、内容类型、区域种类、来源、内容要求和空值处理。可见区的区域种类只能是抬头、正文、小故事、数据卡、结构、文字选项。
|
||||
4. `program_projection` 必须给稳定 ref；状态和拓扑数据不得标为临场或要求模型重新生成。
|
||||
5. 内联变量更新区必须 `用户可见:false`、内容类型为 `state_ops`、来源为 `maintenance_channel`；程序按 API 能力选择 tool call 或结构化区域。`auditor_bundle` 不建立该区。
|
||||
6. 文字选项只规定数量、差异维度和内容要求；不写可点击、按钮动作或回传协议。
|
||||
5. 不建立主回复隐藏填表区（无 `state_ops` / `maintenance_channel`）；表由旁观副 LLM 维护。
|
||||
6. 文字选项须含 `选项`：数量，以及每项的差异维度和组成；不写可点击、按钮动作或回传协议。
|
||||
7. 不选壳，不写布局、HTML、CSS、JS、Tab、SVG 或槽位替换；这些属于正文组成。
|
||||
8. 产物正文不含自评、追问或开放问题。
|
||||
9. `mount` 固定为 `["world-simulator", "narrator"]`；artifact tag 暂时保留 `设计.监控栏`。
|
||||
@@ -184,4 +211,5 @@ feeds: gm,narrator
|
||||
- 展示地图：新增 topology 区域，说明用户应看到哪些节点、连接或当前位置；具体 SVG、表格或 Tab 留给正文组成。
|
||||
- 简单变量内联维护：新增用户不可见的 state_ops 语义区域；运行期优先投递为 tool call，工具不可用时进入结构化整包。
|
||||
- 文字选项：写数量、彼此差异和应提供的行动方向；不声明点击行为。
|
||||
- 坏：只写「要选项」，不定数量、各项差异维度与组成，结果每轮几条选项大同小异。
|
||||
```
|
||||
|
||||
@@ -7,33 +7,37 @@ name: 故事推进
|
||||
id: story-progression
|
||||
artifact: 设计.故事推进
|
||||
declaration: >
|
||||
从偏好库选取本局约束，只选定一种用户输入用法，并写清单轮目标字数、扩写与续写的定义和篇幅分配、OOC 处理；不写遣词。
|
||||
继承交互范式的描写权限与停笔时机，从偏好库选取本局约束，钉推进方式（可无专门推进），只选定一种用户输入用法；每一轮正文分成扩写（演本轮材料）与续写（材料后往下写），写清定义、字数分配与 OOC 处理；不写遣词。
|
||||
when: |
|
||||
已有美学纲领（或等价），但尚未钉死唯一的用户输入用法，或扩写/续写边界与比例仍不清。
|
||||
用户主动谈推进、节奏、停笔、大纲扩写、续写、自动写入时也使用。
|
||||
已有美学纲领（或等价），但尚未钉死唯一的用户输入用法，或扩写/续写边界与比例仍不清，或还不知道故事靠场面自推、等用户决策还是无专门推进。
|
||||
用户主动谈推进、节奏、停笔、大纲扩写、续写、自动写入、单轮字数时也使用。
|
||||
when_not: |
|
||||
本技能只给对话落盘。工序编排不排本步（编排池是「叙事指南与故事推进」）。
|
||||
美学未钉 → 先「美学纲领与交互范式」。
|
||||
只改遣词、笔墨、写法档 → 「叙事指南」。
|
||||
只改遣词、笔墨、写法档、段落形态 → 「叙事指南」。可见区域 → 「回复呈现」。
|
||||
好感公式、进度表、事件池等表规格 → 「变量设计与更新规则」/「生成规则」。
|
||||
固定输入标记（裸写 / `#行动#` / 引号）由程序解析，不在本步重定义或按局反转。
|
||||
boundary: |
|
||||
本能力只交推进合同：唯一输入用法、单轮目标字数、扩写要求、续写要求、篇幅分配、OOC 处理策略。
|
||||
叙事指南管怎么写；变量与生成规则管表与合同；主角详略只写偏好钩子。
|
||||
本能力只交推进合同：推进方式及可选的驱动力、成败与难度、变量钩子；唯一输入用法、单轮目标字数、扩写要求、续写要求、篇幅分配、OOC 处理策略。
|
||||
单轮字数与篇幅只在本步定义；段落形态（长段不碎行）归叙事指南，可见区域归回复呈现。
|
||||
叙事指南管怎么写；变量与生成规则管表与合同，本步只写钩子；主角详略只写偏好钩子。
|
||||
feeds: gm,narrator
|
||||
```
|
||||
|
||||
## opening
|
||||
|
||||
```opening
|
||||
先选定这局唯一的输入用法：大纲扩写、续写，或自动写入。再说清楚单轮大约写多长、扩写和续写各做什么及各占多少；上帝视角/OOC 指令准备怎么处理。
|
||||
先选定这局唯一的输入用法：大纲扩写、续写，或自动写入。再说清楚单轮大约写多长；扩写（主路径不动，在用户已给的节拍之间补连接小动作，把场面写顺）和续写（从材料结束后往下写他人与世界反应）各做什么、各占多少。上帝视角/OOC 指令准备怎么处理。故事要不要专门往前推、用户会不会失败，也可以一并说；没想过就按体验先定。
|
||||
```
|
||||
|
||||
## task
|
||||
|
||||
```task
|
||||
把本局推进合同写入 `设计.故事推进`。先用 `read_library_entry` 从绑定的偏好库读取最相关 1～2 条并本局化，再只选定一种用户输入用法，按该用法写清单轮目标字数、扩写定义、续写定义、两者各自篇幅预算，以及裸写（上帝视角/OOC）指令如何处理。
|
||||
把本局推进合同写入 `设计.故事推进`。先继承上游「美学纲领与交互范式」：把 `交互范式.描写权限`（思考与抉择、言语、主动行动、用户输入后）、`后果与叙事控制` 与 `停笔时机` 改写成推进做法；交互范式已写过停笔权限，也要落进本步的用法与续写要求，不因已有而省略。再用 `read_library_entry` 从绑定的偏好库读取最相关 1～2 条并本局化，然后钉推进方式（不明则写「无专门推进」），只选定一种用户输入用法。
|
||||
|
||||
禁止混用两种用法或写成半套规则。固定标记怎么识别交给程序；本步只写识别之后怎么服从与落笔。不写遣词、示范段落或版式。
|
||||
扩写与续写是每一轮正文里的两段工作，不是两种输入用法。按所选用法写清单轮目标字数、扩写做什么、续写做什么、两者各自篇幅预算，以及裸写（上帝视角/OOC）指令如何处理。
|
||||
|
||||
禁止混用两种用法或写成半套规则。固定标记怎么识别交给程序；本步只写识别之后怎么服从与落笔。不写遣词、段落形态、示范段落或版式。
|
||||
|
||||
材料足够便直接成稿；多种用法都像、必须用户点选时，把缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -41,16 +45,28 @@ feeds: gm,narrator
|
||||
## principles
|
||||
|
||||
```principles
|
||||
1. 全局只允许一种用户输入用法。切分依据是扩写与续写的定义，以及两者在本局的职责和篇幅预算。
|
||||
2. 三种用法互斥，写入时只保留一个完整合同：
|
||||
- 大纲扩写：输入是待演成场面的大纲。扩写要把每个节拍演满，节拍上的人、事、话、结果保持不变，节拍之间补具体动作把处境写出来；续写短，且不代用户做新决定。
|
||||
- 续写：扩写只跟用户本轮已声明的行为与对话；续写为主，停在其他角色等待用户抉择处，不替用户做态度或推进性决定。
|
||||
- 自动写入：输入多为短 meta 或极简扮演；续写宜长；可按主角设定代演用户角色言行。
|
||||
3. 多写“要怎样完成扩写/续写”，少写否定清单；只有会破坏该用法边界的禁点才保留。
|
||||
4. OOC/上帝视角内容由程序从裸写识别；本步写处理策略：服从到什么程度、是否打断当前场面、如何与既定扩写/续写比例兼容。
|
||||
5. 偏好库是参考：读取后改成本局可执行要求，不照抄。
|
||||
6. 单轮篇幅在这里一次钉死：给目标字数或窄区间，并把扩写、续写的占比与字数预算对应起来；其它节点直接消费，不重复定义。
|
||||
7. 不写遣词、写法档、版式或事件池公式。
|
||||
1. `用户输入用法` 与「扩写 / 续写」不是同一层。用法是本局选哪一种合同（三选一）；扩写、续写是每一轮正文里先后做的两段工作。产物必须先按所选用法写清这两段各自做什么，再分配篇幅。
|
||||
2. 扩写：主路径不动，把用户本轮已经给出的材料演成流畅场面。人、事、话、结果保持不变；允许在节拍之间增加使动作接得上的小动作、感官和当场小纠正。只加形容词、比喻或气氛、不增加这些连接动作，不算扩写。另起新冲突、新目标、新决定，或改掉已定结果，算出界。
|
||||
例：用户写「打开冰箱门、放入大象、关闭冰箱门」——主路径就是这三步，结果仍是大象进了冰箱。只写这三句会不流畅，扩写应补门怎么拉开、大象怎么塞进去、门关不上再用力扣上等细节；不能改成没放进去、或中途改去别的房间。
|
||||
- 用法为「大纲扩写」时：材料是待演成场面的大纲，不是终稿句子。每个节拍都要演到；节拍之间必须补连接动作，大纲没写这些动作也要写。另例：大纲「我去相亲、坐了下来，她说要婚后自由出轨权，我拒绝了」——坐下前写先要一杯咖啡，又急忙改成两杯，显出局促；她的要求和拒绝的结果不动。
|
||||
- 用法为「续写」或「自动写入」时：材料是用户本轮已声明的行为与对话。扩写只在这些声明之间补过程与连接动作，可写他人怎么接这一下；不另写用户没声明的新决定。
|
||||
3. 续写：从用户材料结束后往下写。写其他角色与世界如何反应、把场面推到下一个停笔点。默认不可再写用户角色的实际行动、说话或带倾向抉择；最多被动占位（望去、让开）。整轮停在其他角色等用户角色行动抉择之刻。
|
||||
- 用法为「大纲扩写」或「续写」时：禁止替用户角色做决定，不引入材料外的新剧情事实。
|
||||
- 用法为「自动写入」时：可按主角设定代演用户角色言行，把续写写长；关键抉择仍按交互范式的权限处理。
|
||||
4. 三种输入用法互斥，按体验与交互范式选一种，写入时只保留一个完整合同。差别在输入是什么、扩写/续写谁长谁短、续写能不能代演用户角色：
|
||||
- 大纲扩写：用户是写手、要大纲成稿，输入多为剧情节拍时选它。扩写充分、续写短。
|
||||
- 续写（旧称扮演停笔）：代入扮演、轮转等用户决策时选它（交互范式规定关键抉择或言语只由用户做、停笔时机是等用户回应）。输入往往较短；扩写只跟已声明内容，续写多于大纲扩写、短于自动写入。
|
||||
- 自动写入：用户只给短 meta 指令或极简扮演、希望 AI 代演用户角色，且交互范式允许模型代写 user 言行（旁观、操控站位常见）时选它。扩写少、续写宜长。
|
||||
5. 多写“要怎样完成扩写/续写”，少写否定清单；只有会破坏该用法边界的禁点才保留。
|
||||
6. 推进可轻可无：`推进方式` 写「场面自推 / 等用户决策 / 无专门推进」之一，不明则写「无专门推进」。不自己发明好感公式、进度表或事件池；表规格交给「变量设计与更新规则」与「生成规则」。
|
||||
7. 推进子维度按需填，推进方式为「无专门推进」时可全省：
|
||||
- `驱动力`：故事靠谁、靠什么往前（NPC 主动作为、世界压力与时限、用户决策），承接交互范式的 `世界对user的影响` 与 `引入新内容`。
|
||||
- `成败与难度`：用户行动会不会失败、失败怎么写（代价落在哪、挫败场面写多重），承接 `行为后果判定` 与 `设计困境`。体验本身要顺遂碾压时写明「不设失败」以及仍保留的阻力；未写明时不默认无条件顺遂。
|
||||
- `变量钩子`：推进挂在哪些变量上，每条一句「变量名：推进如何看它」；只写钩子，不写公式、阈值或数值表。
|
||||
8. OOC/上帝视角内容由程序从裸写识别；本步写处理策略：服从到什么程度、是否打断当前场面、如何与既定扩写/续写比例兼容。
|
||||
9. 偏好库是参考：读取后改成本局可执行要求，不照抄。
|
||||
10. 单轮篇幅在这里一次钉死：给目标字数或窄区间，并把扩写、续写的占比与字数预算对应起来；其它节点直接消费，不重复定义。段落形态归叙事指南，本步不写。
|
||||
11. 不写遣词、写法档、段落形态、版式或事件池公式。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -70,9 +86,18 @@ feeds: gm,narrator
|
||||
}
|
||||
],
|
||||
"推进方式": "场面自推 / 等用户决策 / 无专门推进",
|
||||
"驱动力": "可选：故事靠谁、靠什么往前；无专门推进时省略",
|
||||
"成败与难度": "可选：用户行动会不会失败、失败怎么写、代价落在哪；无专门推进时省略",
|
||||
"变量钩子": ["可选：变量名：推进如何看它（只写钩子，不写公式）"],
|
||||
"用户输入用法": "大纲扩写 | 续写 | 自动写入",
|
||||
"扩写要求": ["该用法下扩写要做什么、守住什么结果"],
|
||||
"续写要求": ["该用法下续写写到哪里、何时停笔"],
|
||||
"扩写要求": [
|
||||
"主路径（人、事、话、结果）不动；节拍之间补连接小动作，把场面写顺。例：打开冰箱、放入大象、关上冰箱——补怎么塞、门怎么扣，不能改成没放进去",
|
||||
"所选用法下材料是什么、节拍之间补什么（按 principles 第 2 条写具体）"
|
||||
],
|
||||
"续写要求": [
|
||||
"从材料结束后写他人与世界反应，停在等用户角色回应之刻",
|
||||
"所选用法下能否代演用户角色、续写写到哪停（按 principles 第 3 条写具体）"
|
||||
],
|
||||
"单轮篇幅": {
|
||||
"目标字数": "明确数字或窄区间",
|
||||
"扩写占比": "百分比",
|
||||
@@ -89,24 +114,25 @@ feeds: gm,narrator
|
||||
|
||||
硬规则：
|
||||
1. 输出合法 JSON；`用户输入用法` 只能是三者之一，不得并列或混装半套。
|
||||
2. `扩写要求`、`续写要求`、`单轮篇幅` 必须与所选用法一致；两项占比合计 100%，字数预算与目标字数对应。
|
||||
3. `选用偏好` 默认 1～2 项；无合适条目时可 `[]`，但仍须钉死唯一用法。
|
||||
4. 不重定义裸写 / `#行动#` / 引号的识别规则；只写 OOC 处理策略。
|
||||
5. 不在产物正文写自评、追问、开放问题或遣词示范。
|
||||
6. `mount` 固定包含 `world-simulator` 与 `narrator`。
|
||||
2. `扩写要求` 只写「把本轮材料演成场面」怎么做；`续写要求` 只写「材料结束后往下写」怎么做。不要把另一种输入用法的规则抄进这两项。两项占比合计 100%，字数预算与目标字数对应。
|
||||
3. `推进方式` 是单个字符串；`驱动力`、`成败与难度`、`变量钩子` 可选，`变量钩子` 不写公式或数值。
|
||||
4. `选用偏好` 默认 1～2 项；无合适条目时可 `[]`，但仍须钉死唯一用法。
|
||||
5. 不重定义裸写 / `#行动#` / 引号的识别规则；只写 OOC 处理策略。
|
||||
6. 不在产物正文写自评、追问、开放问题、遣词示范或段落形态。
|
||||
7. `mount` 固定包含 `world-simulator` 与 `narrator`。
|
||||
|
||||
## score
|
||||
|
||||
```score
|
||||
唯一性与切分：
|
||||
- 5：只选定一种用法；扩写/续写定义、目标字数和预算切分清楚，无第二种用法残留。
|
||||
- 5：只选定一种用法，且与交互范式的权限和停笔时机相符；扩写（演本轮材料）与续写（材料后往下写）定义清楚，目标字数和预算切分清楚，无第二种用法残留。
|
||||
- 3：主用法明确，但仍有一处边界、占比或字数预算含糊。
|
||||
- 1：混用多种用法，或扩写/续写职责无法区分。
|
||||
- 1：混用多种用法，或把扩写/续写当成另一种输入用法、两段职责无法区分。
|
||||
|
||||
可执行性：
|
||||
- 5：按该合同可直接开玩，扩写、续写、停笔与 OOC 处理都能执行。
|
||||
- 3：大方向可用，但停笔或 OOC 处理仍需猜测。
|
||||
- 1：只有用法名称，没有可执行要求。
|
||||
- 5：按该合同可直接开玩，推进方式、扩写、续写、停笔与 OOC 处理都能执行；有推进时成败与难度写清，不默认无条件顺遂。
|
||||
- 3：大方向可用，但停笔、成败或 OOC 处理仍需猜测。
|
||||
- 1：只有用法名称，没有可执行要求，或自行发明公式与进度表。
|
||||
|
||||
偏好本局化：
|
||||
- 5：已读偏好并改成本局要求，与唯一用法咬合。
|
||||
@@ -117,10 +143,22 @@ feeds: gm,narrator
|
||||
## checklist
|
||||
|
||||
```checklist
|
||||
- [ ] 只选定一种用户输入用法？
|
||||
- [ ] 已继承交互范式的描写权限与停笔时机并落进本步？
|
||||
- [ ] 推进方式已写（不明则「无专门推进」），未发明公式、进度表或事件池？
|
||||
- [ ] 只选定一种用户输入用法，且符合它的适用场景？
|
||||
- [ ] 扩写与续写已按两段工作分别写清（扩写=演本轮材料，续写=材料后往下写），且与所选用法相符？
|
||||
- [ ] 扩写已写明「主路径不动、节拍之间补连接小动作」？只加形容词、或改掉已定结果，都不算写清。
|
||||
- [ ] 单轮目标字数、扩写/续写定义、占比与字数预算已写清且互相一致？
|
||||
- [ ] OOC/裸写处理策略已写，且未重定义固定标记语法？
|
||||
- [ ] 偏好已按需读取并本局化？
|
||||
- [ ] 未写入遣词、写法档或半套混用规则？
|
||||
- [ ] 未写入遣词、写法档、段落形态或半套混用规则？
|
||||
- [ ] 产物正文不含评分、诊断、追问？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：用法钉「续写」；用户写「我推开门」——扩写补门轴、屋里人抬头，不另写用户没说的台词；续写写屋里人的反应，停在等用户开口之刻。
|
||||
好：用法钉「大纲扩写」；用户写「打开冰箱门、放入大象、关闭冰箱门」——扩写补门怎么拉开、大象怎么塞、门关不上再扣上，三步结果不变；续写短。
|
||||
坏：把「扩写」理解成另一种输入用法；只把三句动作各加形容词、场面仍不接；中途改成没放进冰箱或另起新目标；续写里替用户角色补一句台词。
|
||||
```
|
||||
|
||||
@@ -7,7 +7,7 @@ name: 变量设计与更新规则
|
||||
id: variable-design
|
||||
artifact: 设计.变量设计与更新规则
|
||||
declaration: >
|
||||
定义必须跨轮保存的最少真值、合法更新条件、权限与派生映射，供程序选择简单状态工具或复杂维护副 Worker。
|
||||
定义必须跨轮保存的最少真值、合法更新条件、权限与派生映射；有真值时由程序固定启用旁观副 LLM 按准则填表。
|
||||
when: |
|
||||
某项状态若不持久保存，就会导致数值、阶段、资源、门控或已发生事实跨轮漂移时使用。
|
||||
when_not: |
|
||||
@@ -15,8 +15,8 @@ when_not: |
|
||||
节点与连接构成的地图、关系或路径 → 拓扑图谱。
|
||||
大段设定或同类对象合同 → 舞台骨架 / 生成规则。
|
||||
boundary: |
|
||||
本能力定义什么是真值、什么事实允许它改变，以及改变后的约束；不决定 API 投递方式。
|
||||
少量简单变量由程序生成主世界层状态工具；复杂变量与拓扑等由维护合同编译工具汇总。
|
||||
本能力定义什么是真值、什么事实允许它改变，以及改变后的约束。
|
||||
表维护固定走旁观副 LLM（读上一份表、本产物维护准则、本轮主世界正文）；不在此决定内联 tool / state_ops。
|
||||
变量控制上下文负责详细可见范围；正文组成负责展示。本能力不写隐藏段、正则、maintain 包或界面格式。
|
||||
feeds: gm,auditor
|
||||
```
|
||||
@@ -38,7 +38,7 @@ feeds: gm,auditor
|
||||
|
||||
若真值只用于选择分档文本或实例切片，保存真值并引用映射来源，不把派生结果建立成第二个可写真值。若跨越阈值会触发一次性或边沿效果，写成触发规则。
|
||||
|
||||
最后根据变量数量、相互依赖、判断方式和触发复杂度给出维护复杂度事实，供程序选择 `none / inline_tool / inline_structured / auditor_bundle`；这是建议，不绑定 API 能力。
|
||||
可写简短「维护复杂度」备注（交叉依赖、须结合正文推断的点），供旁观副 LLM 参考；**程序有真值即固定开旁观填表**，不再按建议在内联/旁观间切换。勿手写 `play_slots`。
|
||||
|
||||
材料足够便直接成稿；真正会改变变量合同的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
```
|
||||
@@ -52,8 +52,8 @@ feeds: gm,auditor
|
||||
4. 类型决定操作：number/integer 可 `set|delta`；enum/string/boolean 可 `set`；set 可 `add|remove`。不用散文值绕过类型。
|
||||
5. 公开性只区分用户可见与隐藏；更细的主世界层、转述、旁观投影交给变量控制上下文。
|
||||
6. 阈值映射只保存引用和选取条件；长文内容仍在生成规则/具体实例中。
|
||||
7. 维护复杂度按实际依赖判断，不机械按数量：一个简单余额可走 inline tool；一个可扩展关系图不属于简单变量。
|
||||
8. 不在正文中设计隐藏标签、正则或维护 JSON；投递方式由程序根据 API 能力选择。
|
||||
7. 「维护复杂度」只写给旁观看的难点备注（交叉依赖、叙事推断证据），不决定开不开旁观；有真值即旁观填表。
|
||||
8. 不在正文中设计隐藏标签、正则或维护 JSON；主世界不承担隐藏填表通道。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -133,7 +133,7 @@ feeds: gm,auditor
|
||||
}
|
||||
],
|
||||
"维护复杂度": {
|
||||
"建议": "inline_tool|auditor_bundle",
|
||||
"备注": "交叉依赖、须结合正文推断的点；给旁观副 LLM 参考，不驱动开槽",
|
||||
"依据": [
|
||||
"变量数量、交叉依赖、narrative_inference、触发规则与其它可变产物"
|
||||
],
|
||||
@@ -150,7 +150,7 @@ feeds: gm,auditor
|
||||
2. 类型与允许操作必须相容；number/integer 的边界、enum 的枚举值必须明确；不适用的约束字段省略。
|
||||
3. `narrative_inference` 必须写证据要求；正文没有该证据时不更新。
|
||||
4. 派生映射与触发规则可为空；存在时必须引用已声明变量，且派生结果不得成为第二个可写真值。
|
||||
5. `维护复杂度.建议` 只在 `inline_tool|auditor_bundle` 中选择；程序仍按 API 能力决定具体投递方式。
|
||||
5. `维护复杂度` 可选；有则写备注与依据给旁观，勿再写 `建议: inline_*` 指望关旁观。有真值时程序固定 `play_slots.auditor=true`。勿手写 `play_slots`。
|
||||
6. 不输出隐藏段、正则、`maintain.v1/v2`、tool schema、Worker 提示或界面字段。
|
||||
7. 产物正文不含自评、追问或开放问题。
|
||||
8. `mount` 固定为 `["world-simulator", "auditor"]`。
|
||||
@@ -191,15 +191,15 @@ feeds: gm,auditor
|
||||
- [ ] 类型、初值、允许操作、约束和写入者互相一致？
|
||||
- [ ] 每条更新规则都有事实证据和可执行的值规则？
|
||||
- [ ] 派生映射未被重复建成可写真值？
|
||||
- [ ] 维护复杂度依据实际依赖，而不是只按变量数量？
|
||||
- [ ] 维护复杂度若写了，是旁观备注而非开槽开关？
|
||||
- [ ] 未写隐藏标签、正则、维护包或界面格式？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
- 单个数值真值、只有 set/delta 和上下界、无交叉依赖：建议 inline_tool；程序按 API 探测生成类型明确的小工具。
|
||||
- 多个真值需要结合正文共同判断，且阈值会触发映射或补实例：建议 auditor_bundle；维护编译工具统一收集。
|
||||
- 单个数值真值或关系矩阵：都写清更新规则；有真值即由旁观副 LLM 填表，主世界只写正文。
|
||||
- 多个真值需结合正文共同判断：在维护复杂度.备注里点出推断难点与证据，供旁观读。
|
||||
- 当前态度由好感区间映射得到：只保存好感，态度作为派生映射，不建立第二个可写“当前态度”。
|
||||
- 不合适：每轮重写一段“当前性格”长字符串，或只写“随剧情变化”而没有证据与操作。
|
||||
```
|
||||
|
||||
@@ -14,7 +14,7 @@ artifact: 设计.worker规格
|
||||
declaration: >
|
||||
程序根据前面已经确认的写作、维护和秘密隔离需求，整理本局需要启用的固定执行单元
|
||||
when: |
|
||||
落档或开玩时自动执行；从叙事、维护复杂度、信息隔离和写作路径中推导固定槽位。
|
||||
落档或开玩时自动执行；从叙事、信息隔离、写作路径推导固定槽位；有变量真值时程序固定开旁观填表。
|
||||
when_not: |
|
||||
前置语义尚未确定时沿用安全默认值，不向用户重复询问。
|
||||
boundary: |
|
||||
@@ -57,7 +57,8 @@ boundary: |
|
||||
2. 默认世界模拟：`gm: true, narrator: false, auditor: false, perspective: false`；
|
||||
调度序 gm → perspective? → narrator? → auditor?；
|
||||
主世界层必开（用户明确只要旁观/写手路径除外）；
|
||||
要独立文风改写才 narrator true；有变量设计/表维护需求 → auditor true；
|
||||
要独立文风改写才 narrator true；
|
||||
**旁观填表由程序固定**：有「设计.变量设计与更新规则」真值时落档会开 `auditor=true`，本步不必猜；手写 play_slots 时旁观与之对齐即可。
|
||||
仅信息隔离才开 perspective。
|
||||
3. 写手路径：`outline` + `chapter-writer` 默认都开；用户明确只要正文则可关 outline。
|
||||
4. 可写简短 `mount_notes`（哪类上游产物挂哪槽），不粘贴长文。
|
||||
@@ -70,8 +71,8 @@ boundary: |
|
||||
|
||||
```principles
|
||||
1. 只勾选，不发明：ref 必须在白名单内；无机遇裁定槽。
|
||||
2. 默认：世界模拟 = 仅主世界层；转述/旁观/perspective 默认关。主世界层必要。
|
||||
3. 要独立文风层再开转述；有变量/表/规则要盯再开旁观。
|
||||
2. 默认：世界模拟 = 仅主世界层；转述/perspective 默认关。主世界层必要。
|
||||
3. 要独立文风层再开转述。表维护固定旁观副 LLM：有变量真值即 auditor true，不再按 inline 建议关旁观。
|
||||
4. 变量 / Data / Progressive 不是执行单元；旁观维护只出 maintain.v1，不写真相。
|
||||
5. 删掉检验仍适用：关某个槽要说得清损失什么。
|
||||
6. 本步不输出完整 Worker 集、不写 tables 全文、不写随机范围表（交给细化终稿 / 变量设计 / 随机范围整理）。
|
||||
@@ -85,10 +86,9 @@ boundary: |
|
||||
|
||||
- 是否必须独立转述改写正文？（要才 narrator=true；默认不要，主世界层原文即终稿）
|
||||
- 是否有「主世界层不该知道的角色秘密」？（有才 perspective=true）
|
||||
- 有变量/表要盯却关了旁观？或无状态却硬开旁观？
|
||||
- 扩写：要不要先验大纲再写章？
|
||||
|
||||
不要问「还想加什么 Worker」；不要问要不要机遇裁定。
|
||||
不要问「还想加什么 Worker」；不要问要不要机遇裁定；旁观开否由程序按有无变量真值固定。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -123,7 +123,7 @@ boundary: |
|
||||
- `path=writing` 时必须有 `writing_slots`；`play_slots` 可省略。
|
||||
- `play_slots.gm` 世界模拟路径下应为 true。
|
||||
- `narrator` 缺省按 false（主世界层兼呈现）；仅用户明确要独立文风层时 true。
|
||||
- `auditor` 缺省按 false；有变量/表维护意图才 true。
|
||||
- `auditor` 缺省按 false；有变量真值时落档程序固定为 true（旁观副 LLM 填表）。对话落盘不必本步猜。
|
||||
- 不要输出 `chance` 字段；旧产物若有 `chance: true`，本步改为 false/删除，并指向「随机范围整理」。
|
||||
- 不要输出自造 `actors[]` / 自由 `workers[]`。
|
||||
- 旧产物若含 `actors[]`：本步应改写为槽位勾选，不再追加自定义 ref。
|
||||
@@ -132,8 +132,8 @@ boundary: |
|
||||
|
||||
```checklist
|
||||
- [ ] 是否只有白名单槽，无自造 ref、无机遇裁定？
|
||||
- [ ] 主世界层是否开启（世界模拟路径）？旁观是否与变量需求一致？
|
||||
- [ ] auditor / perspective / 只要正文等非常规选择是否有理由？
|
||||
- [ ] 主世界层是否开启（世界模拟路径）？有变量真值时旁观是否按程序约定开启？
|
||||
- [ ] perspective / 只要正文等非常规选择是否有理由？
|
||||
- [ ] 是否误把变量管理做成槽？是否误交完整 设计.worker集？
|
||||
```
|
||||
|
||||
|
||||
@@ -12,8 +12,7 @@ when: |
|
||||
核心体验依赖外部环境、社会力量或世界状况持续作出一致反馈，而现有材料还不足以支撑这些反馈时使用。
|
||||
是否换区不是判断标准：跨越多地可以需要舞台骨架；一室之内的朝堂、审讯或家庭关系同样可以需要。
|
||||
when_not: |
|
||||
熟悉现实基底已经够用，地点和外部格局不会改变核心体验或世界反馈时不使用。
|
||||
既成硬法则归实现机制；多份人物或要先审字段的名片归生成规则，单条字段很少的具名内容才归具体实例；重点在节点与连接本身时归拓扑图谱。
|
||||
熟悉现实基底已经够用，地点和外部格局不会改变核心体验或世界反馈时不使用。 既成硬法则归实现机制；多份人物或要先审字段的名片归生成规则，单条字段很少的具名内容才归具体实例；重点在节点与连接本身时归拓扑图谱。
|
||||
boundary: |
|
||||
本能力只交会上台并反复有用的外部事实，其余沿用熟悉常识或明确留白。
|
||||
不重写核心体验，不解释硬法则起源，不制作人物名片、完整地图、历史或百科。
|
||||
@@ -38,11 +37,12 @@ feeds: gm
|
||||
|
||||
```principles
|
||||
1. 舞台以世界怎样回应玩家为准，不以是否换区、地图大小或题材类型为准。
|
||||
2. 熟悉常识默认继承，只写相对基底的必要变造，以及会反复参与体验的外部事实。
|
||||
3. 核心运转可以是社会力量、阶层关系、周期事件、资源条件、禁忌或日常秩序；只说明它如何作用于玩家和主要舞台。
|
||||
4. 主要舞台可以是空间、机构、社会层或一间房；详细度以能稳定反馈为止。
|
||||
5. 既成硬法则只引用其结果，规则本身交给实现机制；多份或要先审字段的具名人物交给生成规则，单条字段很少的具名内容才交给具体实例。
|
||||
6. 用户未触及的地区、历史、物种、民俗和势力不补全，在留白边界中说明即可。
|
||||
2. 尺度由核心体验会碰到的边界决定。同是“现代都市”基底：体验作为韩国顶级财阀，尺度升到地球级势力、国家与资本网络，主要舞台写到财团、宴会、听证层；体验普通大学生活，尺度是一座城或一所学校，主要舞台写到系馆与社团层。按体验正推，写“需要跨国资本压迫感 → 升到国家层”，不写“因为是校园文所以不要国际政治”。
|
||||
3. 熟悉常识默认继承，只写相对基底的必要变造，以及会反复参与体验的外部事实。
|
||||
4. 核心运转可以是社会力量、阶层关系、周期事件、资源条件、禁忌或日常秩序；只说明它如何作用于玩家和主要舞台。用户已点名的灾变、祭祀、异界出口等世界状况必须写进核心运转。
|
||||
5. 主要舞台可以是空间、机构、社会层或一间房；写到能稳定反馈为止：玩家在这里做什么、会碰到谁或什么、得到什么反馈，不只点名。
|
||||
6. 既成硬法则只引用其结果，规则本身交给实现机制；多份或要先审字段的具名人物交给生成规则，单条字段很少的具名内容才交给具体实例。多张名片共享的台上格局与世界状况写在本能力，生成规则只引用，不在规则里重抄。
|
||||
7. 用户未触及的地区、历史、物种、民俗和势力不补全，在留白边界中说明即可。类型（西幻、修仙、末日等）只作体裁与氛围参照，不是必选题单：不因题材自动塞入该类型常见的地理与势力；不为显得严谨补起源史、伪科学或神话谱系，除非它们本身会上台。
|
||||
```
|
||||
|
||||
## output
|
||||
@@ -57,6 +57,7 @@ feeds: gm
|
||||
"正文": {
|
||||
"舞台基底": {
|
||||
"熟悉基底": "现实或常见类型中可直接继承的最短描述",
|
||||
"尺度": "体验会碰到的边界，如国家与资本网络、一座城、一所学校、一间殿；熟悉基底已说明时省略",
|
||||
"必要变造": ["只列会改变世界反馈的差异；无则省略"]
|
||||
},
|
||||
"核心运转": [
|
||||
@@ -69,7 +70,7 @@ feeds: gm
|
||||
"主要舞台": [
|
||||
{
|
||||
"名称": "会反复上台的空间、机构或社会层",
|
||||
"玩家如何接触": "玩家在这里做什么、受到什么反馈",
|
||||
"玩家如何接触": "玩家在这里做什么、会碰到谁或什么、受到什么反馈；写到能稳定反馈的具体程度，不只点名",
|
||||
"关键运转": ["仅引用与此处直接相关的核心运转"]
|
||||
}
|
||||
],
|
||||
@@ -96,7 +97,7 @@ feeds: gm
|
||||
克制度：
|
||||
- 5：只写会上台且反复有用的内容，熟悉常识默认继承，其余明确留白。
|
||||
- 3：夹带少量暂时不会影响游玩的地图、历史或设定。
|
||||
- 1：写成完整世界百科，核心舞台反而不突出。
|
||||
- 1：写成完整世界百科，或因题材自动塞满该类型常见地理与势力，核心舞台反而不突出。
|
||||
```
|
||||
|
||||
## checklist
|
||||
@@ -107,5 +108,20 @@ feeds: gm
|
||||
- [ ] 核心运转只含会反复作用的关系与状况？
|
||||
- [ ] 主要舞台详细度够用即止，未触及范围已留白？
|
||||
- [ ] 未重写体验、硬法则、人物名片、完整历史或百科？
|
||||
- [ ] 尺度由体验正推？多张名片共享的格局已写在本步？
|
||||
- [ ] 产物正文不含评分、诊断、追问和开放问题？
|
||||
```
|
||||
|
||||
## examples
|
||||
|
||||
```examples
|
||||
好：
|
||||
- 体验普通大学生活：熟悉基底为现代都市，尺度是一所学校；主要舞台写到系馆、宿舍、社团，各写玩家在那里做什么、受到什么反馈；国家政治列入留白边界。
|
||||
- 核心运转写世界状况：“每年寒灾”“南极通往怪物世界的出口”“村中活人祭祀”各写如何作用，如寒灾按年逼近、决定储粮与迁徙的代价；祭祀每年要交出活人，决定村里谁有话语权。
|
||||
- 一室朝堂：主要舞台只有一座殿，核心运转是皇权、派系与言官的相互牵制，如何作用写清每次奏对会被谁记账、牵动哪一方。
|
||||
|
||||
坏：
|
||||
- 只写山脉河流，不写用户已点名的祭祀、灾变、异界出口：世界怎样回应玩家仍无从判断。
|
||||
- “地点1：XX咖啡馆，店主叫…”：这是具体实例的活；本能力只写这类场所在舞台上怎样作用于玩家。
|
||||
- 把多张名片共享的台上格局丢给生成规则写“共通前提”：格局写在本能力，生成规则只引用。
|
||||
```
|
||||
|
||||
@@ -39,7 +39,7 @@ feeds: gm,narrator
|
||||
- `append`：按顺序追加记录，由程序执行上限或压缩规则。
|
||||
- `data`：对程序权威数据应用结构化投影，不接收模型伪造的完整状态。
|
||||
5. 隐藏维护通道不创建 DOM。模型只提交允许它生成的槽位；状态和拓扑更新由程序投影。
|
||||
6. 写完整 HTML/CSS/JS，创作时一次生成并固定。美化用配色、字体层级、边框、间距，以及 CSS 或简单内联 SVG，按本次能写稳的程度完成。JS 使用宿主提供的 `zeroLayer.onInit(renderCard)` 与 `zeroLayer.onPatch(applyPatch)`；不得依赖正则替换。地图 SVG 只从结构化数据绘制。
|
||||
6. 写完整 HTML/CSS/JS，创作时一次生成并固定。每个可见槽位一个空的 `data-block` 节点，程序按契约把内容灌进去。markdown/text 槽禁止用 `textContent` 写入；本地必须写时用 `present.markdown`（若宿主提供）或等价安全渲染。根容器（如 `#zero-layer-app`）`width: 100%`，铺满宿主可用宽度，禁止整壳 `max-width` + `margin: 0 auto` 收成居中卡片；若要控阅读行长，只给正文柱设 `max-width`。不要 `100vh`，不要整页或侧栏滚动条。美化用配色、字体层级、边框、间距，以及 CSS 或简单内联 SVG，按本次能写稳的程度完成。JS 使用宿主提供的 `zeroLayer.onInit(renderCard)` 与 `zeroLayer.onPatch(applyPatch)`；不得依赖正则替换。地图 SVG 只从结构化数据绘制。
|
||||
7. 给一份非正史初始卡和一次更新预览，覆盖所有可见槽位与空态。
|
||||
|
||||
HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变主工作面的缺口交给 `self_score.gap`。产物正文不放评分、诊断、追问或开放问题。
|
||||
@@ -50,13 +50,14 @@ HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变
|
||||
```principles
|
||||
1. 固定外壳只创建一次；每轮只更新目标槽位，不重发整页代码。
|
||||
2. 正文或核心工作面保持最高阅读优先级，辅助 Tab 不平均争夺注意力。
|
||||
3. 更新模式来自 0层回复呈现，前端不能擅自把 replace 改成 append。
|
||||
4. 程序投影槽只消费权威数据；正文模型不输出状态副本、地图 SVG 或完整卡快照。
|
||||
5. 每次更新带目标槽位和模式；程序校验允许列表与 revision 后原子提交。
|
||||
6. 动态文字安全渲染；Markdown 使用宿主安全能力，不能拼接未清洗 HTML。
|
||||
7. 允许本地 Tab、折叠、缩放和 SVG 图谱；禁止联网、表单、弹窗、顶层导航和宿主行动回传。
|
||||
8. 正规路径使用结构化 slot_updates；正则只属于外部兼容适配器，不写进本产物。
|
||||
9. 美化落在配色、字体层级、边框、间距和简单内联 SVG 上，写进 CSS，按本次能写稳的程度完成。
|
||||
3. 外壳铺满宿主可用宽度（`width: 100%`）；行长只限正文柱，不限整壳；缺则按宿主全宽排。
|
||||
4. 更新模式来自 0层回复呈现，前端不能擅自把 replace 改成 append。
|
||||
5. 程序投影槽只消费权威数据；正文模型不输出状态副本、地图 SVG 或完整卡快照。
|
||||
6. 每次更新带目标槽位和模式；程序校验允许列表与 revision 后原子提交。
|
||||
7. 动态文字安全渲染；Markdown 使用宿主安全能力，不能拼接未清洗 HTML。
|
||||
8. 允许本地 Tab、折叠、缩放和 SVG 图谱；禁止联网、表单、弹窗、顶层导航和宿主行动回传。
|
||||
9. 正规路径使用结构化 slot_updates；正则只属于外部兼容适配器，不写进本产物。
|
||||
10. 美化落在配色、字体层级、边框、间距和简单内联 SVG 上，写进 CSS，按本次能写稳的程度完成。
|
||||
```
|
||||
|
||||
## probe
|
||||
@@ -130,7 +131,7 @@ HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变
|
||||
],
|
||||
"frontend": {
|
||||
"html": "<div id=\"zero-layer-app\">完整稳定外壳</div>",
|
||||
"css": "完整样式与响应式规则",
|
||||
"css": "完整样式；根容器 width:100%；行长只限正文柱；含响应式规则",
|
||||
"js": "使用 zeroLayer.onInit(renderCard) 与 zeroLayer.onPatch(applyPatch) 的完整代码"
|
||||
},
|
||||
"预览": {
|
||||
@@ -153,7 +154,7 @@ HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变
|
||||
|
||||
硬规则：
|
||||
1. 输出合法 JSON；槽位落点覆盖 0层回复呈现的全部可见槽位，id 与更新模式一致。
|
||||
2. `frontend.html/css/js` 完整且总量不超过 60 KB；JS 必须实现初始化与增量更新入口。
|
||||
2. `frontend.html/css/js` 完整且总量不超过 60 KB；JS 必须实现初始化与增量更新入口。根容器铺满宿主宽度；不要整壳 `max-width` 居中卡片。
|
||||
3. 后续回合只提交 `zero-layer.update.v1`，不重发 HTML 或完整卡。
|
||||
4. `data` 槽只接受程序权威投影；隐藏维护通道不进入 DOM。
|
||||
5. 更新必须带 `base_revision`，由程序校验后原子提交；失败时保持旧卡。
|
||||
@@ -182,9 +183,9 @@ HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变
|
||||
- 1：模型每轮重编权威状态。
|
||||
|
||||
界面结构：
|
||||
- 5：主工作面明确，Tab 和辅助区服务本局体验，窄屏仍可用。
|
||||
- 3：总体可读，但一个辅助区抢占注意力。
|
||||
- 1：机械地堆满 Tab，没有主次。
|
||||
- 5：主工作面明确；外壳铺满宿主宽度；Tab 与辅助区服务本局体验，窄屏仍可用。
|
||||
- 3：总体可读，但一个辅助区抢占注意力，或整壳被限宽成居中卡片。
|
||||
- 1：机械地堆满 Tab，没有主次，或外壳明显窄于宿主。
|
||||
|
||||
实现与安全：
|
||||
- 5：代码完整、安全渲染、无联网与回传，结构化区域有降级。
|
||||
@@ -199,7 +200,7 @@ HTML、CSS、JS 合计不超过 60 KB。材料足够便直接成稿；会改变
|
||||
- [ ] 外壳只初始化一次，后续只应用 slot_updates？
|
||||
- [ ] revision、原子提交与失败保持旧值完整？
|
||||
- [ ] 权威状态没有被模型重复生成，隐藏维护区未进 DOM？
|
||||
- [ ] 前端完整、安全、无正则依赖与外部联网？
|
||||
- [ ] 前端完整、安全、根容器铺满宿主宽度、无正则依赖与外部联网？
|
||||
- [ ] 美化已用配色、字体层级、边框、间距或简单 SVG 写进 CSS，程度限于本次能写稳的部分？
|
||||
```
|
||||
|
||||
|
||||
@@ -23,7 +23,7 @@ id: alongside
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：正文和结构同屏。窄屏改为上下。选项在最下，直接可见。结构用创作时固定的绘图器，游玩灌入拓扑。
|
||||
不替换：外壳铺满宿主宽度；正文和结构同屏。窄屏改为上下。选项在最下，直接可见。结构用创作时固定的绘图器，游玩灌入拓扑。
|
||||
可替换：两栏谁更宽、抬头文字、结构块的短名、配色和字体。本局没有结构时改用阅读或正文为主。
|
||||
```
|
||||
|
||||
@@ -32,19 +32,19 @@ id: alongside
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<p id="heading"></p>
|
||||
<p id="heading" data-block="heading"></p>
|
||||
<div class="pair">
|
||||
<article id="body"></article>
|
||||
<article id="body" data-block="body"></article>
|
||||
<section id="structure">
|
||||
<div id="graph"></div>
|
||||
<div id="graph" data-block="structure"></div>
|
||||
</section>
|
||||
</div>
|
||||
<ol id="options"></ol>
|
||||
<ol id="options" data-block="options"></ol>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c1915; background: #f7f3ea; }
|
||||
#app { max-width: 56rem; margin: 0 auto; padding: 1.25rem; font: 16px/1.6 "Palatino Linotype", "Songti SC", serif; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1.25rem; font: 16px/1.6 "Palatino Linotype", "Songti SC", serif; }
|
||||
.pair { display: grid; grid-template-columns: 1.2fr 0.8fr; gap: 1rem; }
|
||||
#structure { border: 1px solid #d5cbb8; border-radius: 6px; padding: 0.8rem; background: #fffdf8; }
|
||||
#graph { min-height: 8rem; }
|
||||
@@ -54,17 +54,8 @@ css:
|
||||
js:
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putText("heading", blocks.heading);
|
||||
putText("body", blocks.body);
|
||||
putGraph("graph", blocks.structure);
|
||||
putList("options", blocks.options);
|
||||
});
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
}
|
||||
function putGraph(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
@@ -110,15 +101,4 @@ function putGraph(id, value) {
|
||||
});
|
||||
el.appendChild(svg);
|
||||
}
|
||||
function putList(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (!Array.isArray(value) || !value.length) { el.remove(); return; }
|
||||
value.forEach(function (line) {
|
||||
var li = document.createElement("li");
|
||||
li.textContent = String(line);
|
||||
el.appendChild(li);
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
@@ -25,7 +25,7 @@ id: data-cards
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：数据区在抬头下一直展开。正文在其下较宽的一栏，篇幅仍是数百字。窄栏在较窄的一栏。选项在两栏之下，直接可见。窄屏时宽栏在上。
|
||||
不替换：外壳铺满宿主宽度；数据区在抬头下一直展开。正文在其下较宽的一栏，篇幅仍是数百字。窄栏在较窄的一栏。选项在两栏之下，直接可见。窄屏时宽栏在上。
|
||||
可替换：数据项的名称、条数和当前值，窄栏的标题和条目，配色和字体。本局没有窄栏时删掉窄栏，正文占回整宽。
|
||||
```
|
||||
|
||||
@@ -34,21 +34,21 @@ id: data-cards
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<p id="heading"></p>
|
||||
<p id="heading" data-block="heading"></p>
|
||||
<div id="cards"></div>
|
||||
<div class="pair">
|
||||
<article id="body"></article>
|
||||
<article id="body" data-block="body"></article>
|
||||
<aside class="side">
|
||||
<p id="side-title"></p>
|
||||
<p id="side-title" data-block="side_title"></p>
|
||||
<div id="side"></div>
|
||||
</aside>
|
||||
</div>
|
||||
<ol id="options"></ol>
|
||||
<ol id="options" data-block="options"></ol>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c2420; background: #d5ddd4; }
|
||||
#app { max-width: 52rem; margin: 0 auto; padding: 1rem; font: 15px/1.75 "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1rem; font: 15px/1.75 "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif; }
|
||||
#heading { margin: 0 0 0.6rem; font-size: 13px; }
|
||||
#cards { display: grid; grid-template-columns: repeat(4, 1fr); gap: 0.45rem; margin-bottom: 0.75rem; }
|
||||
#cards.stack { grid-template-columns: 1fr; }
|
||||
@@ -66,19 +66,9 @@ css:
|
||||
js:
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putText("heading", blocks.heading);
|
||||
putText("side-title", blocks.side_title);
|
||||
putText("body", blocks.body);
|
||||
putCards("cards", blocks.cards);
|
||||
putLines("side", blocks.side);
|
||||
putList("options", blocks.options);
|
||||
});
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
}
|
||||
function putCards(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
@@ -111,15 +101,4 @@ function putLines(id, value) {
|
||||
el.appendChild(p);
|
||||
});
|
||||
}
|
||||
function putList(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (!Array.isArray(value) || !value.length) { el.remove(); return; }
|
||||
value.forEach(function (line) {
|
||||
var li = document.createElement("li");
|
||||
li.textContent = String(line);
|
||||
el.appendChild(li);
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
@@ -14,7 +14,7 @@ id: dual-text
|
||||
可加: 短文、结构、数据、抬头、文字选项
|
||||
摆放:
|
||||
抬头: 最上的一行短内容
|
||||
正文: 占满阅读宽度，按数百字来写
|
||||
正文: 占满宿主阅读面；行长可限在正文柱，按数百字来写
|
||||
短文、结构: 标签或折叠，点开才看。标签用短显示名
|
||||
数据: 若有，默认可收起
|
||||
文字选项: 始终直接可见，不放进标签或折叠
|
||||
@@ -24,8 +24,8 @@ id: dual-text
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：正文占满阅读宽度。其余块收在标签或折叠里，默认不展开。选项在这些控件之外，直接可见。
|
||||
可替换：标签和折叠的短名、收起块里的文字、配色和字体。本局没有的块删掉。数据要一直展开时改用数据常驻。
|
||||
不替换：外壳铺满宿主宽度；正文是主阅读面。其余块收在标签或折叠里，默认不展开。选项在这些控件之外，直接可见。
|
||||
可替换：标签和折叠的短名、收起块里的文字、配色和字体、正文柱行长上限。本局没有的块删掉。数据要一直展开时改用数据常驻。
|
||||
```
|
||||
|
||||
## sample
|
||||
@@ -33,47 +33,26 @@ id: dual-text
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<p id="heading"></p>
|
||||
<article id="body"></article>
|
||||
<p id="heading" data-block="heading"></p>
|
||||
<article id="body" data-block="body"></article>
|
||||
<details>
|
||||
<summary id="vignette-title"></summary>
|
||||
<div id="vignette"></div>
|
||||
<summary id="vignette-title" data-block="vignette_title"></summary>
|
||||
<div id="vignette" data-block="vignette"></div>
|
||||
</details>
|
||||
<ol id="options"></ol>
|
||||
<ol id="options" data-block="options"></ol>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c1915; background: #f7f3ea; }
|
||||
#app { max-width: 40rem; margin: 0 auto; padding: 1.25rem; font: 17px/1.75 "Palatino Linotype", "Songti SC", serif; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1.25rem; font: 17px/1.75 "Palatino Linotype", "Songti SC", serif; }
|
||||
#heading, #body, details, #options { max-width: 40rem; margin-inline: auto; }
|
||||
#heading { margin: 0 0 0.8rem; font-size: 0.95rem; }
|
||||
#vignette { font-size: 0.92rem; margin-top: 0.6rem; }
|
||||
#options { margin: 0.8rem 0 0; padding-left: 1.2rem; }
|
||||
#options { margin: 0.8rem auto 0; padding-left: 1.2rem; }
|
||||
summary { cursor: pointer; }
|
||||
|
||||
js:
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putText("heading", blocks.heading);
|
||||
putText("body", blocks.body);
|
||||
putText("vignette-title", blocks.vignette_title);
|
||||
putText("vignette", blocks.vignette);
|
||||
putList("options", blocks.options);
|
||||
present.onData(function () {
|
||||
/* 文本区由程序按 data-block 灌数；折叠开合是浏览器原生行为 */
|
||||
});
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
}
|
||||
function putList(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (!Array.isArray(value) || !value.length) { el.remove(); return; }
|
||||
value.forEach(function (line) {
|
||||
var li = document.createElement("li");
|
||||
li.textContent = String(line);
|
||||
el.appendChild(li);
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
@@ -23,8 +23,8 @@ id: paged
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：同时只显示一页。页用一行标签切换。抬头在标签之外。
|
||||
可替换：每个标签的短名、页内文字和数据项、配色和字体。本局未启用的页不建标签。
|
||||
不替换：外壳铺满宿主宽度；同时只显示一页。页用一行标签切换。抬头在标签之外。
|
||||
可替换：每个标签的短名、页内文字和数据项、配色和字体、页内正文柱行长上限。本局未启用的页不建标签。
|
||||
```
|
||||
|
||||
## sample
|
||||
@@ -32,14 +32,15 @@ id: paged
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<p id="heading"></p>
|
||||
<p id="heading" data-block="heading"></p>
|
||||
<div id="tabs"></div>
|
||||
<div id="page"></div>
|
||||
<div id="page" data-block="page"></div>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c1915; background: #f7f3ea; }
|
||||
#app { max-width: 44rem; margin: 0 auto; padding: 1.25rem; font: 16px/1.65 "Segoe UI", "Source Han Sans SC", sans-serif; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1.25rem; font: 16px/1.65 "Segoe UI", "Source Han Sans SC", sans-serif; }
|
||||
#heading, #tabs, #page { max-width: 44rem; margin-inline: auto; }
|
||||
#tabs { display: flex; gap: 0.4rem; margin-bottom: 0.9rem; }
|
||||
#tabs button { border: 0; border-bottom: 2px solid transparent; background: transparent; padding: 0.35rem 0.6rem; font: inherit; color: inherit; }
|
||||
#tabs button[aria-current="true"] { border-bottom-color: #8a5a2b; }
|
||||
@@ -49,7 +50,6 @@ var pages = [];
|
||||
var current = 0;
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putText("heading", blocks.heading);
|
||||
pages = Array.isArray(blocks.pages) ? blocks.pages : [];
|
||||
current = 0;
|
||||
drawTabs();
|
||||
@@ -73,12 +73,7 @@ function drawTabs() {
|
||||
}
|
||||
function drawPage() {
|
||||
var el = document.getElementById("page");
|
||||
el.textContent = pages[current] ? pages[current].text : "";
|
||||
}
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
var text = pages[current] ? pages[current].text : "";
|
||||
el.innerHTML = text ? present.markdown(String(text)) : "";
|
||||
}
|
||||
```
|
||||
|
||||
@@ -13,7 +13,7 @@ id: piece
|
||||
主区: 一份成稿
|
||||
可加窄带: 文字选项
|
||||
摆放:
|
||||
成稿: 居中。内部小节在创作时按范例重写
|
||||
成稿: 外壳铺满宿主；成稿正文柱可限行长并水平居中
|
||||
文字选项: 成稿之后，始终直接可见
|
||||
未启用的区不留空位。
|
||||
```
|
||||
@@ -21,8 +21,8 @@ id: piece
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：一份成稿容器居中。选项在容器之后，直接可见。
|
||||
可替换：容器内部的标记、分段、配色和字体。按本局这一份成品重写内部。样本内部与本局不同时整段替换，保留容器和选项的前后关系。
|
||||
不替换：外壳铺满宿主宽度；一份成稿容器。选项在容器之后，直接可见。
|
||||
可替换：容器内部的标记、分段、配色和字体、成稿柱行长上限。按本局这一份成品重写内部。样本内部与本局不同时整段替换，保留容器和选项的前后关系。
|
||||
```
|
||||
|
||||
## sample
|
||||
@@ -30,37 +30,19 @@ id: piece
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<article id="piece"></article>
|
||||
<ol id="options"></ol>
|
||||
<article id="piece" data-block="piece"></article>
|
||||
<ol id="options" data-block="options"></ol>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c1916; background: #ebe6dc; }
|
||||
#app { max-width: 34rem; margin: 0 auto; padding: 1.6rem 1.25rem 2rem; font: 16px/1.65 "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif; }
|
||||
#piece { margin: 0; }
|
||||
#options { margin: 1rem 0 0; padding-left: 1.15rem; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1.6rem 1.25rem 2rem; font: 16px/1.65 "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif; }
|
||||
#piece, #options { max-width: 34rem; margin-inline: auto; }
|
||||
#piece { margin: 0 auto; }
|
||||
#options { margin: 1rem auto 0; padding-left: 1.15rem; }
|
||||
|
||||
js:
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putText("piece", blocks.piece);
|
||||
putList("options", blocks.options);
|
||||
present.onData(function () {
|
||||
/* 成稿与选项由程序按 data-block 灌数 */
|
||||
});
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
}
|
||||
function putList(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (!Array.isArray(value) || !value.length) { el.remove(); return; }
|
||||
value.forEach(function (line) {
|
||||
var li = document.createElement("li");
|
||||
li.textContent = String(line);
|
||||
el.appendChild(li);
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
@@ -14,7 +14,7 @@ id: reading
|
||||
可加窄带: 抬头、文字选项
|
||||
摆放:
|
||||
抬头: 正文之前的一行短内容
|
||||
正文: 阅读柱，占主要宽度
|
||||
正文: 阅读柱；外壳铺满宿主，行长可限在正文柱
|
||||
文字选项: 正文之后，始终直接可见，不放进标签或折叠。不显示「文字选项」这个区名
|
||||
未启用的区不留空位。
|
||||
```
|
||||
@@ -22,8 +22,8 @@ id: reading
|
||||
## note
|
||||
|
||||
```note
|
||||
不替换：单栏阅读柱。抬头在正文前，独占一行。选项在正文后，直接可见。
|
||||
可替换：抬头文字、正文文字、配色和字体。本局没有抬头时删掉抬头。
|
||||
不替换：外壳铺满宿主宽度；单栏阅读柱。抬头在正文前，独占一行。选项在正文后，直接可见。
|
||||
可替换：抬头文字、正文文字、配色和字体、正文柱行长上限。本局没有抬头时删掉抬头。
|
||||
```
|
||||
|
||||
## sample
|
||||
@@ -31,53 +31,21 @@ id: reading
|
||||
```sample
|
||||
html:
|
||||
<div id="app">
|
||||
<p id="heading"></p>
|
||||
<article id="body"></article>
|
||||
<ol id="options"></ol>
|
||||
<p id="heading" data-block="heading"></p>
|
||||
<article id="body" data-block="body"></article>
|
||||
<ol id="options" data-block="options"></ol>
|
||||
</div>
|
||||
|
||||
css:
|
||||
:root { color: #1c1915; background: #f7f3ea; }
|
||||
#app { max-width: 38rem; margin: 0 auto; padding: 1.5rem 1.25rem 2.5rem; font: 17px/1.75 "Palatino Linotype", "Songti SC", serif; }
|
||||
#app { width: 100%; box-sizing: border-box; margin: 0; padding: 1.5rem 1.25rem 2.5rem; font: 17px/1.75 "Palatino Linotype", "Songti SC", serif; }
|
||||
#heading, #body, #options { max-width: 38rem; margin-inline: auto; }
|
||||
#heading { margin: 0 0 0.85rem; font-size: 0.95rem; letter-spacing: 0.04em; }
|
||||
#body { font-size: 1.05rem; }
|
||||
#options { margin: 1.25rem 0 0; padding-left: 1.2rem; }
|
||||
#options { margin: 1.25rem auto 0; padding-left: 1.2rem; }
|
||||
|
||||
js:
|
||||
present.onData(function (data) {
|
||||
var blocks = (data && data.blocks) || {};
|
||||
putHeading("heading", blocks.heading);
|
||||
putText("body", blocks.body);
|
||||
putList("options", blocks.options);
|
||||
present.onData(function () {
|
||||
/* 正文/抬头/选项由程序按 data-block 灌数；本地交互再写这里 */
|
||||
});
|
||||
function putHeading(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
if (typeof value === "string") { el.textContent = value; return; }
|
||||
Object.keys(value).forEach(function (key) {
|
||||
var item = document.createElement("span");
|
||||
item.textContent = key + " " + String(value[key]);
|
||||
item.style.marginRight = "1rem";
|
||||
el.appendChild(item);
|
||||
});
|
||||
}
|
||||
function putText(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
if (value == null || value === "") { el.remove(); return; }
|
||||
el.textContent = String(value);
|
||||
}
|
||||
function putList(id, value) {
|
||||
var el = document.getElementById(id);
|
||||
if (!el) return;
|
||||
el.replaceChildren();
|
||||
if (!Array.isArray(value) || !value.length) { el.remove(); return; }
|
||||
value.forEach(function (line) {
|
||||
var li = document.createElement("li");
|
||||
li.textContent = String(line);
|
||||
el.appendChild(li);
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
+26
-3
@@ -9,7 +9,20 @@ import {
|
||||
writeFileSync,
|
||||
} from "node:fs";
|
||||
import path from "node:path";
|
||||
import type { BookProject, BookSummary } from "../types/book.js";
|
||||
import type {
|
||||
BookCreationMode,
|
||||
BookProject,
|
||||
BookSummary,
|
||||
} from "../types/book.js";
|
||||
|
||||
function normalizeBookCreationMode(
|
||||
raw: string | undefined | null,
|
||||
): BookCreationMode {
|
||||
if (raw === "listen_serial" || raw === "recipe" || raw === "dictate") {
|
||||
return raw;
|
||||
}
|
||||
return "dictate";
|
||||
}
|
||||
import type { PersistedBookSession } from "../types/book-session.js";
|
||||
import type { RunSnapshot } from "../types/run-snapshot.js";
|
||||
import { ensureUserDataDirs, getUserDataDir } from "../config/user-data-dir.js";
|
||||
@@ -57,6 +70,7 @@ export function listBooks(): BookSummary[] {
|
||||
title: b.title,
|
||||
activeSkillId: b.activeSkillId ?? b.orchestratorId,
|
||||
activeSkillName: b.activeSkillName ?? b.orchestratorName,
|
||||
creationMode: normalizeBookCreationMode(b.creationMode),
|
||||
preview: b.preview,
|
||||
updatedAt: b.updatedAt,
|
||||
orchestratorId: b.orchestratorId,
|
||||
@@ -67,19 +81,27 @@ export function listBooks(): BookSummary[] {
|
||||
export function getBook(id: string): BookProject | null {
|
||||
try {
|
||||
const raw = readFileSync(bookPath(id), "utf8");
|
||||
return JSON.parse(raw) as BookProject;
|
||||
const book = JSON.parse(raw) as BookProject;
|
||||
return {
|
||||
...book,
|
||||
creationMode: normalizeBookCreationMode(book.creationMode),
|
||||
};
|
||||
} catch {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
export function createBook(input: { title?: string }): BookProject {
|
||||
export function createBook(input: {
|
||||
title?: string;
|
||||
creationMode?: BookCreationMode;
|
||||
}): BookProject {
|
||||
ensureUserDataDirs();
|
||||
mkdirSync(booksDir(), { recursive: true });
|
||||
const now = new Date().toISOString();
|
||||
const book: BookProject = {
|
||||
id: randomUUID(),
|
||||
title: input.title?.trim() || "未命名作品",
|
||||
creationMode: normalizeBookCreationMode(input.creationMode),
|
||||
preview: "新建作品，描述你想创作什么…",
|
||||
sessionIds: [],
|
||||
createdAt: now,
|
||||
@@ -100,6 +122,7 @@ export function updateBook(
|
||||
| "activeSessionId"
|
||||
| "activeSkillId"
|
||||
| "activeSkillName"
|
||||
| "creationMode"
|
||||
| "orchestratorId"
|
||||
| "orchestratorName"
|
||||
>
|
||||
|
||||
+161
-29
@@ -6,42 +6,140 @@ import {
|
||||
type DictateProduct,
|
||||
} from "./types.js";
|
||||
|
||||
export type DictateContextPhase = "select" | "fill" | "reply";
|
||||
|
||||
/** 五段字数，只给用量统计，不另存正文。 */
|
||||
export type DictateSegmentChars = {
|
||||
rules: number;
|
||||
history: number;
|
||||
products: number;
|
||||
injected: number;
|
||||
tail: number;
|
||||
};
|
||||
|
||||
export type DictateBuiltContext = {
|
||||
messages: ChatMessage[];
|
||||
segmentChars: DictateSegmentChars;
|
||||
};
|
||||
|
||||
/**
|
||||
* 拼给模型的上下文：系统提示 + 全部产物（按相对序）+ 全量对话。
|
||||
* 清空对话后，只剩产物块，产物即浓缩上下文。
|
||||
* 创作请求的块顺序：固定规则、更早的对话、全部产物、本轮注入、末尾。
|
||||
* 选题不带产物正文和本轮注入。回复步只带本轮刚落盘产物与末句用户话。
|
||||
* 更早对话只追加，最后一轮用户输入放在末尾。
|
||||
*/
|
||||
export function buildDictateContext(params: {
|
||||
systemPrompt: string;
|
||||
products: DictateProduct[];
|
||||
dialogue: DictateChatTurn[];
|
||||
phase?: DictateContextPhase;
|
||||
injected?: string;
|
||||
tail?: string;
|
||||
}): DictateBuiltContext {
|
||||
const phase = params.phase ?? "fill";
|
||||
if (phase === "reply") {
|
||||
return buildReplyContext(params);
|
||||
}
|
||||
const { earlier, lastUser } = splitEarlierDialogue(params.dialogue);
|
||||
const history = formatEarlierDialogue(earlier);
|
||||
const products =
|
||||
phase === "fill" ? formatProducts(sortDictateProducts(params.products)) : "";
|
||||
const injected = phase === "fill" ? params.injected?.trim() ?? "" : "";
|
||||
const tail = joinBlocks([params.tail?.trim() ?? "", lastUser ? `最后一轮用户输入\n${lastUser}` : ""]);
|
||||
|
||||
const fillDuty =
|
||||
phase === "fill"
|
||||
? "## 本步收尾\n\n写完本轮节点表后，同一次请求必须调用 reply_user。只 insert 不算完成。"
|
||||
: "";
|
||||
const sections = [
|
||||
fillDuty,
|
||||
history,
|
||||
products ? `## 当前产物（已写入黑板，按相对序；对话清空后仍保留）\n\n${products}` : "",
|
||||
injected
|
||||
? `## 本轮注入\n\n写完下列节点后，同一次请求必须调用 reply_user。\n\n本轮节点表\n\n${injected}`
|
||||
: "",
|
||||
tail ? `## 本轮\n\n${tail}` : "",
|
||||
].filter(Boolean);
|
||||
|
||||
const user = sections.join("\n\n");
|
||||
return {
|
||||
messages: [
|
||||
{ role: "system", content: params.systemPrompt },
|
||||
{ role: "user", content: user },
|
||||
],
|
||||
segmentChars: {
|
||||
rules: params.systemPrompt.length,
|
||||
history: history.length,
|
||||
products: products.length,
|
||||
injected: injected.length,
|
||||
tail: tail.length,
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
/** 回复兜底：只给本轮产物 + 末句，不带全黑板与节点规格。 */
|
||||
function buildReplyContext(params: {
|
||||
systemPrompt: string;
|
||||
products: DictateProduct[];
|
||||
dialogue: DictateChatTurn[];
|
||||
tail?: string;
|
||||
}): DictateBuiltContext {
|
||||
const { lastUser } = splitEarlierDialogue(params.dialogue);
|
||||
const products = formatRoundProducts(params.products);
|
||||
const tail = joinBlocks([
|
||||
params.tail?.trim() ?? "",
|
||||
lastUser ? `最后一轮用户输入\n${lastUser}` : "",
|
||||
]);
|
||||
const sections = [
|
||||
products ? `## 本轮刚落盘\n\n${products}` : "## 本轮刚落盘\n\n（无正文产物；按删除与用户原话回复）",
|
||||
tail ? `## 本轮\n\n${tail}` : "",
|
||||
].filter(Boolean);
|
||||
const user = sections.join("\n\n");
|
||||
return {
|
||||
messages: [
|
||||
{ role: "system", content: params.systemPrompt },
|
||||
{ role: "user", content: user },
|
||||
],
|
||||
segmentChars: {
|
||||
rules: params.systemPrompt.length,
|
||||
history: 0,
|
||||
products: products.length,
|
||||
injected: 0,
|
||||
tail: tail.length,
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
function formatRoundProducts(products: DictateProduct[]): string {
|
||||
if (products.length === 0) return "";
|
||||
return sortDictateProducts(products)
|
||||
.map((product) => {
|
||||
const score = formatSelfScore(product);
|
||||
return [`### ${product.tag}`, score, product.content].filter(Boolean).join("\n");
|
||||
})
|
||||
.join("\n\n");
|
||||
}
|
||||
|
||||
function formatSelfScore(product: DictateProduct): string {
|
||||
const dims = product.selfScore?.dims;
|
||||
if (!dims?.length) return "";
|
||||
const lines = dims.map((dim) => {
|
||||
const gap = dim.gap?.trim() ? `；缺口：${dim.gap.trim()}` : "";
|
||||
const options =
|
||||
dim.options?.length ? `；选项：${dim.options.map((item) => item.trim()).filter(Boolean).join(" / ")}` : "";
|
||||
return `- ${dim.name}：${dim.score}${gap}${options}`;
|
||||
});
|
||||
return `self_score：\n${lines.join("\n")}`;
|
||||
}
|
||||
|
||||
export function buildDictateMessages(params: {
|
||||
systemPrompt: string;
|
||||
products: DictateProduct[];
|
||||
dialogue: DictateChatTurn[];
|
||||
phase?: DictateContextPhase;
|
||||
injected?: string;
|
||||
tail?: string;
|
||||
}): ChatMessage[] {
|
||||
const parts: string[] = [];
|
||||
const products = sortDictateProducts(params.products);
|
||||
|
||||
if (products.length > 0) {
|
||||
parts.push("## 当前产物（已写入黑板，按相对序；对话清空后仍保留）");
|
||||
for (const p of products) {
|
||||
const ord = effectiveDictateOrder(p);
|
||||
parts.push(`### ${p.tag} 〔order=${ord}〕\n${p.content}`);
|
||||
}
|
||||
} else {
|
||||
parts.push("## 当前产物\n（尚无）");
|
||||
}
|
||||
|
||||
parts.push("## 对话（全量）");
|
||||
if (params.dialogue.length === 0) {
|
||||
parts.push("（尚无对话；等用户开口）");
|
||||
} else {
|
||||
for (const turn of params.dialogue) {
|
||||
const who = turn.role === "user" ? "用户" : "助手";
|
||||
parts.push(`【${who}】\n${turn.text}`);
|
||||
}
|
||||
}
|
||||
|
||||
return [
|
||||
{ role: "system", content: params.systemPrompt },
|
||||
{ role: "user", content: parts.join("\n\n") },
|
||||
];
|
||||
return buildDictateContext(params).messages;
|
||||
}
|
||||
|
||||
/** 从会话消息里抽出转述式对话（用户输入 + 助手回复） */
|
||||
@@ -62,3 +160,37 @@ export function extractDictateDialogue(
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
function splitEarlierDialogue(dialogue: DictateChatTurn[]): {
|
||||
earlier: DictateChatTurn[];
|
||||
lastUser: string;
|
||||
} {
|
||||
const last = dialogue.at(-1);
|
||||
if (last?.role === "user") {
|
||||
return { earlier: dialogue.slice(0, -1), lastUser: last.text.trim() };
|
||||
}
|
||||
return { earlier: dialogue, lastUser: "" };
|
||||
}
|
||||
|
||||
function formatEarlierDialogue(dialogue: DictateChatTurn[]): string {
|
||||
if (dialogue.length === 0) return "## 更早的对话\n\n（尚无更早对话）";
|
||||
const lines = dialogue.map((turn) => {
|
||||
const who = turn.role === "user" ? "用户" : "助手";
|
||||
return `【${who}】\n${turn.text}`;
|
||||
});
|
||||
return `## 更早的对话\n\n${lines.join("\n\n")}`;
|
||||
}
|
||||
|
||||
function formatProducts(products: DictateProduct[]): string {
|
||||
if (products.length === 0) return "（尚无）";
|
||||
return products
|
||||
.map((product) => {
|
||||
const ord = effectiveDictateOrder(product);
|
||||
return `### ${product.tag} 〔order=${ord}〕\n${product.content}`;
|
||||
})
|
||||
.join("\n\n");
|
||||
}
|
||||
|
||||
function joinBlocks(blocks: string[]): string {
|
||||
return blocks.map((block) => block.trim()).filter(Boolean).join("\n\n");
|
||||
}
|
||||
|
||||
@@ -1,5 +1,6 @@
|
||||
import {
|
||||
catalogModulesForIntake,
|
||||
isDictateProgramModule,
|
||||
type ModuleCatalog,
|
||||
type ModuleCatalogEntry,
|
||||
} from "../skills/creation-flow.js";
|
||||
@@ -33,6 +34,53 @@ const ALL_PRESENTATION_MODULE_IDS = new Set(
|
||||
Object.values(PRESENTATION_MODULE_IDS).flat(),
|
||||
);
|
||||
|
||||
/** 技能页按节点在对话落盘里的角色分组，不按旧配方包。 */
|
||||
export const MODULE_BROWSE_GROUPS = [
|
||||
{
|
||||
id: "plan",
|
||||
label: "本局方案",
|
||||
hint: "抽出体验并给内容节点分档，不进入游玩",
|
||||
},
|
||||
{
|
||||
id: "content",
|
||||
label: "内容节点",
|
||||
hint: "本局方案按档取用；这句话说清了再写",
|
||||
},
|
||||
{
|
||||
id: "present",
|
||||
label: "呈现链",
|
||||
hint: "普通楼层与 0 层卡各用一套，两套不混用",
|
||||
},
|
||||
{
|
||||
id: "program",
|
||||
label: "落档程序",
|
||||
hint: "开玩前由程序编译，创作对话不写入",
|
||||
},
|
||||
] as const;
|
||||
|
||||
export type ModuleBrowseGroupId = (typeof MODULE_BROWSE_GROUPS)[number]["id"];
|
||||
|
||||
export function moduleBrowseGroup(module: {
|
||||
id: string;
|
||||
auto?: boolean;
|
||||
}): ModuleBrowseGroupId {
|
||||
if (module.id === "creation-plan") return "plan";
|
||||
if (isDictateProgramModule(module)) return "program";
|
||||
if (ALL_PRESENTATION_MODULE_IDS.has(module.id)) return "present";
|
||||
return "content";
|
||||
}
|
||||
|
||||
export function modulePresentationMode(
|
||||
moduleId: string,
|
||||
): CreationPresentationMode | null {
|
||||
for (const mode of ["message", "zero_layer"] as const) {
|
||||
if ((PRESENTATION_MODULE_IDS[mode] as readonly string[]).includes(moduleId)) {
|
||||
return mode;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
export type CreationPlanItem = {
|
||||
module_id: string;
|
||||
tier: CreationPlanTier;
|
||||
|
||||
+12
-5
@@ -1,4 +1,15 @@
|
||||
export { buildDictateSystemPrompt, DICTATE_SYSTEM_PROMPT } from "./prompt.js";
|
||||
export {
|
||||
buildDictateReplyPrompt,
|
||||
buildDictateSelectPrompt,
|
||||
buildDictateSystemPrompt,
|
||||
DICTATE_SYSTEM_PROMPT,
|
||||
formatDictateRoundTail,
|
||||
} from "./prompt.js";
|
||||
export {
|
||||
buildDictateMessages,
|
||||
buildDictateContext,
|
||||
extractDictateDialogue,
|
||||
} from "./context.js";
|
||||
export {
|
||||
CREATION_INTAKE_MODE_TAG,
|
||||
DICTATE_MODE_VALUE,
|
||||
@@ -16,10 +27,6 @@ export {
|
||||
type DictateSelfScore,
|
||||
type DictateChatTurn,
|
||||
} from "./types.js";
|
||||
export {
|
||||
buildDictateMessages,
|
||||
extractDictateDialogue,
|
||||
} from "./context.js";
|
||||
export { DICTATE_TOOL_DEFINITIONS } from "./tools.js";
|
||||
export { runDictateTurn, type DictateTurnResult } from "./turn.js";
|
||||
export {
|
||||
|
||||
@@ -15,6 +15,7 @@ import {
|
||||
/** 对话落盘常用简化 tag → 能力 id（catalog.artifact 之外） */
|
||||
export const DICTATE_INSERT_TAG_ALIASES: Readonly<Record<string, string>> = {
|
||||
"设计.开场白": "opening-setup",
|
||||
"设计.开场白与开场变量": "opening-setup",
|
||||
"设计.美学纲领": "aesthetics-interaction",
|
||||
"设计.文风": "narrative-guide",
|
||||
"设计.监控栏": "status-bar",
|
||||
@@ -54,7 +55,8 @@ export function resolveModuleForInsertTag(
|
||||
if (byFamily) return byFamily;
|
||||
}
|
||||
|
||||
const aliasId = DICTATE_INSERT_TAG_ALIASES[t];
|
||||
const family = t.includes("#") ? t.slice(0, t.indexOf("#")).trim() : t;
|
||||
const aliasId = DICTATE_INSERT_TAG_ALIASES[t] ?? DICTATE_INSERT_TAG_ALIASES[family];
|
||||
if (aliasId) {
|
||||
return catalog.modules.find((m) => m.id === aliasId) ?? null;
|
||||
}
|
||||
|
||||
@@ -1,7 +1,9 @@
|
||||
/**
|
||||
* 对话落盘 · 落档/开玩时自动把产物 tag 挂到游玩 worker。
|
||||
* 只做「固定上下文绑定」；变量清单、可见性、分档映射仍靠对话产出，此处不发明。
|
||||
* 保存时按上游产物指纹决定是否重跑：没改过就保留已有排序和运行配置。
|
||||
*/
|
||||
import { createHash } from "node:crypto";
|
||||
import type { Blackboard } from "../blackboard/blackboard.js";
|
||||
import {
|
||||
CONTEXT_ORDER_SCHEMA,
|
||||
@@ -38,6 +40,12 @@ import {
|
||||
VALUE_MAP_TAG,
|
||||
listValueMapTargetTags,
|
||||
} from "../skills/value-map.js";
|
||||
import {
|
||||
applyMaintenanceStrategyToPlaySlots,
|
||||
MAINTENANCE_STRATEGY_TAG,
|
||||
resolveMaintenanceStrategyFromBoard,
|
||||
writeMaintenanceStrategy,
|
||||
} from "../skills/maintenance-strategy.js";
|
||||
import {
|
||||
DICTATE_ORDER_META_KEY,
|
||||
effectiveDictateOrder,
|
||||
@@ -57,6 +65,7 @@ const SKIP_BIND_TAGS = new Set([
|
||||
CREATION_SELECTED_RECIPE_TAG,
|
||||
VARIABLE_CATALOG_TAG,
|
||||
VALUE_MAP_TAG,
|
||||
MAINTENANCE_STRATEGY_TAG,
|
||||
]);
|
||||
|
||||
function shouldSkipBindTag(tag: string): boolean {
|
||||
@@ -77,11 +86,16 @@ const STYLE_TAG_RE =
|
||||
/** 这些配方开玩时刚需叙事转述（主世界出内容、转述出文风） */
|
||||
const RECIPES_REQUIRE_NARRATOR = new Set(["文本生成器"]);
|
||||
|
||||
/** 上次编译时所依据的上游指纹；两边产物一致且未过期则跳过重跑 */
|
||||
export const DICTATE_PLAY_BIND_INPUT_META = "dictatePlayBindInput";
|
||||
|
||||
export type DictatePlayBindResult = {
|
||||
workerSetJson: string;
|
||||
contextOrder: ContextOrderDoc;
|
||||
boundTags: string[];
|
||||
changed: boolean;
|
||||
/** 上游产物未改，保留已有上下文排序与运行配置 */
|
||||
skipped: boolean;
|
||||
};
|
||||
|
||||
function isStyleProductTag(tag: string): boolean {
|
||||
@@ -314,7 +328,7 @@ function resolvePlaySlots(
|
||||
const parsed = existingRaw?.trim()
|
||||
? parseWorkerSetYaml(existingRaw)
|
||||
: null;
|
||||
const slots = { ...(parsed?.play_slots ?? defaultPlaySlots()) };
|
||||
let slots = { ...(parsed?.play_slots ?? defaultPlaySlots()) };
|
||||
const recipeRef = parseSelectedRecipeRef(
|
||||
board.getContentByTag(CREATION_SELECTED_RECIPE_TAG),
|
||||
);
|
||||
@@ -322,6 +336,9 @@ function resolvePlaySlots(
|
||||
slots.gm = true;
|
||||
slots.narrator = true;
|
||||
}
|
||||
// 变量设计.维护复杂度 → auditor true/false（内联则关旁观）
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(board);
|
||||
slots = applyMaintenanceStrategyToPlaySlots(slots, strategy);
|
||||
return slots;
|
||||
}
|
||||
|
||||
@@ -336,6 +353,8 @@ export function bindDictateProductsToPlaySpec(board: Blackboard): DictatePlayBin
|
||||
board.getContentByTag(WORKER_SET_DRAFT_TAG)?.trim() ||
|
||||
"";
|
||||
const playSlots = resolvePlaySlots(existing || undefined, board);
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(board);
|
||||
writeMaintenanceStrategy(board, strategy, "runtime:play-bind");
|
||||
const runtimeTags = collectRuntimeStateTags(board);
|
||||
const contextOrder = buildDictateContextOrder({
|
||||
products,
|
||||
@@ -371,26 +390,100 @@ export function bindDictateProductsToPlaySpec(board: Blackboard): DictatePlayBin
|
||||
? row.brief
|
||||
: "对话落盘：主世界出内容，叙事转述出文风";
|
||||
}
|
||||
if (playSlots.auditor) {
|
||||
const base =
|
||||
typeof row.brief === "string" && row.brief.trim() ? String(row.brief) : "对话落盘落档自动生成";
|
||||
if (!base.includes("旁观")) {
|
||||
row.brief = `${base}；旁观副 LLM 填表（固定）`;
|
||||
}
|
||||
}
|
||||
const withSlots = JSON.stringify(row, null, 2);
|
||||
const workerSetJson = mergeContextOrderIntoWorkerSetJson(withSlots, contextOrder);
|
||||
|
||||
const boundTags = products.map((p) => p.tag);
|
||||
const changed = workerSetJson.trim() !== existing.trim();
|
||||
return { workerSetJson, contextOrder, boundTags, changed };
|
||||
return { workerSetJson, contextOrder, boundTags, changed, skipped: false };
|
||||
}
|
||||
|
||||
/** 写入黑板：设计.worker集 + 设计.上下文投影排序 */
|
||||
/**
|
||||
* 上游指纹：会挂进游玩上下文的产物正文与次序，加上会改变槽位的配方、变量目录、映射和已启用槽。
|
||||
* 不含投影结果（变量.当前、上下文排序、worker 集正文），避免保存后自己把自己标脏。
|
||||
*/
|
||||
export function dictatePlayBindInputKey(board: Blackboard): string {
|
||||
const products = collectDictateBindProducts(board).map((product) => ({
|
||||
tag: product.tag,
|
||||
order: product.order ?? null,
|
||||
content: product.content,
|
||||
}));
|
||||
const existing =
|
||||
board.getContentByTag(WORKER_SET_FINAL_TAG)?.trim() ||
|
||||
board.getContentByTag(WORKER_SET_DRAFT_TAG)?.trim() ||
|
||||
"";
|
||||
const payload = {
|
||||
products,
|
||||
recipe: board.getContentByTag(CREATION_SELECTED_RECIPE_TAG)?.trim() ?? "",
|
||||
catalog: board.getContentByTag(VARIABLE_CATALOG_TAG)?.trim() ?? "",
|
||||
maps: board.getContentByTag(VALUE_MAP_TAG)?.trim() ?? "",
|
||||
playSlots: resolvePlaySlots(existing || undefined, board),
|
||||
};
|
||||
return createHash("sha256").update(JSON.stringify(payload)).digest("hex");
|
||||
}
|
||||
|
||||
function storedPlayBindKey(
|
||||
board: Blackboard,
|
||||
tag: string,
|
||||
): string | undefined {
|
||||
const stamp = board.getLatestByTag(tag)?.metadata?.[DICTATE_PLAY_BIND_INPUT_META];
|
||||
return typeof stamp === "string" && stamp.trim() ? stamp : undefined;
|
||||
}
|
||||
|
||||
function freshPlayBind(
|
||||
board: Blackboard,
|
||||
key: string,
|
||||
): DictatePlayBindResult | null {
|
||||
const orderItem = board.getLatestByTag(CONTEXT_ORDER_TAG);
|
||||
const workerItem = board.getLatestByTag(WORKER_SET_FINAL_TAG);
|
||||
const orderRaw = orderItem?.content?.trim() ?? "";
|
||||
const workerRaw = workerItem?.content?.trim() ?? "";
|
||||
if (!orderRaw || !workerRaw) return null;
|
||||
if (storedPlayBindKey(board, CONTEXT_ORDER_TAG) !== key) return null;
|
||||
if (storedPlayBindKey(board, WORKER_SET_FINAL_TAG) !== key) return null;
|
||||
try {
|
||||
const contextOrder = JSON.parse(orderRaw) as ContextOrderDoc;
|
||||
if (!contextOrder || typeof contextOrder !== "object") return null;
|
||||
return {
|
||||
workerSetJson: workerItem!.content,
|
||||
contextOrder,
|
||||
boundTags: collectDictateBindProducts(board).map((product) => product.tag),
|
||||
changed: false,
|
||||
skipped: true,
|
||||
};
|
||||
} catch {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
/** 写入黑板：设计.worker集 + 设计.上下文投影排序。上游未改则不覆盖。 */
|
||||
export function applyDictatePlayBind(board: Blackboard): DictatePlayBindResult {
|
||||
const key = dictatePlayBindInputKey(board);
|
||||
const fresh = freshPlayBind(board, key);
|
||||
if (fresh) return fresh;
|
||||
|
||||
const result = bindDictateProductsToPlaySpec(board);
|
||||
const orderItem = board.getLatestByTag(CONTEXT_ORDER_TAG);
|
||||
const workerItem = board.getLatestByTag(WORKER_SET_FINAL_TAG);
|
||||
const stamp = { [DICTATE_PLAY_BIND_INPUT_META]: key };
|
||||
board.write({
|
||||
tag: WORKER_SET_FINAL_TAG,
|
||||
content: result.workerSetJson,
|
||||
source: "runtime",
|
||||
metadata: { ...(workerItem?.metadata ?? {}), ...stamp },
|
||||
});
|
||||
board.write({
|
||||
tag: CONTEXT_ORDER_TAG,
|
||||
content: JSON.stringify(result.contextOrder, null, 2),
|
||||
source: "runtime",
|
||||
metadata: { ...(orderItem?.metadata ?? {}), ...stamp },
|
||||
});
|
||||
return result;
|
||||
}
|
||||
|
||||
+99
-42
@@ -1,16 +1,13 @@
|
||||
/**
|
||||
* 对话落盘 · 创作进料专用系统提示
|
||||
*
|
||||
* 拼装：本提示（system）+ 当前产物（按相对序）+ 全量对话。
|
||||
* 仅用于创作层；开玩后由游玩管线接棒。
|
||||
* 对话落盘 · 创作系统提示。
|
||||
* 系统侧只放跨轮不变的规则。方案进度、能力目录和本轮输入放在请求末尾。
|
||||
*/
|
||||
|
||||
export function buildDictateSystemPrompt(opts?: {
|
||||
/** 已有方案 + 分档状态；存在时不再注入全能力目录 */
|
||||
export type DictatePromptTailOpts = {
|
||||
/** 已有方案的短进度；放在末尾，不进系统提示 */
|
||||
creationPlanState?: string;
|
||||
/** 尚无方案时使用的短能力目录 */
|
||||
/** 尚无方案时的短能力目录 */
|
||||
capabilityCatalog?: string;
|
||||
/** 尚无方案时可引用的旧配方案例目录 */
|
||||
recipeExamples?: string;
|
||||
/** @deprecated 仅兼容旧调用，作为案例而非已选路径 */
|
||||
recipeName?: string;
|
||||
@@ -18,43 +15,38 @@ export function buildDictateSystemPrompt(opts?: {
|
||||
recipeBrief?: string;
|
||||
/** @deprecated 使用 capabilityCatalog */
|
||||
moduleGuide?: string;
|
||||
}): string {
|
||||
const legacyExample =
|
||||
opts?.recipeName || opts?.recipeBrief
|
||||
? `
|
||||
【旧配方案例 · 兼容参考】
|
||||
${opts.recipeName ? `- ${opts.recipeName}` : ""}
|
||||
${opts.recipeBrief ?? ""}
|
||||
它只提供可改造的方法，不是用户已选路径，也不规定能力顺序。
|
||||
`
|
||||
: "";
|
||||
const decisionBlock = opts?.creationPlanState?.trim()
|
||||
? `${opts.creationPlanState.trim()}
|
||||
/** select：只定计划。fill：按注入规格写正文。reply：只交用户可见回复。 */
|
||||
phase?: "select" | "fill" | "reply";
|
||||
/** 本轮要落实或重写的 tag 与错误 */
|
||||
checklist?: string;
|
||||
};
|
||||
|
||||
方案是收口前的核对清单，不是本轮待办。写哪项先对照该能力旁的何时用 / 何时不用；档位只说明值不值得做。条件成立，且用户这句话已经说清的，通常 2～3 个：先 read_module，再连续 insert。说清指用户已经给出这份产物要写的内容，不是能从题材推演出一整份。标成「必须」或「有必要」、但这句话没覆盖到或使用条件不成立的，本轮不写，按已有材料给 2～3 个选项来问，选定或改写后再写入。写开场或结束创作前，列出仍未落的「必须」和「有必要」，问用户哪些还没设计、要补哪几项或明确不要。用户改口或方案明显失效时，先整份覆盖 \`设计.本局创作方案\`；使用条件仍留在能力旁。`
|
||||
: `【当前状态 · 尚无本局创作方案】
|
||||
先抽出用户要的核心体验，再对照能力目录主动标出有帮助的能力，生成 \`creation-plan.v1\`，用 insert 写入 \`设计.本局创作方案\`。先写 \`presentation_mode: "message"|"zero_layer"\`，再覆盖公共能力与该模式对应的三项呈现能力；两套呈现不得同时纳入。分档看能力对这份体验和可运行性的贡献：用户没点名、但缺了体验会明显变差的，标进「必须」或「有必要」，具体还没定的选择写进 missing。方案只表达本局方法、能力档位、理由与缺口，不生成固定流水线。
|
||||
/** 第一次请求：只从目录里点出这段话要填的节点，不写产物。 */
|
||||
export function buildDictateSelectPrompt(): string {
|
||||
return `你只做选题。对照能力目录和用户这句话，调用 select_nodes，选出已经够写的节点。
|
||||
|
||||
${opts?.capabilityCatalog?.trim() || opts?.moduleGuide?.trim() || "（能力目录暂不可用；目录恢复前不要猜造方案。）"}
|
||||
${opts?.recipeExamples?.trim() ? `\n\n【旧配方案例目录】\n${opts.recipeExamples.trim()}` : ""}
|
||||
${legacyExample}`;
|
||||
nodes 里的 module_id 必须来自目录。通常选这段话已经说清的 2～4 个。尚无本局方案、且这句话已经给出核心体验时，把 creation-plan 放进 nodes。
|
||||
用户的体验目标、举例和比方只说明要追求什么、像什么。范围还没和用户确认时，不为把它推成世界常态而选舞台骨架或实现机制。
|
||||
这一步不写产物正文，不回复用户。`;
|
||||
}
|
||||
|
||||
export function buildDictateSystemPrompt(_opts?: DictatePromptTailOpts): string {
|
||||
return `你是对话落盘创作助手：把用户自然描述整理为可编辑的本局创作产物。作用对象是创作黑板；开玩后由游玩 Worker 接棒。
|
||||
|
||||
**本提示只用于创作。** 开玩后叙事由游玩 worker 执行；创作期只落规格与引导，不写游玩回合正文。
|
||||
|
||||
${decisionBlock}
|
||||
**填写收尾：** 节点表产物写完后，同一次请求必须再调用 reply_user。只 insert、把话写在普通正文里，都不算完成。回复写法见文末「用户可见回复」。
|
||||
|
||||
## 写入规则
|
||||
设定 / 变量 / 映射必须经 tool 写入；聊天只做确认与引导，不当落盘载体。
|
||||
设定、变量、映射和删除通过本步的工具调用落盘。reply_user 只写给用户看的回复。
|
||||
|
||||
### insert — 写入/覆盖固定产物
|
||||
- position：\`用户.*\` / \`设计.*\`
|
||||
- content：正文；order 可选（同一变化频率组内，基础材料用较小值，越接近生成时使用的合同越大）
|
||||
- 同 position 再 insert = 整份覆盖改写；变量目录与映射改用 declare_variable / declare_map
|
||||
- 可反复能力每个实例用 \`基名#唯一id\`；改某实例时覆盖同一完整 tag
|
||||
- 只有方案中该填的内容都完成后才写开场白；先单独调用 prepare_opening，让程序编译信息可见范围、游玩执行方式、上下文排序和运行配置；工具成功后的下一步再生成并 insert 开场白
|
||||
- 开场白与生成规则、具体实例一样可增殖：每条入口 insert \`设计.开场白#场景短码\`（0 层卡用 \`设计.0层开场白与初态#场景短码\`）。改某一条则覆盖同一完整 tag。全部保存；游玩时左右切换，选中的一条作为 0 层。不要在回复里把多条入口收成一道先选定的题
|
||||
- 只有方案中该填的内容都完成后才写开场白。程序在这次填写前编译信息可见范围、游玩执行方式、上下文排序和运行配置；本步按清单写出开场正文
|
||||
- 开场白与生成规则、具体实例一样可增殖：每条入口写入 \`设计.开场白#场景短码\`（0 层卡用 \`设计.0层开场白与初态#场景短码\`）。改某一条则覆盖同一完整 tag。全部保存；游玩时左右切换，选中的一条作为 0 层。不要在回复里把多条入口收成一道先选定的题
|
||||
|
||||
### delete — 删除固定产物
|
||||
- position：已写入的 \`用户.*\` / \`设计.*\` tag，整份移除
|
||||
@@ -73,22 +65,87 @@ ${decisionBlock}
|
||||
- 移除一条 → remove_map（同 id 再 declare_map = 覆盖改写）
|
||||
|
||||
## 推进原则
|
||||
1. 本局创作方案是决策信息，不是 DAG；不要按 items 顺序逐项执行。有方案时，写哪项先对照该能力的何时用 / 何时不用。
|
||||
2. 写任何内容能力前先调用 read_module 读取当前提示版本；需要库条目时，再用 read_library_entry 读取当前能力绑定库条目（默认 1～2；确需更多须给 reason）。不要凭目录短说明猜产物。
|
||||
3. 本轮只写入用户这句话已经说清的能力，通常 2～3 个。对这些能力先 read_module，再 insert / declare_*。每次 insert 都是独立产物写入。
|
||||
4. 写某能力前以 read_module 返回的 task / principles / output / score 为准；尚未读到专用提示时，不猜造其复杂结构。能力提示里的「材料足够便直接成稿」只用于这句已经说清的能力。
|
||||
5. 「必须」和「有必要」要补齐：即使用户没点过能力名，也根据已抽出的核心体验和已有输入给 2～3 个选项来问，选定或改写后再写入。不要因为原话没出现就把有帮助的能力降成不推进。用户明确不要的「有必要」可以降档。「有一定效果」可顺带问一次。「没有意义」表示对这份体验没有帮助，不推进。未落的「必须」和「有必要」留到写开场或结束创作前，问用户哪些还没设计。
|
||||
6. 用户改口、新增方向或现有方案明显不成立时，先覆盖本局创作方案，再据此增删或改写其它产物。
|
||||
1. 本局创作方案是决策信息，不是 DAG；不要按 items 顺序逐项执行。尚无方案时，按末尾目录生成 \`creation-plan.v1\`，写入 \`设计.本局创作方案\`。有方案时，写哪项先对照该能力的何时用 / 何时不用。
|
||||
2. 本轮只处理用户这句话已经说清的能力，通常 2～3 个。说清指用户已经给出这份产物要写的内容，不是能从题材推演出一整份。
|
||||
3. 先定范围再写。用户的话按最窄的成立范围落盘：体验目标说明玩家追求的对象，举例和比方说明用户想要的感觉或基底，都不说明整个世界都是这样。例：体验是「在各个世界征服高洁女性、打败恶役反派」，落成每个世界里有这样的目标人物和对手；世界本身仍按题材常识（如普通西幻），不写成全世界女性都高洁，也不为此发明神殿戒律。用「和现实一样」时顺口提到的现实事实，只确认基底是现实，不单独成条。宽窄拿不准、且会写出不同产物时，按最窄范围写或先不写，在回复里请用户确认范围。
|
||||
4. 填写时以「本轮注入」里的 task、principles、output、score 和库正文为规格。注入尚未给出的能力，不猜造其复杂结构。
|
||||
5. 变量和映射不写入普通产物。每次产物写入都是独立的一份。
|
||||
6. 「必须」和「有必要」要补齐：即使用户没点过能力名，也根据已抽出的核心体验和已有输入给 2～3 个选项来问，选定或改写后再写入。不要因为原话没出现就把有帮助的能力降成不推进。用户明确不要的「有必要」可以降档。「有一定效果」可顺带问一次。「没有意义」表示对这份体验没有帮助，不推进。未落的「必须」和「有必要」留到写开场或结束创作前，问用户哪些还没设计。
|
||||
7. 用户改口、新增方向或现有方案明显不成立时，先覆盖本局创作方案，再据此增删或改写其它产物。
|
||||
|
||||
## 用户可见回复
|
||||
自然回应用户本轮有效判断、矛盾和值得保留之处，再简要说明形成或改写了什么。完整 JSON 由产物卡展示，不粘贴到回复。
|
||||
写正文组成或 0 层正文组成时，在回复里直接给出一份隔离内容供预览：与写入的 frontend 同一套骨架，示例内容已经放进去，单独可看。产物仍只保存 frontend 与数据契约。
|
||||
这句话没覆盖到的「必须」或「有必要」，用 2～3 个从用户原话推出的选项询问，不先写成产物。已经说清并写入的，说明形成了什么，不再另找问题。同轮多个已说清产物只做一次综合回应，不逐模块套固定标题。
|
||||
|
||||
## 其它
|
||||
对话明显过长时 clear_dialogue（产物与变量/映射保留）。
|
||||
${dictateReplyGuide()}
|
||||
`;
|
||||
}
|
||||
|
||||
/** 方案进度、目录和本轮任务。拼在请求末尾。 */
|
||||
export function formatDictateRoundTail(opts?: DictatePromptTailOpts): string {
|
||||
const legacyExample =
|
||||
opts?.recipeName || opts?.recipeBrief
|
||||
? `【旧配方案例 · 兼容参考】
|
||||
${opts.recipeName ? `- ${opts.recipeName}` : ""}
|
||||
${opts.recipeBrief ?? ""}
|
||||
它只提供可改造的方法，不是用户已选路径，也不规定能力顺序。`
|
||||
: "";
|
||||
const state = opts?.creationPlanState?.trim()
|
||||
? `${opts.creationPlanState.trim()}
|
||||
|
||||
方案是收口前的核对清单，不是本轮待办。写哪项先对照该能力旁的何时用 / 何时不用；档位只说明值不值得做。条件成立，且用户这句话已经说清的，通常 2～3 个。标成「必须」或「有必要」、但这句话没覆盖到或使用条件不成立的，本轮不写，给 2～3 个选项来问。写开场或结束创作前，列出仍未落的「必须」和「有必要」。用户改口或方案明显失效时，先整份覆盖 \`设计.本局创作方案\`。`
|
||||
: `【当前状态 · 尚无本局创作方案】
|
||||
先抽出用户要的核心体验，再对照能力目录生成 \`creation-plan.v1\`，写入 \`设计.本局创作方案\`。先写 \`presentation_mode: "message"|"zero_layer"\`，再覆盖公共能力与该模式对应的三项呈现能力；两套呈现不得同时纳入。分档看能力对这份体验和可运行性的贡献。方案只表达本局方法、能力档位、理由与缺口，不生成固定流水线。
|
||||
|
||||
${opts?.capabilityCatalog?.trim() || opts?.moduleGuide?.trim() || "（能力目录暂不可用；目录恢复前不要猜造方案。）"}
|
||||
${opts?.recipeExamples?.trim() ? `\n【旧配方案例目录】\n${opts.recipeExamples.trim()}` : ""}
|
||||
${legacyExample}`;
|
||||
|
||||
if (opts?.phase === "reply") {
|
||||
const state = opts?.creationPlanState?.trim()
|
||||
? `${opts.creationPlanState.trim()}
|
||||
|
||||
写回复时用这份进度核对：用户这句没覆盖到的「必须」或「有必要」，给 2～3 个选项来问；已写入的不再强行追问。`
|
||||
: "";
|
||||
return [
|
||||
"本步只调用 reply_user。对照本轮刚落盘产物和「本轮关注」写综合回复，含自评最稳处、缺口例子与选项。",
|
||||
opts?.checklist?.trim() ?? "",
|
||||
state,
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n\n");
|
||||
}
|
||||
|
||||
const phase =
|
||||
opts?.phase === "fill"
|
||||
? `本步先写「本轮节点表」里尚未通过的产物（有 self_score 维度时 insert 带上；已通过的以工具结果为准），写完后同一次请求必须调用 reply_user，正文按系统提示「用户可见回复」。`
|
||||
: "本步是第一次请求。只调用 select_nodes：这段话用于填充哪些节点。这一步不写产物正文，不回复用户。";
|
||||
|
||||
// 填写时把收尾要求放在方案进度之前，避免被大段进度挤到注意力末尾。
|
||||
return opts?.phase === "fill"
|
||||
? [phase, opts?.checklist?.trim() ?? "", state.trim()].filter(Boolean).join("\n\n")
|
||||
: [state.trim(), opts?.checklist?.trim() ?? "", phase].filter(Boolean).join("\n\n");
|
||||
}
|
||||
|
||||
/** @deprecated 用 buildDictateSystemPrompt() */
|
||||
export const DICTATE_SYSTEM_PROMPT = buildDictateSystemPrompt();
|
||||
|
||||
/** 用户可见回复正文要求（评分、缺口、建议）。填写与回复兜底共用。 */
|
||||
export function dictateReplyBodyGuide(): string {
|
||||
return `自然回应用户这句里值得保留、要纠正或互相矛盾的地方，再说明形成了什么。产物正文留在产物卡。正文组成、0 层正文组成和开场的画面也以产物卡为准。
|
||||
用 self_score 点出最稳的一处，以及分数一般或还有更好写法的缺口。每个缺口写三样：一个从原话推出来的短例子、一句它会让体验变成什么样、2～3 个可以接着说的选项。这份产物没有 self_score 时，用这句里还没定的具体点代替缺口。
|
||||
本轮按最窄范围写下、或因范围不明先没写的点，排在缺口最前面问：说明你按什么范围理解，给 2～3 个宽窄选项（如只是这个目标人物 / 这一类人物都这样 / 整个世界的规则）。
|
||||
这句话没覆盖到的「必须」或「有必要」，写出下一步可以提及哪些，还可能缺哪些节点也写在这里；还没有方案时，按这句已经说清的设定给 2～3 个可继续的方向。
|
||||
最后补一个和这句相关、用户可能想接着定的点，放在回复里，这一步不为它调用 insert。`;
|
||||
}
|
||||
|
||||
/** 用户可见回复。只放在默认系统提示里，填写尾巴和工具说明不再抄一遍。 */
|
||||
export function dictateReplyGuide(): string {
|
||||
return `insert 全部完成后，同一次请求必须调用 reply_user，生成给用户看的综合回复。多个产物合成一段。
|
||||
${dictateReplyBodyGuide()}`;
|
||||
}
|
||||
|
||||
/** 程序侧兜底回复请求：只看本轮已落盘产物与能力回复口径，只调 reply_user。不向填写步透露。 */
|
||||
export function buildDictateReplyPrompt(): string {
|
||||
return `你只生成给用户看的综合回复。对照本轮刚落盘的产物、附带的「本轮关注」和用户原话，调用 reply_user。
|
||||
多个产物合成一段，不要逐模块套标题。不要再 insert、delete 或改设定。
|
||||
|
||||
${dictateReplyBodyGuide()}`;
|
||||
}
|
||||
|
||||
@@ -143,6 +143,74 @@ function bodyOfFragment(content: string): Record<string, unknown> | null {
|
||||
return body as Record<string, unknown>;
|
||||
}
|
||||
|
||||
/** 多开场切换用的落盘族。每条必须是 基名#场景短码。 */
|
||||
export const OPENING_SLOT_FAMILY = "设计.开场白";
|
||||
/** 旧整包 tag。insert 时改写到 OPENING_SLOT_FAMILY#槽，避免再并列一份。 */
|
||||
export const LEGACY_OPENING_BUNDLE_FAMILY = "设计.开场白与开场变量";
|
||||
|
||||
const OPENING_SLOT_REQUIRED =
|
||||
"开场每条单独落盘，position 用「设计.开场白#场景短码」（例如 设计.开场白#opening-cycle2），供之后左右切换。不要写裸「设计.开场白」，也不要写「设计.开场白与开场变量」。改某一条则 insert 同一完整 tag。没有变量的卡不写开场变量，tag 仍然带场景短码。";
|
||||
|
||||
/** 旧整包族改写成开场族，槽位原样保留。 */
|
||||
export function rewriteLegacyOpeningTag(tag: string): string {
|
||||
const raw = tag.trim();
|
||||
if (dictateProductFamily(raw) !== LEGACY_OPENING_BUNDLE_FAMILY) return raw;
|
||||
const slot = dictateProductSlot(raw);
|
||||
return slot ? buildRepeatableProductTag(OPENING_SLOT_FAMILY, slot) : OPENING_SLOT_FAMILY;
|
||||
}
|
||||
|
||||
function openingCandidateRows(body: Record<string, unknown>): Record<string, unknown>[] {
|
||||
const raw = body.开场白候选 ?? body.候选;
|
||||
if (!Array.isArray(raw)) return [];
|
||||
return raw.filter(
|
||||
(row): row is Record<string, unknown> =>
|
||||
Boolean(row) && typeof row === "object" && !Array.isArray(row),
|
||||
);
|
||||
}
|
||||
|
||||
/**
|
||||
* 从单张开场正文抽出场景短码。多条候选挤在一份正文里时拒绝，要求拆成多次 insert。
|
||||
*/
|
||||
export function extractOpeningSlotFromContent(
|
||||
content: string,
|
||||
): { slot?: string; error?: string } {
|
||||
const parsed = tryParseJsonDoc(content);
|
||||
if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) return {};
|
||||
const doc = parsed as Record<string, unknown>;
|
||||
const nested = bodyOfFragment(content);
|
||||
const body = nested ?? doc;
|
||||
const direct =
|
||||
asString(doc.id) ??
|
||||
asString(doc.候选id) ??
|
||||
asString(doc.场景短码) ??
|
||||
asString(body.id) ??
|
||||
asString(body.候选id) ??
|
||||
asString(body.场景短码);
|
||||
if (direct) {
|
||||
const slot = sanitizeRepeatableSlot(direct);
|
||||
return slot ? { slot } : {};
|
||||
}
|
||||
const rows = openingCandidateRows(body);
|
||||
if (rows.length > 1) {
|
||||
return {
|
||||
error:
|
||||
"多条开场不要写进同一份产物。每条 insert 一次「设计.开场白#场景短码」。",
|
||||
};
|
||||
}
|
||||
const one = rows[0];
|
||||
const fromOne = one
|
||||
? asString(one.id) ?? asString(one.候选id) ?? asString(one.场景短码)
|
||||
: undefined;
|
||||
const fallback =
|
||||
rows.length <= 1
|
||||
? asString(body.默认候选id) ?? asString(body.默认候选ID)
|
||||
: undefined;
|
||||
const picked = fromOne ?? fallback;
|
||||
if (!picked) return {};
|
||||
const slot = sanitizeRepeatableSlot(picked);
|
||||
return slot ? { slot } : {};
|
||||
}
|
||||
|
||||
/**
|
||||
* 从正文抽出建议槽位：生成规则 → 规则名；具体实例 → 名称 / 记录里的姓名。
|
||||
*/
|
||||
@@ -207,8 +275,9 @@ export function resolveRepeatableInsertTag(params: {
|
||||
existingTags: readonly string[];
|
||||
repeatableFamilies: ReadonlySet<string>;
|
||||
}): ResolveRepeatableInsertResult {
|
||||
const raw = params.tag.trim();
|
||||
if (!raw) return { error: "position 为空" };
|
||||
const requested = params.tag.trim();
|
||||
if (!requested) return { error: "position 为空" };
|
||||
const raw = rewriteLegacyOpeningTag(requested);
|
||||
|
||||
const family = dictateProductFamily(raw);
|
||||
if (!params.repeatableFamilies.has(family)) {
|
||||
@@ -217,13 +286,21 @@ export function resolveRepeatableInsertTag(params: {
|
||||
|
||||
const explicit = dictateProductSlot(raw);
|
||||
const extracted = extractRepeatableSlotFromContent(family, params.content);
|
||||
const slot = explicit || extracted || "";
|
||||
const openingSlot =
|
||||
family === OPENING_SLOT_FAMILY && !explicit
|
||||
? extractOpeningSlotFromContent(params.content)
|
||||
: {};
|
||||
if (openingSlot.error) return { error: openingSlot.error };
|
||||
const slot = explicit || extracted || openingSlot.slot || "";
|
||||
|
||||
const siblings = params.existingTags
|
||||
.map((t) => t.trim())
|
||||
.filter((t) => t && dictateProductFamily(t) === family);
|
||||
|
||||
if (!slot) {
|
||||
if (family === OPENING_SLOT_FAMILY) {
|
||||
return { error: OPENING_SLOT_REQUIRED };
|
||||
}
|
||||
// 允许覆盖基名自身；若已有其它 #槽位，禁止再用裸基名（须写明 #id）
|
||||
const otherSiblings = siblings.filter((t) => t !== family);
|
||||
if (otherSiblings.length > 0 && raw === family) {
|
||||
@@ -235,11 +312,11 @@ export function resolveRepeatableInsertTag(params: {
|
||||
}
|
||||
|
||||
const resolved = buildRepeatableProductTag(family, slot);
|
||||
if (raw === family || raw !== resolved) {
|
||||
if (requested !== resolved) {
|
||||
return {
|
||||
tag: resolved,
|
||||
remapped: raw !== resolved,
|
||||
note: raw !== resolved ? `可增殖已拆到 ${resolved}` : undefined,
|
||||
remapped: true,
|
||||
note: `已写入 ${resolved}。之后改这一条用同一完整 tag。`,
|
||||
};
|
||||
}
|
||||
return { tag: resolved };
|
||||
|
||||
@@ -0,0 +1,721 @@
|
||||
/**
|
||||
* 一句用户话：先用 select_nodes 定目标节点，程序注入这些节点的写法，
|
||||
* 再在同一条消息前缀上循环 insert。已通过的不再覆盖。
|
||||
* 没有未通过项时，同一次请求里的 reply_user 才给用户。
|
||||
* 填写步若未调用 reply_user，程序再打一轮瘦回复（只带本轮刚落盘产物）；填写提示不提及此事。
|
||||
*/
|
||||
import { randomUUID } from "node:crypto";
|
||||
import type { DriverStepOutcome, DriverToolCall } from "../runtime/driver.js";
|
||||
import type { AgentDriver } from "../runtime/driver.js";
|
||||
import type { ChatMessage, LlmProvider, UsageDetail, UsageSink } from "../llm/client.js";
|
||||
import { DICTATE_REPLY_TOOL, DICTATE_SELECT_TOOL, DICTATE_WRITE_LOOP_TOOLS } from "./tools.js";
|
||||
import { buildDictateContext, type DictateSegmentChars } from "./context.js";
|
||||
import {
|
||||
CREATION_PLAN_SCHEMA,
|
||||
CREATION_PLAN_TAG,
|
||||
formatCreationPlanCapabilityCatalog,
|
||||
formatCreationPlanProgress,
|
||||
parseCreationPlan,
|
||||
} from "./creation-plan.js";
|
||||
import { isDictateOpeningTag } from "./experience-anchor.js";
|
||||
import {
|
||||
buildDictateReplyPrompt,
|
||||
buildDictateSelectPrompt,
|
||||
buildDictateSystemPrompt,
|
||||
formatDictateRoundTail,
|
||||
type DictatePromptTailOpts,
|
||||
} from "./prompt.js";
|
||||
import {
|
||||
parseDictateModuleReadState,
|
||||
readLibraryEntries,
|
||||
serializeDictateModuleReadState,
|
||||
DICTATE_MODULE_READ_STATE_TAG,
|
||||
type DictateModuleReadPayload,
|
||||
} from "./read-module.js";
|
||||
import { resolveModuleForInsertTag } from "./insert-feedback.js";
|
||||
import type { DictateChatTurn, DictateProduct } from "./types.js";
|
||||
import { isAllowedProductTag } from "./types.js";
|
||||
import { VALUE_MAP_TAG } from "../skills/value-map.js";
|
||||
import { VARIABLE_CATALOG_TAG } from "../skills/variable-catalog.js";
|
||||
|
||||
const READ_EXEMPT = new Set([
|
||||
"narrative",
|
||||
"variable-context",
|
||||
"context-order",
|
||||
"worker-spec",
|
||||
"refine",
|
||||
]);
|
||||
|
||||
type Apply = (calls: DriverToolCall[]) => DriverStepOutcome | Promise<DriverStepOutcome>;
|
||||
|
||||
export type PlannedRoundParams = {
|
||||
llm: LlmProvider;
|
||||
dialogue: DictateChatTurn[];
|
||||
handlers: {
|
||||
listProducts: () => DictateProduct[];
|
||||
readTag: (tag: string) => string | undefined;
|
||||
writeTag: (tag: string, content: string) => void;
|
||||
prepareOpening?: () => void;
|
||||
/** 本轮刚写入 tag 的能力回复口径（探测/原则/自评关注） */
|
||||
lookupInsertFeedback?: (tag: string) => string | undefined;
|
||||
onThinkingDelta?: (delta: string) => void;
|
||||
onThinkingDone?: (text: string) => void;
|
||||
};
|
||||
moduleCatalog?: import("../skills/creation-flow.js").ModuleCatalog | null;
|
||||
moduleReadIndex?: ReadonlyMap<string, DictateModuleReadPayload>;
|
||||
moduleGuide?: string;
|
||||
recipeExamples?: string;
|
||||
recipeName?: string;
|
||||
recipeBrief?: string;
|
||||
systemPrompt?: string;
|
||||
};
|
||||
|
||||
type PlanInsert = {
|
||||
tag: string;
|
||||
moduleId: string;
|
||||
libraryId?: string;
|
||||
entryIds: string[];
|
||||
reason?: string;
|
||||
};
|
||||
|
||||
type SelectPlan = {
|
||||
clarify: string[];
|
||||
inserts: PlanInsert[];
|
||||
deletes: string[];
|
||||
/** 模型对象上的字段名，空计划时写进回复，方便核对解析有没有对上 */
|
||||
keys: string[];
|
||||
};
|
||||
|
||||
|
||||
type ItemFailure = { tag: string; error: string };
|
||||
|
||||
|
||||
const INSERT_ATTEMPTS = 4;
|
||||
|
||||
export function createPlannedDictateDriver(params: PlannedRoundParams): AgentDriver {
|
||||
return {
|
||||
async run(input) {
|
||||
const reply = await runPlannedDictateRound(params, input.handleStep);
|
||||
return { iterations: reply.modelCalls, stop: { kind: "text", content: reply.text } };
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
async function runPlannedDictateRound(
|
||||
params: PlannedRoundParams,
|
||||
apply: Apply,
|
||||
): Promise<{ text: string; modelCalls: number }> {
|
||||
const turnId = randomUUID();
|
||||
const fillSystem = params.systemPrompt?.trim() || buildDictateSystemPrompt();
|
||||
const tailOpts = tailState(params);
|
||||
const selectTail = formatDictateRoundTail({ ...tailOpts, phase: "select" });
|
||||
const selectCtx = buildDictateContext({
|
||||
systemPrompt: buildDictateSelectPrompt(),
|
||||
products: params.handlers.listProducts(),
|
||||
dialogue: params.dialogue,
|
||||
phase: "select",
|
||||
tail: selectTail,
|
||||
});
|
||||
const selected = await params.llm.completeWithTools(selectCtx.messages, {
|
||||
caller: "dictate_agent",
|
||||
label: "对话落盘 选题",
|
||||
tools: [DICTATE_SELECT_TOOL],
|
||||
toolChoice: { type: "function", function: { name: "select_nodes" } },
|
||||
usageDetail: detail(turnId, "select", 0, selectCtx.segmentChars),
|
||||
usageSink: sinkFor(turnId, "select"),
|
||||
});
|
||||
if (selected.reasoning?.trim()) params.handlers.onThinkingDone?.(selected.reasoning.trim());
|
||||
const selectCall = selected.toolCalls.find((call) => call.name === "select_nodes") ?? selected.toolCalls[0];
|
||||
const plan = parseSelect(selectCall ? safeJson(selectCall.arguments) : {}, params);
|
||||
if (plan.clarify.length >= 2 && plan.inserts.length === 0 && plan.deletes.length === 0) {
|
||||
return {
|
||||
modelCalls: 1,
|
||||
text: plan.clarify.map((item, index) => `${index + 1}. ${item}`).join("\n"),
|
||||
};
|
||||
}
|
||||
if (plan.inserts.length === 0 && plan.deletes.length === 0) {
|
||||
const fields = plan.keys.length ? plan.keys.join("、") : "空对象";
|
||||
return {
|
||||
modelCalls: 1,
|
||||
text: `这一轮没有写入产物。选题返回的字段是 ${fields}，没有可执行的写入项。可以说得更具体些，或点明先写哪一块。`,
|
||||
};
|
||||
}
|
||||
|
||||
const blocked = plan.inserts.filter((item) => item.tag.startsWith("!"));
|
||||
const inserts = plan.inserts.filter((item) => !item.tag.startsWith("!"));
|
||||
const openings = inserts.filter((item) => isDictateOpeningTag(item.tag));
|
||||
const rest = inserts.filter((item) => !isDictateOpeningTag(item.tag));
|
||||
const blockedLines = blocked.map(
|
||||
(item) => `${item.tag.slice(1)}（${item.reason || "计划未通过程序核验"}）`,
|
||||
);
|
||||
if (openings.length > 0) params.handlers.prepareOpening?.();
|
||||
const injected = injectGroup(params, [...rest, ...openings]);
|
||||
const tail = formatDictateRoundTail({
|
||||
...tailOpts,
|
||||
phase: "fill",
|
||||
checklist: formatChecklist([...rest, ...openings], plan.deletes, []),
|
||||
});
|
||||
const ctx = buildDictateContext({
|
||||
systemPrompt: fillSystem,
|
||||
products: params.handlers.listProducts(),
|
||||
dialogue: params.dialogue,
|
||||
phase: "fill",
|
||||
injected: injected.text,
|
||||
tail,
|
||||
});
|
||||
const messages: ChatMessage[] = [...ctx.messages];
|
||||
const passed = new Set<string>();
|
||||
const pending = new Set([...rest, ...openings].map((item) => item.tag));
|
||||
const pendingDeletes = new Set(plan.deletes);
|
||||
const errors = new Map<string, string>();
|
||||
let replyFromTool = "";
|
||||
let lastProse = "";
|
||||
let modelCalls = 1;
|
||||
|
||||
for (let attempt = 0; attempt < INSERT_ATTEMPTS && (pending.size > 0 || pendingDeletes.size > 0); attempt++) {
|
||||
const step = attempt === 0 ? "fill" : "fill-retry";
|
||||
const filled = await params.llm.completeWithTools(messages, {
|
||||
caller: "dictate_agent",
|
||||
label: attempt === 0 ? "对话落盘 填写" : "对话落盘 重写",
|
||||
tools: DICTATE_WRITE_LOOP_TOOLS,
|
||||
toolChoice: "required",
|
||||
usageDetail: {
|
||||
...detail(turnId, step, attempt, ctx.segmentChars),
|
||||
plannedTags: [...pending, ...pendingDeletes],
|
||||
},
|
||||
usageSink: sinkFor(turnId, step),
|
||||
});
|
||||
modelCalls += 1;
|
||||
if (filled.reasoning?.trim()) params.handlers.onThinkingDone?.(filled.reasoning.trim());
|
||||
const prose = filled.content?.trim() ?? "";
|
||||
if (prose) lastProse = prose;
|
||||
const calls: DriverToolCall[] = filled.toolCalls.map((call, index) => ({
|
||||
id: call.id || `call-${attempt}-${index}`,
|
||||
name: call.name,
|
||||
arguments: call.arguments,
|
||||
}));
|
||||
if (calls.length === 0) break;
|
||||
const settled = await settleWriteStep(params, apply, calls, passed, pending, pendingDeletes, errors);
|
||||
messages.push(assistantToolMessage(filled.content, calls));
|
||||
for (const item of settled.results) {
|
||||
messages.push({ role: "tool", content: item.content, tool_call_id: item.callId });
|
||||
}
|
||||
if (pending.size === 0 && pendingDeletes.size === 0) {
|
||||
replyFromTool = settled.reply;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
let reply = replyFromTool;
|
||||
if (
|
||||
!replyFromTool &&
|
||||
pending.size === 0 &&
|
||||
pendingDeletes.size === 0 &&
|
||||
passed.size > 0
|
||||
) {
|
||||
reply = await runLeanReply(params, turnId, passed, plan.deletes, lastProse);
|
||||
modelCalls += 1;
|
||||
} else if (!reply) {
|
||||
reply = lastProse;
|
||||
}
|
||||
|
||||
const failed = [...pending, ...pendingDeletes].map(
|
||||
(tag) => `${tag}（${errors.get(tag) ?? "尚未通过"}）`,
|
||||
);
|
||||
const pendingLines = [...blockedLines, ...failed];
|
||||
const text = pendingLines.length
|
||||
? [reply, `尚未完成：${pendingLines.join("、")}。`].filter(Boolean).join("\n")
|
||||
: reply;
|
||||
return { text, modelCalls };
|
||||
}
|
||||
|
||||
async function runLeanReply(
|
||||
params: PlannedRoundParams,
|
||||
turnId: string,
|
||||
passed: Set<string>,
|
||||
plannedDeletes: string[],
|
||||
fallbackProse: string,
|
||||
): Promise<string> {
|
||||
const writtenTags = [...passed].filter((tag) => !tag.startsWith("del:"));
|
||||
const deletedTags = [...passed]
|
||||
.filter((tag) => tag.startsWith("del:"))
|
||||
.map((tag) => tag.slice(4))
|
||||
.filter(Boolean);
|
||||
const products = params.handlers
|
||||
.listProducts()
|
||||
.filter((product) => writtenTags.some((tag) => product.tag === tag || product.tag.startsWith(`${tag}#`)));
|
||||
const feedbackBlocks = writtenTags
|
||||
.map((tag) => params.handlers.lookupInsertFeedback?.(tag)?.trim())
|
||||
.filter((text): text is string => Boolean(text));
|
||||
if (writtenTags.length > 1 && feedbackBlocks.length > 1) {
|
||||
feedbackBlocks.unshift(
|
||||
"本轮写入了多个产物。用一段自然回复综合说明本轮变化、共同强项与关键缺口；不要逐模块套标题或重复模板。只有缺口确需用户决定时才提问。",
|
||||
);
|
||||
}
|
||||
const tailOpts = tailState(params);
|
||||
const deleteNote =
|
||||
deletedTags.length || plannedDeletes.length
|
||||
? `本轮已删除：${(deletedTags.length ? deletedTags : plannedDeletes).join("、")}`
|
||||
: "";
|
||||
const replyTail = [
|
||||
formatDictateRoundTail({
|
||||
...tailOpts,
|
||||
phase: "reply",
|
||||
checklist: [
|
||||
writtenTags.length ? `本轮已写入：${writtenTags.join("、")}` : "",
|
||||
deleteNote,
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n"),
|
||||
}),
|
||||
feedbackBlocks.length ? `## 本轮关注（按能力回复口径）\n\n${feedbackBlocks.join("\n\n")}` : "",
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n\n");
|
||||
const replyCtx = buildDictateContext({
|
||||
systemPrompt: buildDictateReplyPrompt(),
|
||||
products,
|
||||
dialogue: params.dialogue,
|
||||
phase: "reply",
|
||||
tail: replyTail,
|
||||
});
|
||||
const filled = await params.llm.completeWithTools(replyCtx.messages, {
|
||||
caller: "dictate_agent",
|
||||
label: "对话落盘 回复",
|
||||
tools: [DICTATE_REPLY_TOOL],
|
||||
toolChoice: { type: "function", function: { name: DICTATE_REPLY_TOOL.function.name } },
|
||||
usageDetail: {
|
||||
...detail(turnId, "reply", 0, replyCtx.segmentChars),
|
||||
plannedTags: writtenTags,
|
||||
},
|
||||
usageSink: sinkFor(turnId, "reply"),
|
||||
});
|
||||
if (filled.reasoning?.trim()) params.handlers.onThinkingDone?.(filled.reasoning.trim());
|
||||
const call = filled.toolCalls.find((item) => item.name === DICTATE_REPLY_TOOL.function.name);
|
||||
if (call) {
|
||||
const text = replyText({
|
||||
id: call.id || "reply",
|
||||
name: call.name,
|
||||
arguments: call.arguments,
|
||||
});
|
||||
if (text) return text;
|
||||
}
|
||||
return filled.content?.trim() || fallbackProse;
|
||||
}
|
||||
|
||||
function assistantToolMessage(content: string | null, calls: DriverToolCall[]): ChatMessage {
|
||||
return {
|
||||
role: "assistant",
|
||||
content,
|
||||
tool_calls: calls.map((call) => ({
|
||||
id: call.id,
|
||||
type: "function" as const,
|
||||
function: { name: call.name, arguments: call.arguments },
|
||||
})),
|
||||
};
|
||||
}
|
||||
|
||||
async function settleWriteStep(
|
||||
params: PlannedRoundParams,
|
||||
apply: Apply,
|
||||
calls: DriverToolCall[],
|
||||
passed: Set<string>,
|
||||
pending: Set<string>,
|
||||
pendingDeletes: Set<string>,
|
||||
errors: Map<string, string>,
|
||||
): Promise<{ results: Array<{ callId: string; content: string }>; reply: string }> {
|
||||
const results: Array<{ callId: string; content: string }> = [];
|
||||
const forward: DriverToolCall[] = [];
|
||||
let reply = "";
|
||||
for (const call of calls) {
|
||||
if (call.name === DICTATE_REPLY_TOOL.function.name) continue;
|
||||
const tag = callTag(call, params);
|
||||
if (call.name === "insert" && tag && passed.has(tag)) {
|
||||
results.push({
|
||||
callId: call.id,
|
||||
content: JSON.stringify({ error: "已通过，本步不要再写", tag }),
|
||||
});
|
||||
continue;
|
||||
}
|
||||
if (call.name === "delete" && tag && passed.has(`del:${tag}`)) {
|
||||
results.push({
|
||||
callId: call.id,
|
||||
content: JSON.stringify({ error: "已通过，本步不要再删除", tag }),
|
||||
});
|
||||
continue;
|
||||
}
|
||||
if (call.name === "insert" && tag) {
|
||||
const args = asRecord(safeJson(call.arguments));
|
||||
forward.push({ ...call, arguments: JSON.stringify({ ...args, position: tag }) });
|
||||
continue;
|
||||
}
|
||||
forward.push(call);
|
||||
}
|
||||
const outcome = await apply(forward);
|
||||
if (outcome.kind === "continue") {
|
||||
for (const result of outcome.results) {
|
||||
results.push(result);
|
||||
const call = forward.find((item) => item.id === result.callId);
|
||||
const payload = asRecord(safeJson(result.content));
|
||||
const tag = call ? callTag(call, params) : "";
|
||||
if (payload.ok === true && tag) {
|
||||
if (call?.name === "delete") {
|
||||
passed.add(`del:${tag}`);
|
||||
pendingDeletes.delete(tag);
|
||||
} else {
|
||||
passed.add(tag);
|
||||
for (const planned of [...pending]) {
|
||||
if (planned === tag || tag.startsWith(`${planned}#`)) pending.delete(planned);
|
||||
}
|
||||
}
|
||||
errors.delete(tag);
|
||||
} else if (typeof payload.error === "string" && tag) {
|
||||
const planned = [...pending, ...pendingDeletes].find(
|
||||
(item) => item === tag || tag.startsWith(`${item}#`),
|
||||
);
|
||||
errors.set(planned ?? tag, payload.error);
|
||||
}
|
||||
}
|
||||
}
|
||||
for (const call of calls) {
|
||||
if (call.name !== DICTATE_REPLY_TOOL.function.name) continue;
|
||||
const text = replyText(call);
|
||||
if (pending.size > 0 || pendingDeletes.size > 0) {
|
||||
const left = [...pending, ...pendingDeletes].join("、");
|
||||
results.push({
|
||||
callId: call.id,
|
||||
content: JSON.stringify({
|
||||
error: `还有未通过：${left}。这份回复先不给用户。只再写入这些。`,
|
||||
}),
|
||||
});
|
||||
continue;
|
||||
}
|
||||
reply = text;
|
||||
results.push({ callId: call.id, content: JSON.stringify({ ok: true }) });
|
||||
}
|
||||
return { results, reply };
|
||||
}
|
||||
|
||||
function callTag(call: DriverToolCall, params: PlannedRoundParams): string {
|
||||
const args = asRecord(safeJson(call.arguments));
|
||||
const raw = String(args.position ?? args.tag ?? "").trim();
|
||||
if (!raw) return "";
|
||||
return canonicalProductTag(raw, "", params.moduleCatalog)?.tag ?? raw;
|
||||
}
|
||||
|
||||
function replyText(call: DriverToolCall): string {
|
||||
const args = asRecord(safeJson(call.arguments));
|
||||
return String(args.text ?? args.reply ?? args.content ?? args.message ?? "").trim();
|
||||
}
|
||||
|
||||
function sinkFor(_turnId: string, _step: string): UsageSink {
|
||||
return {};
|
||||
}
|
||||
|
||||
function detail(
|
||||
turnId: string,
|
||||
step: UsageDetail["step"],
|
||||
retryIndex: number,
|
||||
segmentChars: DictateSegmentChars,
|
||||
): UsageDetail {
|
||||
return { turnId, step, retryIndex, segmentChars };
|
||||
}
|
||||
|
||||
function tailState(params: PlannedRoundParams): DictatePromptTailOpts {
|
||||
const products = params.handlers.listProducts();
|
||||
const plan = parseCreationPlan(
|
||||
products.find((product) => product.tag === CREATION_PLAN_TAG)?.content,
|
||||
);
|
||||
if (plan) {
|
||||
return {
|
||||
creationPlanState: formatCreationPlanProgress({
|
||||
plan,
|
||||
catalog: params.moduleCatalog,
|
||||
products,
|
||||
}),
|
||||
};
|
||||
}
|
||||
return {
|
||||
capabilityCatalog:
|
||||
formatCreationPlanCapabilityCatalog(params.moduleCatalog) || params.moduleGuide,
|
||||
recipeExamples: params.recipeExamples,
|
||||
recipeName: params.recipeName,
|
||||
recipeBrief: params.recipeBrief,
|
||||
moduleGuide: params.moduleGuide,
|
||||
};
|
||||
}
|
||||
|
||||
function parseSelect(raw: unknown, params: PlannedRoundParams): SelectPlan {
|
||||
const row = asRecord(raw);
|
||||
const clarify = stringList(row.clarify);
|
||||
let deletes = stringList(row.deletes).filter((tag) => isAllowedProductTag(tag));
|
||||
const inserts: PlanInsert[] = [];
|
||||
let items = Array.isArray(row.nodes)
|
||||
? row.nodes
|
||||
: Array.isArray(row.inserts)
|
||||
? row.inserts
|
||||
: [];
|
||||
if (items.length === 0 && clarify.length < 2 && deletes.length === 0) {
|
||||
const recovered = recoverSelectWork(row);
|
||||
items = recovered.items;
|
||||
deletes = recovered.deletes.filter((tag) => isAllowedProductTag(tag));
|
||||
}
|
||||
for (const item of items) {
|
||||
const rec = asRecord(item);
|
||||
const moduleIdHint = String(rec.module_id ?? rec.moduleId ?? "").trim();
|
||||
let tag = String(rec.tag ?? rec.position ?? "").trim();
|
||||
if (!tag && moduleIdHint) {
|
||||
tag =
|
||||
params.moduleCatalog?.modules.find((module) => module.id === moduleIdHint)?.artifact?.trim() ??
|
||||
"";
|
||||
}
|
||||
if (!tag) continue;
|
||||
const canonical = canonicalProductTag(tag, moduleIdHint, params.moduleCatalog);
|
||||
if (canonical) tag = canonical.tag;
|
||||
const resolvedHint = canonical?.moduleId || moduleIdHint;
|
||||
const module =
|
||||
params.moduleReadIndex?.get(resolvedHint) ??
|
||||
params.moduleReadIndex?.get(tag) ??
|
||||
undefined;
|
||||
const fromTag = resolveModuleForInsertTag(tag, params.moduleCatalog);
|
||||
const moduleId = module?.module_id || canonical?.moduleId || fromTag?.id || moduleIdHint;
|
||||
const entryIds = stringList(rec.entry_ids);
|
||||
const reason = String(rec.reason ?? "").trim();
|
||||
if (!isAllowedProductTag(tag) && tag !== VARIABLE_CATALOG_TAG && tag !== VALUE_MAP_TAG) {
|
||||
inserts.push({ tag: `!${tag || "空"}`, moduleId, entryIds, reason: "产物 tag 不合法" });
|
||||
continue;
|
||||
}
|
||||
const payload = moduleId ? params.moduleReadIndex?.get(moduleId) : undefined;
|
||||
if (payload && !READ_EXEMPT.has(payload.module_id) && payload.libraries.length) {
|
||||
const libraryId = String(rec.library_id ?? payload.libraries[0]?.library_id ?? "").trim();
|
||||
if (entryIds.length > 2 && !reason) {
|
||||
inserts.push({
|
||||
tag: `!${tag}`,
|
||||
moduleId,
|
||||
entryIds,
|
||||
reason: "默认最多两个库条目",
|
||||
});
|
||||
continue;
|
||||
}
|
||||
if (libraryId && entryIds.length) {
|
||||
const bound = payload.libraries.some((library) => library.library_id === libraryId);
|
||||
if (!bound) {
|
||||
inserts.push({ tag: `!${tag}`, moduleId, entryIds, reason: `能力未绑定库 ${libraryId}` });
|
||||
continue;
|
||||
}
|
||||
}
|
||||
}
|
||||
inserts.push({
|
||||
tag,
|
||||
moduleId,
|
||||
libraryId: String(rec.library_id ?? "").trim() || undefined,
|
||||
entryIds,
|
||||
reason: reason || undefined,
|
||||
});
|
||||
}
|
||||
return { clarify, inserts, deletes, keys: Object.keys(row) };
|
||||
}
|
||||
|
||||
/** 选题没交 inserts 时，从方案正文、operations 或别名字段收出本轮要写的能力，再自动发填写。 */
|
||||
function recoverSelectWork(row: Record<string, unknown>): {
|
||||
items: unknown[];
|
||||
deletes: string[];
|
||||
} {
|
||||
if (isCreationPlanShape(row)) {
|
||||
return { items: [{ tag: CREATION_PLAN_TAG, module_id: "creation-plan" }], deletes: [] };
|
||||
}
|
||||
const items: unknown[] = [];
|
||||
const deletes: string[] = [];
|
||||
for (const list of [row.modules, row.selections, row.selected, row.writes, row.products, row.operations]) {
|
||||
for (const item of asItemList(list)) {
|
||||
const rec = flattenItem(item);
|
||||
if (
|
||||
typeof rec.tier === "string" &&
|
||||
typeof rec.why === "string" &&
|
||||
rec.tag == null &&
|
||||
rec.position == null
|
||||
) {
|
||||
continue;
|
||||
}
|
||||
const op = operationName(rec);
|
||||
if (op === "read_module" || op === "read_library_entry" || op === "prepare_opening") {
|
||||
continue;
|
||||
}
|
||||
const moduleId = String(rec.module_id ?? rec.moduleId ?? "").trim();
|
||||
const tag = String(rec.tag ?? rec.position ?? rec.artifact ?? "").trim();
|
||||
if (op === "delete" || op === "remove") {
|
||||
if (tag) deletes.push(tag);
|
||||
continue;
|
||||
}
|
||||
if (
|
||||
op === "declare_variable" ||
|
||||
op === "undeclare_variable" ||
|
||||
op === "declare_map" ||
|
||||
op === "remove_map"
|
||||
) {
|
||||
continue;
|
||||
}
|
||||
if (!tag && !moduleId) continue;
|
||||
items.push({
|
||||
tag,
|
||||
module_id: moduleId,
|
||||
...(rec.library_id != null ? { library_id: rec.library_id } : {}),
|
||||
...(rec.entry_ids != null ? { entry_ids: rec.entry_ids } : {}),
|
||||
...(rec.reason != null ? { reason: rec.reason } : {}),
|
||||
});
|
||||
}
|
||||
}
|
||||
return { items, deletes };
|
||||
}
|
||||
|
||||
function isCreationPlanShape(row: Record<string, unknown>): boolean {
|
||||
if (row.schema === CREATION_PLAN_SCHEMA) return true;
|
||||
if (!Array.isArray(row.items) || row.items.length === 0) return false;
|
||||
return row.items.every((item) => {
|
||||
const rec = asRecord(item);
|
||||
return (
|
||||
typeof rec.module_id === "string" &&
|
||||
rec.module_id.trim().length > 0 &&
|
||||
typeof rec.tier === "string" &&
|
||||
rec.tag == null &&
|
||||
rec.position == null
|
||||
);
|
||||
});
|
||||
}
|
||||
|
||||
function injectGroup(
|
||||
params: PlannedRoundParams,
|
||||
group: PlanInsert[],
|
||||
): { text: string } {
|
||||
const blocks: string[] = [];
|
||||
const state = parseDictateModuleReadState(
|
||||
params.handlers.readTag(DICTATE_MODULE_READ_STATE_TAG),
|
||||
);
|
||||
let changed = false;
|
||||
for (const item of group) {
|
||||
const payload = item.moduleId ? params.moduleReadIndex?.get(item.moduleId) : undefined;
|
||||
if (!payload) continue;
|
||||
const catalogEntry = resolveModuleForInsertTag(item.tag, params.moduleCatalog);
|
||||
if (!READ_EXEMPT.has(payload.module_id) && !catalogEntry?.auto) {
|
||||
state.modules[payload.module_id] = payload.version;
|
||||
state.active_module_id = payload.module_id;
|
||||
changed = true;
|
||||
}
|
||||
const parts = [
|
||||
`### ${payload.module_id} @ ${payload.version}`,
|
||||
`产物 tag：${item.tag}`,
|
||||
payload.score_dimensions.length
|
||||
? `self_score 维度（这次写入必填，name 必须一致）：${payload.score_dimensions.join("、")}`
|
||||
: "",
|
||||
`#### task\n${payload.task}`,
|
||||
`#### principles\n${payload.principles}`,
|
||||
`#### output\n${payload.output}`,
|
||||
`#### score\n${payload.score}`,
|
||||
];
|
||||
if (item.libraryId && item.entryIds.length) {
|
||||
const loaded = readLibraryEntries({
|
||||
module: payload,
|
||||
libraryId: item.libraryId,
|
||||
entryIds: item.entryIds,
|
||||
reason: item.reason,
|
||||
});
|
||||
if (loaded.entries) {
|
||||
for (const entry of loaded.entries) {
|
||||
const row = entry as { id?: string; content?: string; name?: string };
|
||||
parts.push(`#### 库 ${item.libraryId} / ${row.id ?? ""}\n${row.content ?? ""}`);
|
||||
}
|
||||
} else if (loaded.error) {
|
||||
parts.push(`#### 库读取失败\n${loaded.error}`);
|
||||
}
|
||||
}
|
||||
blocks.push(parts.filter(Boolean).join("\n\n"));
|
||||
}
|
||||
if (changed) {
|
||||
params.handlers.writeTag(DICTATE_MODULE_READ_STATE_TAG, serializeDictateModuleReadState(state));
|
||||
}
|
||||
return { text: blocks.join("\n\n") };
|
||||
}
|
||||
|
||||
function canonicalProductTag(
|
||||
tag: string,
|
||||
moduleIdHint: string,
|
||||
catalog: PlannedRoundParams["moduleCatalog"],
|
||||
): { tag: string; moduleId: string } | null {
|
||||
const bare = tag.replace(/^(?:用户|设计)\./, "").split("#")[0]?.trim() ?? "";
|
||||
const entry =
|
||||
catalog?.modules.find((module) => module.id === moduleIdHint) ??
|
||||
resolveModuleForInsertTag(tag, catalog) ??
|
||||
catalog?.modules.find(
|
||||
(module) =>
|
||||
module.name === bare ||
|
||||
module.id === bare ||
|
||||
module.artifact === tag ||
|
||||
module.artifact.endsWith(`.${bare}`),
|
||||
);
|
||||
if (!entry?.artifact?.trim()) return null;
|
||||
const hash = tag.indexOf("#");
|
||||
const suffix = hash > 0 ? tag.slice(hash) : "";
|
||||
const next =
|
||||
entry.repeatable && suffix.startsWith("#") && suffix.length > 1
|
||||
? `${entry.artifact.trim()}${suffix}`
|
||||
: entry.artifact.trim();
|
||||
return { tag: next, moduleId: entry.id };
|
||||
}
|
||||
|
||||
|
||||
function formatChecklist(group: PlanInsert[], deletes: string[], errors: ItemFailure[]): string {
|
||||
const nodes = group.map((item) => item.moduleId || item.tag).filter(Boolean);
|
||||
const lines = [
|
||||
nodes.length ? `第一轮选定的节点：${nodes.join("、")}` : "",
|
||||
group.length ? `本轮写入：${group.map((item) => item.tag).join("、")}` : "",
|
||||
deletes.length ? `本轮删除：${deletes.join("、")}` : "",
|
||||
errors.length
|
||||
? `只重写未成功项：\n${errors.map((item) => `- ${item.tag}：${item.error}`).join("\n")}`
|
||||
: "",
|
||||
];
|
||||
return lines.filter(Boolean).join("\n");
|
||||
}
|
||||
|
||||
function asItemList(value: unknown): unknown[] {
|
||||
if (Array.isArray(value)) return value;
|
||||
const rec = asRecord(value);
|
||||
const values = Object.values(rec);
|
||||
if (values.length > 0 && values.every((item) => item && typeof item === "object")) {
|
||||
return values;
|
||||
}
|
||||
return [];
|
||||
}
|
||||
|
||||
function flattenItem(item: unknown): Record<string, unknown> {
|
||||
if (typeof item === "string") {
|
||||
const text = item.trim();
|
||||
if (!text) return {};
|
||||
if (text.startsWith("设计.") || text.startsWith("用户.")) return { tag: text };
|
||||
return { module_id: text };
|
||||
}
|
||||
const rec = asRecord(item);
|
||||
const nested = asRecord(rec.arguments ?? rec.args ?? rec.payload);
|
||||
return { ...rec, ...nested };
|
||||
}
|
||||
|
||||
function operationName(rec: Record<string, unknown>): string {
|
||||
return String(rec.op ?? rec.name ?? rec.type ?? rec.action ?? "insert").trim();
|
||||
}
|
||||
|
||||
function asRecord(value: unknown): Record<string, unknown> {
|
||||
if (!value || typeof value !== "object" || Array.isArray(value)) return {};
|
||||
return value as Record<string, unknown>;
|
||||
}
|
||||
|
||||
function stringList(value: unknown): string[] {
|
||||
if (!Array.isArray(value)) return [];
|
||||
return value.map((item) => String(item ?? "").trim()).filter(Boolean);
|
||||
}
|
||||
|
||||
function safeJson(raw: string): unknown {
|
||||
try {
|
||||
return JSON.parse(raw);
|
||||
} catch {
|
||||
return {};
|
||||
}
|
||||
}
|
||||
+68
-2
@@ -3,6 +3,66 @@ import type { ToolDefinition } from "../llm/client.js";
|
||||
/**
|
||||
* 对话落盘创作期工具：产物 / 变量 / 映射均须 toolcall；聊天只做确认与建议。
|
||||
*/
|
||||
/** 第一次请求仅此一个工具：点出下一步要写的节点。 */
|
||||
export const DICTATE_SELECT_TOOL: ToolDefinition = {
|
||||
type: "function",
|
||||
function: {
|
||||
name: "select_nodes",
|
||||
description:
|
||||
"规定下一步要写入的目标节点。nodes 只填能力目录中的 module_id。这一步不写产物正文。",
|
||||
parameters: {
|
||||
type: "object",
|
||||
additionalProperties: false,
|
||||
properties: {
|
||||
nodes: {
|
||||
type: "array",
|
||||
description: "这段用户话要填充的节点。",
|
||||
items: {
|
||||
type: "object",
|
||||
additionalProperties: false,
|
||||
properties: {
|
||||
module_id: { type: "string", description: "能力目录中的 id" },
|
||||
},
|
||||
required: ["module_id"],
|
||||
},
|
||||
},
|
||||
},
|
||||
required: ["nodes"],
|
||||
},
|
||||
},
|
||||
};
|
||||
|
||||
/** 写入循环的收尾。未通过的节点还在时，程序不把这份正文给用户。 */
|
||||
export const DICTATE_REPLY_TOOL: ToolDefinition = {
|
||||
type: "function",
|
||||
function: {
|
||||
name: "reply_user",
|
||||
description:
|
||||
"本步必调。与 insert 放在同一次请求：节点表写完后立刻提交给用户的综合回复。正文按系统提示「用户可见回复」写。",
|
||||
parameters: {
|
||||
type: "object",
|
||||
additionalProperties: false,
|
||||
properties: {
|
||||
text: {
|
||||
type: "string",
|
||||
description:
|
||||
"综合回复。先回应原话并说明形成了什么，再写自评最稳处；每个分数一般的缺口含短例子、一句体验变化和 2～3 个可接着说的选项；接着写下一步可以提及哪些，并补一个相关设定。多个产物合成一段。",
|
||||
},
|
||||
},
|
||||
required: ["text"],
|
||||
},
|
||||
},
|
||||
};
|
||||
|
||||
const WRITE_LOOP_TOOL_NAMES = new Set([
|
||||
"insert",
|
||||
"delete",
|
||||
"declare_variable",
|
||||
"undeclare_variable",
|
||||
"declare_map",
|
||||
"remove_map",
|
||||
]);
|
||||
|
||||
export const DICTATE_TOOL_DEFINITIONS: ToolDefinition[] = [
|
||||
{
|
||||
type: "function",
|
||||
@@ -80,12 +140,12 @@ export const DICTATE_TOOL_DEFINITIONS: ToolDefinition[] = [
|
||||
position: {
|
||||
type: "string",
|
||||
description:
|
||||
"产物 tag，须以「用户.」或「设计.」开头。例：用户.需求、设计.正文组成、设计.开场白、设计.开场白#dorm；可增殖例：设计.生成规则#rule-id、设计.具体实例#batch-id、设计.主角设定#dorm。勿用本工具写变量目录/映射。",
|
||||
"产物 tag，须以「用户.」或「设计.」开头。例：用户.需求、设计.正文组成、设计.开场白#opening-cycle2；可增殖例：设计.生成规则#rule-id、设计.具体实例#batch-id、设计.主角设定#dorm。开场必须带场景短码，不要写裸「设计.开场白」或「设计.开场白与开场变量」。勿用本工具写变量目录/映射。",
|
||||
},
|
||||
content: {
|
||||
type: "string",
|
||||
description:
|
||||
"落入该位置的正文。设计.正文组成=格式 JSON；设计.开场白[#短码]=叙事正文（有正文组成时为 present.v1 JSON；否则 Markdown），必须含字面 @玩家，禁止写入用户角色.名字/简介；设计.主角设定[#同短码]=名字/背景/特殊设定 JSON，与开场同短码即绑定该开局",
|
||||
"落入该位置的正文。设计.正文组成=格式 JSON；设计.开场白#场景短码=这一张开场的 present.v1，blocks 必须含正文组成数据契约里的全部模型生成区（用来灌进已固定前端并替换示例），不要只写 shell+body，也不要填程序投影区；必须含字面 @玩家，名字不进 blocks。这条开场的玩家角色写在 meta.用户角色（名字+一到三句客观事实），程序同步为同短码主角设定。这张开场有初值才在同一份正文里写开场变量，没有就不写。设计.主角设定=共通的名字/背景/特殊设定 JSON；各开场不同的玩家角色写进开场 meta，不另写主角设定#短码",
|
||||
},
|
||||
order: {
|
||||
type: "number",
|
||||
@@ -283,3 +343,9 @@ export const DICTATE_TOOL_DEFINITIONS: ToolDefinition[] = [
|
||||
},
|
||||
},
|
||||
];
|
||||
|
||||
/** 第二次起的工具。不含 read_module：写法已由程序注入。同一轮重试不改变这份清单。 */
|
||||
export const DICTATE_WRITE_LOOP_TOOLS: ToolDefinition[] = [
|
||||
...DICTATE_TOOL_DEFINITIONS.filter((tool) => WRITE_LOOP_TOOL_NAMES.has(tool.function.name)),
|
||||
DICTATE_REPLY_TOOL,
|
||||
];
|
||||
|
||||
+146
-3
@@ -1,9 +1,9 @@
|
||||
import type { LlmProvider } from "../llm/client.js";
|
||||
import {
|
||||
createLocalLlmDriver,
|
||||
DriverIterationLimitError,
|
||||
type AgentDriver,
|
||||
} from "../runtime/driver.js";
|
||||
import { createPlannedDictateDriver } from "./round.js";
|
||||
import { buildDictateMessages } from "./context.js";
|
||||
import { buildDictateSystemPrompt } from "./prompt.js";
|
||||
import { DICTATE_TOOL_DEFINITIONS } from "./tools.js";
|
||||
@@ -41,8 +41,14 @@ import {
|
||||
serializeVariableCatalog,
|
||||
upsertVariableField,
|
||||
removeVariableField,
|
||||
seedVariablesFromCatalog,
|
||||
type UpsertVariableInput,
|
||||
} from "../skills/variable-catalog.js";
|
||||
import {
|
||||
VARIABLE_DESIGN_TAG,
|
||||
syncVariableDesignToRuntime,
|
||||
writeVariableDesignSync,
|
||||
} from "../skills/variable-design-compile.js";
|
||||
import {
|
||||
VALUE_MAP_TAG,
|
||||
parseValueMapDoc,
|
||||
@@ -57,6 +63,46 @@ import {
|
||||
} from "../blackboard/table-cells.js";
|
||||
import { isDictateOpeningTag } from "./experience-anchor.js";
|
||||
import { canonicalizeInsertedContent } from "../parse/json-doc.js";
|
||||
import {
|
||||
openingContractError,
|
||||
parseReplyDataContract,
|
||||
replyContractSummary,
|
||||
type ReplyDataContract,
|
||||
} from "../skills/reply-contract.js";
|
||||
|
||||
function seedCurrentFromCatalog(handlers: DictateTurnHandlers): string[] {
|
||||
const seeded = seedVariablesFromCatalog({
|
||||
catalogRaw: handlers.readTag(VARIABLE_CATALOG_TAG),
|
||||
currentRaw: handlers.readTag("变量.当前"),
|
||||
initialRaw: handlers.readTag("运行.初始变量"),
|
||||
actor: "system:declare_variable",
|
||||
});
|
||||
const wrote: string[] = [];
|
||||
if (seeded.initial) {
|
||||
handlers.writeTag("运行.初始变量", seeded.initial);
|
||||
wrote.push("运行.初始变量");
|
||||
}
|
||||
if (seeded.current) {
|
||||
handlers.writeTag("变量.当前", seeded.current);
|
||||
wrote.push("变量.当前");
|
||||
}
|
||||
return wrote;
|
||||
}
|
||||
|
||||
function syncDesignProductToTable(
|
||||
handlers: DictateTurnHandlers,
|
||||
designContent?: string,
|
||||
): string[] {
|
||||
const sync = syncVariableDesignToRuntime({
|
||||
designRaw: designContent ?? handlers.readTag(VARIABLE_DESIGN_TAG),
|
||||
catalogRaw: handlers.readTag(VARIABLE_CATALOG_TAG),
|
||||
currentRaw: handlers.readTag("变量.当前"),
|
||||
initialRaw: handlers.readTag("运行.初始变量"),
|
||||
});
|
||||
return writeVariableDesignSync(sync, (tag, content) =>
|
||||
handlers.writeTag(tag, content),
|
||||
);
|
||||
}
|
||||
|
||||
export type DictateTurnHandlers = {
|
||||
listProducts: () => DictateProduct[];
|
||||
@@ -127,7 +173,28 @@ export async function runDictateTurn(params: {
|
||||
repeatableFamilies?: ReadonlySet<string> | readonly string[];
|
||||
driver?: AgentDriver;
|
||||
}): Promise<DictateTurnResult> {
|
||||
const driver = params.driver ?? createLocalLlmDriver(params.llm);
|
||||
const driver =
|
||||
params.driver ??
|
||||
createPlannedDictateDriver({
|
||||
llm: params.llm,
|
||||
dialogue: params.dialogue,
|
||||
handlers: {
|
||||
listProducts: params.handlers.listProducts,
|
||||
readTag: params.handlers.readTag,
|
||||
writeTag: params.handlers.writeTag,
|
||||
prepareOpening: params.handlers.prepareOpening,
|
||||
lookupInsertFeedback: params.handlers.lookupInsertFeedback,
|
||||
onThinkingDelta: params.handlers.onThinkingDelta,
|
||||
onThinkingDone: params.handlers.onThinkingDone,
|
||||
},
|
||||
moduleCatalog: params.moduleCatalog,
|
||||
moduleReadIndex: params.moduleReadIndex,
|
||||
moduleGuide: params.moduleGuide,
|
||||
recipeExamples: params.recipeExamples,
|
||||
recipeName: params.recipeName,
|
||||
recipeBrief: params.recipeBrief,
|
||||
systemPrompt: params.systemPrompt,
|
||||
});
|
||||
const wroteTags: string[] = [];
|
||||
const deletedTags: string[] = [];
|
||||
let clearedDialogue = false;
|
||||
@@ -217,9 +284,31 @@ export async function runDictateTurn(params: {
|
||||
call.name,
|
||||
`${module.module_id}@${module.version}`,
|
||||
);
|
||||
const replyContract =
|
||||
module.module_id === "opening-setup" ||
|
||||
module.module_id === "zero-layer-opening-setup"
|
||||
? contractForOpeningTag(
|
||||
module.module_id === "zero-layer-opening-setup"
|
||||
? "设计.0层开场白与初态"
|
||||
: "设计.开场白",
|
||||
params.handlers.listProducts(),
|
||||
)
|
||||
: undefined;
|
||||
return {
|
||||
callId: call.id,
|
||||
content: JSON.stringify({ ok: true, ...module }),
|
||||
content: JSON.stringify({
|
||||
ok: true,
|
||||
...module,
|
||||
...(replyContract
|
||||
? { reply_contract: replyContractSummary(replyContract) }
|
||||
: module.module_id === "opening-setup" ||
|
||||
module.module_id === "zero-layer-opening-setup"
|
||||
? {
|
||||
reply_contract_missing:
|
||||
"还没有可执行的正文组成数据契约。先写正文组成，再写开场。",
|
||||
}
|
||||
: {}),
|
||||
}),
|
||||
};
|
||||
}
|
||||
if (call.name === "read_library_entry") {
|
||||
@@ -385,6 +474,18 @@ export async function runDictateTurn(params: {
|
||||
const order = parseDictateOrder(args.order);
|
||||
const productsBeforeWrite = params.handlers.listProducts();
|
||||
const isOpening = isDictateOpeningTag(pieceTag);
|
||||
const contractMiss = openingContractMiss(
|
||||
pieceTag,
|
||||
piece.content,
|
||||
params.handlers.listProducts(),
|
||||
);
|
||||
if (contractMiss) {
|
||||
params.handlers.onToolCall?.(call.name, contractMiss.error);
|
||||
return {
|
||||
callId: call.id,
|
||||
content: JSON.stringify(contractMiss),
|
||||
};
|
||||
}
|
||||
const explicitlyPrepared = openingPrepared;
|
||||
if (isOpening && !explicitlyPrepared) {
|
||||
params.handlers.prepareOpening?.();
|
||||
@@ -464,6 +565,18 @@ export async function runDictateTurn(params: {
|
||||
const order = parseDictateOrder(args.order);
|
||||
const productsBeforeWrite = params.handlers.listProducts();
|
||||
const isOpening = isDictateOpeningTag(tag);
|
||||
const contractMiss = openingContractMiss(
|
||||
tag,
|
||||
content,
|
||||
params.handlers.listProducts(),
|
||||
);
|
||||
if (contractMiss) {
|
||||
params.handlers.onToolCall?.(call.name, contractMiss.error);
|
||||
return {
|
||||
callId: call.id,
|
||||
content: JSON.stringify(contractMiss),
|
||||
};
|
||||
}
|
||||
const explicitlyPrepared = openingPrepared;
|
||||
if (isOpening && !explicitlyPrepared) {
|
||||
params.handlers.prepareOpening?.();
|
||||
@@ -480,6 +593,10 @@ export async function runDictateTurn(params: {
|
||||
: {}),
|
||||
...(selfScore.value ? { selfScore: selfScore.value } : {}),
|
||||
});
|
||||
if (tag === VARIABLE_DESIGN_TAG || tag.startsWith(`${VARIABLE_DESIGN_TAG}#`)) {
|
||||
const synced = syncDesignProductToTable(params.handlers, content);
|
||||
wroteTags.push(...synced);
|
||||
}
|
||||
if (!isOpening) openingPrepared = false;
|
||||
wroteTags.push(tag);
|
||||
const detail =
|
||||
@@ -557,6 +674,8 @@ export async function runDictateTurn(params: {
|
||||
serializeVariableCatalog(doc),
|
||||
);
|
||||
wroteTags.push(VARIABLE_CATALOG_TAG);
|
||||
const seeded = seedCurrentFromCatalog(params.handlers);
|
||||
wroteTags.push(...seeded);
|
||||
const detail = `${input.key} visible=${doc.fields.find((f) => f.key === input.key)?.user_visible !== false}`;
|
||||
params.handlers.onToolCall?.(call.name, detail);
|
||||
return {
|
||||
@@ -826,6 +945,30 @@ function safeParseArgs(raw: string): Record<string, unknown> {
|
||||
}
|
||||
}
|
||||
|
||||
function contractForOpeningTag(
|
||||
tag: string,
|
||||
products: DictateProduct[],
|
||||
): ReplyDataContract | undefined {
|
||||
const want = tag.includes("0层") ? "设计.0层正文组成" : "设计.正文组成";
|
||||
const raw =
|
||||
products.find((product) => product.tag === want)?.content ??
|
||||
products.find((product) => product.tag === "设计.回复格式")?.content;
|
||||
return parseReplyDataContract(raw);
|
||||
}
|
||||
|
||||
function openingContractMiss(
|
||||
tag: string,
|
||||
content: string,
|
||||
products: DictateProduct[],
|
||||
): { error: string; reply_contract?: ReturnType<typeof replyContractSummary> } | undefined {
|
||||
if (!isDictateOpeningTag(tag)) return undefined;
|
||||
const contract = contractForOpeningTag(tag, products);
|
||||
if (!contract) return undefined;
|
||||
const error = openingContractError(content, contract);
|
||||
if (!error) return undefined;
|
||||
return { error, reply_contract: replyContractSummary(contract) };
|
||||
}
|
||||
|
||||
function requiresModuleRead(moduleId: string, auto?: boolean): boolean {
|
||||
if (auto) return false;
|
||||
return !new Set([
|
||||
|
||||
@@ -0,0 +1,173 @@
|
||||
import type { ChatMessage, LlmProvider } from "../llm/client.js";
|
||||
import { tryParseJsonDoc } from "../parse/json-doc.js";
|
||||
import {
|
||||
intakeSystemPrompt,
|
||||
nextWaveSystemPrompt,
|
||||
outlineSystemPrompt,
|
||||
redirectSystemPrompt,
|
||||
sectionSystemPrompt,
|
||||
} from "./prompt.js";
|
||||
import type { ListenSerialState } from "./types.js";
|
||||
|
||||
export type IntakeLlmResult =
|
||||
| { type: "ask"; reply: string }
|
||||
| {
|
||||
type: "propose";
|
||||
reply: string;
|
||||
thrill: string;
|
||||
world: string;
|
||||
structure: string;
|
||||
};
|
||||
|
||||
export type OutlineLlmResult = {
|
||||
titles: string[];
|
||||
payoffTitle: string;
|
||||
reply: string;
|
||||
thrill?: string;
|
||||
};
|
||||
|
||||
export type SectionLlmResult = {
|
||||
text: string;
|
||||
summary: string;
|
||||
fulfilled: boolean;
|
||||
};
|
||||
|
||||
export type RedirectLlmResult = {
|
||||
titles: string[];
|
||||
payoffTitle: string;
|
||||
thrillOverride: string;
|
||||
reply: string;
|
||||
};
|
||||
|
||||
async function completeJson(
|
||||
llm: LlmProvider,
|
||||
system: string,
|
||||
user: string,
|
||||
caller: string,
|
||||
): Promise<Record<string, unknown>> {
|
||||
const messages: ChatMessage[] = [
|
||||
{ role: "system", content: system },
|
||||
{ role: "user", content: user },
|
||||
];
|
||||
const result = await llm.complete(messages, {
|
||||
responseFormat: "json_object",
|
||||
caller,
|
||||
label: "听书连载",
|
||||
});
|
||||
const parsed = tryParseJsonDoc(result.content);
|
||||
if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) {
|
||||
throw new Error("听书连载：模型未返回合法 JSON 对象");
|
||||
}
|
||||
return parsed as Record<string, unknown>;
|
||||
}
|
||||
|
||||
function str(v: unknown): string {
|
||||
return typeof v === "string" ? v.trim() : v == null ? "" : String(v).trim();
|
||||
}
|
||||
|
||||
function strList(v: unknown): string[] {
|
||||
if (!Array.isArray(v)) return [];
|
||||
return v.map((x) => str(x)).filter(Boolean);
|
||||
}
|
||||
|
||||
export async function llmIntake(
|
||||
llm: LlmProvider,
|
||||
dialogue: Array<{ role: "user" | "assistant"; text: string }>,
|
||||
): Promise<IntakeLlmResult> {
|
||||
const transcript = dialogue
|
||||
.map((t) => `${t.role === "user" ? "用户" : "助手"}：${t.text}`)
|
||||
.join("\n\n");
|
||||
const obj = await completeJson(
|
||||
llm,
|
||||
intakeSystemPrompt(),
|
||||
transcript || "（用户尚未说话）",
|
||||
"listen_serial_intake",
|
||||
);
|
||||
const type = str(obj.type) === "propose" ? "propose" : "ask";
|
||||
const reply = str(obj.reply) || (type === "ask" ? "还需要一点材料。" : "请确认本局草案。");
|
||||
if (type === "ask") return { type, reply };
|
||||
return {
|
||||
type: "propose",
|
||||
reply,
|
||||
thrill: str(obj.thrill),
|
||||
world: str(obj.world),
|
||||
structure: str(obj.structure),
|
||||
};
|
||||
}
|
||||
|
||||
export async function llmOutline(
|
||||
llm: LlmProvider,
|
||||
state: ListenSerialState,
|
||||
userNote?: string,
|
||||
): Promise<OutlineLlmResult> {
|
||||
const obj = await completeJson(
|
||||
llm,
|
||||
outlineSystemPrompt(state),
|
||||
userNote?.trim() || "请生成本段大纲。",
|
||||
"listen_serial_outline",
|
||||
);
|
||||
const titles = strList(obj.titles);
|
||||
const payoffTitle = str(obj.payoff_title) || titles[titles.length - 1] || "";
|
||||
return {
|
||||
titles,
|
||||
payoffTitle,
|
||||
reply: str(obj.reply) || "大纲已就绪，开始写正文。",
|
||||
};
|
||||
}
|
||||
|
||||
export async function llmNextWave(
|
||||
llm: LlmProvider,
|
||||
state: ListenSerialState,
|
||||
): Promise<OutlineLlmResult> {
|
||||
const obj = await completeJson(
|
||||
llm,
|
||||
nextWaveSystemPrompt(state),
|
||||
"请开下一波爽点大纲。",
|
||||
"listen_serial_next_wave",
|
||||
);
|
||||
const titles = strList(obj.titles);
|
||||
return {
|
||||
titles,
|
||||
payoffTitle: str(obj.payoff_title) || titles[titles.length - 1] || "",
|
||||
reply: str(obj.reply) || "下一波大纲已就绪。",
|
||||
thrill: str(obj.thrill) || undefined,
|
||||
};
|
||||
}
|
||||
|
||||
export async function llmSection(
|
||||
llm: LlmProvider,
|
||||
state: ListenSerialState,
|
||||
nextTitle: string,
|
||||
): Promise<SectionLlmResult> {
|
||||
const obj = await completeJson(
|
||||
llm,
|
||||
sectionSystemPrompt(state, nextTitle),
|
||||
`请写标题「${nextTitle}」的正文。`,
|
||||
"listen_serial_section",
|
||||
);
|
||||
return {
|
||||
text: str(obj.text),
|
||||
summary: str(obj.summary),
|
||||
fulfilled: Boolean(obj.fulfilled),
|
||||
};
|
||||
}
|
||||
|
||||
export async function llmRedirect(
|
||||
llm: LlmProvider,
|
||||
state: ListenSerialState,
|
||||
opinion: string,
|
||||
): Promise<RedirectLlmResult> {
|
||||
const obj = await completeJson(
|
||||
llm,
|
||||
redirectSystemPrompt(state, opinion),
|
||||
opinion,
|
||||
"listen_serial_redirect",
|
||||
);
|
||||
const titles = strList(obj.titles);
|
||||
return {
|
||||
titles,
|
||||
payoffTitle: str(obj.payoff_title) || titles[titles.length - 1] || "",
|
||||
thrillOverride: str(obj.thrill_override),
|
||||
reply: str(obj.reply) || "已按你的意见改后面的走向。",
|
||||
};
|
||||
}
|
||||
@@ -0,0 +1,41 @@
|
||||
export {
|
||||
DEFAULT_BUFFER_TARGET_CHARS,
|
||||
LISTEN_SERIAL_MODE_VALUE,
|
||||
LISTEN_PHASE_TAG,
|
||||
LISTEN_THRILL_TAG,
|
||||
LISTEN_WORLD_TAG,
|
||||
LISTEN_STRUCTURE_TAG,
|
||||
LISTEN_OUTLINE_TAG,
|
||||
LISTEN_SECTIONS_TAG,
|
||||
LISTEN_SUMMARIES_TAG,
|
||||
isListenSerialModeValue,
|
||||
type ListenPhase,
|
||||
type ListenSerialState,
|
||||
type ListenSerialView,
|
||||
type OutlineItem,
|
||||
type SectionItem,
|
||||
} from "./types.js";
|
||||
|
||||
export {
|
||||
readListenState,
|
||||
toListenView,
|
||||
unreadChars,
|
||||
ensureBufferTarget,
|
||||
isConfirmText,
|
||||
isNextWaveText,
|
||||
isEndText,
|
||||
makeOutlineItems,
|
||||
applyRedirectCut,
|
||||
writeListenBrief,
|
||||
writeListenOutline,
|
||||
writeListenPhase,
|
||||
writeListenSections,
|
||||
writeListenSummaries,
|
||||
} from "./state.js";
|
||||
|
||||
export {
|
||||
runListenSerialTurn,
|
||||
markListenConsumed,
|
||||
type ListenChatTurn,
|
||||
type ListenSerialTurnResult,
|
||||
} from "./turn.js";
|
||||
@@ -0,0 +1,139 @@
|
||||
import type { ListenSerialState } from "./types.js";
|
||||
import { MAX_SUMMARIES_IN_CONTEXT } from "./types.js";
|
||||
|
||||
const ROLE = `你是「听书连载」写手：快速产出打发时间的廉价爽文（短剧/短篇量级）。
|
||||
文笔可以平、设定可以薄、前后可以略矛盾；但核心爽点必须兑现，节奏要直接。
|
||||
不要写成长篇网文铺垫；允许跳过前史，直接切到打脸/兑现段。
|
||||
输出必须是单个 JSON 对象，不要 markdown 围栏。`;
|
||||
|
||||
export function intakeSystemPrompt(): string {
|
||||
return `${ROLE}
|
||||
|
||||
当前阶段：开书收集。根据用户几句话，要么轻回问（最多 1～2 个关键缺口），要么给出可确认的本局草案。
|
||||
|
||||
JSON schema:
|
||||
{
|
||||
"type": "ask" | "propose",
|
||||
"reply": "给用户看的短回复（中文）",
|
||||
"thrill": "核心爽点一句话（propose 必填）",
|
||||
"world": "一句世界：基础世界+变造（propose 必填）",
|
||||
"structure": "结构调味，如起承转合/三段式+本局怎么用（propose 必填）"
|
||||
}
|
||||
|
||||
规则：
|
||||
- type=ask 时不要编造 thrill/world/structure。
|
||||
- type=propose 时三项都要具体、短；爽点要一眼能懂（身份反差/法律制裁/天道报应等）。
|
||||
- 用户材料已够就直接 propose，别为问而问。`;
|
||||
}
|
||||
|
||||
export function outlineSystemPrompt(state: ListenSerialState): string {
|
||||
return `${ROLE}
|
||||
|
||||
当前阶段：为本段生成大纲标题链，规划到「爽点兑现」为止（不要一次规划整本人生）。
|
||||
|
||||
已钉死：
|
||||
- 核心爽点：${state.thrill}
|
||||
- 一句世界：${state.world}
|
||||
- 结构：${state.structure}
|
||||
|
||||
JSON schema:
|
||||
{
|
||||
"titles": ["标题1", "标题2", "..."],
|
||||
"payoff_title": "其中作为本段兑现的那条标题（通常是最后一条）",
|
||||
"reply": "给用户看的一句说明"
|
||||
}
|
||||
|
||||
规则：
|
||||
- titles 6～12 条为宜，每条是短标题，不是正文。
|
||||
- 允许跳过最低境界等前史，直接切到兑现相关桥段。
|
||||
- payoff_title 必须是 titles 中的一项。`;
|
||||
}
|
||||
|
||||
export function sectionSystemPrompt(state: ListenSerialState, nextTitle: string): string {
|
||||
const summaries = state.summaries.slice(-MAX_SUMMARIES_IN_CONTEXT);
|
||||
return `${ROLE}
|
||||
|
||||
当前阶段：写下一节正文，并附小总结。
|
||||
|
||||
已钉死：
|
||||
- 核心爽点：${state.thrill}
|
||||
- 一句世界：${state.world}
|
||||
- 结构：${state.structure}
|
||||
|
||||
近期小总结（仅此记忆，勿要求全文一致）：
|
||||
${summaries.length ? summaries.map((s, i) => `${i + 1}. ${s}`).join("\n") : "（尚无）"}
|
||||
|
||||
本节大纲标题：${nextTitle}
|
||||
本节是否兑现节点：${
|
||||
state.outline.find((o) => o.title === nextTitle && o.status === "pending")
|
||||
?.payoff
|
||||
? "是——写到爽点落地"
|
||||
: "否——推进并留下钩子"
|
||||
}
|
||||
|
||||
JSON schema:
|
||||
{
|
||||
"text": "本节正文（约 600～1200 字，中文）",
|
||||
"summary": "本节小总结（一两句，供后文上下文）",
|
||||
"fulfilled": true/false
|
||||
}
|
||||
|
||||
规则：
|
||||
- fulfilled=true 仅当本节已兑现核心爽点（通常是兑现标题那节）。
|
||||
- 文笔廉价直白即可；不要长篇内心独白。`;
|
||||
}
|
||||
|
||||
export function redirectSystemPrompt(
|
||||
state: ListenSerialState,
|
||||
opinion: string,
|
||||
): string {
|
||||
const pending = state.outline
|
||||
.filter((o) => o.status === "pending")
|
||||
.map((o) => o.title);
|
||||
const done = state.outline
|
||||
.filter((o) => o.status === "done")
|
||||
.map((o) => o.title);
|
||||
return `${ROLE}
|
||||
|
||||
当前阶段：用户插嘴改向。只重排「尚未消费」的后续大纲；已听过的情节可不动，允许前后矛盾。
|
||||
|
||||
已钉死爽点（尽量仍朝它收，除非用户意见明确改爽点）：${state.thrill}
|
||||
一句世界：${state.world}
|
||||
用户意见：${opinion}
|
||||
|
||||
已完成标题：${done.length ? done.join(" / ") : "（无）"}
|
||||
原未写标题（将被替换）：${pending.length ? pending.join(" / ") : "（无）"}
|
||||
|
||||
JSON schema:
|
||||
{
|
||||
"titles": ["新的后续标题1", "..."],
|
||||
"payoff_title": "新的兑现标题",
|
||||
"thrill_override": "若用户改了爽点则给新爽点，否则空字符串",
|
||||
"reply": "给用户看的短确认"
|
||||
}
|
||||
|
||||
规则：
|
||||
- titles 规划到新的兑现为止；短、直接、服务用户意见。
|
||||
- 不要复述已完成标题。`;
|
||||
}
|
||||
|
||||
export function nextWaveSystemPrompt(state: ListenSerialState): string {
|
||||
const recent = state.summaries.slice(-MAX_SUMMARIES_IN_CONTEXT);
|
||||
return `${ROLE}
|
||||
|
||||
上一波爽点已兑现，用户要「下一波」。在同一句世界下开新的兑现段大纲。
|
||||
|
||||
核心爽点（可沿用或小幅升级）：${state.thrill}
|
||||
一句世界：${state.world}
|
||||
结构：${state.structure}
|
||||
近期小总结：
|
||||
${recent.length ? recent.map((s, i) => `${i + 1}. ${s}`).join("\n") : "（无）"}
|
||||
|
||||
JSON schema:
|
||||
{
|
||||
"titles": ["标题1", "..."],
|
||||
"payoff_title": "本波兑现标题",
|
||||
"thrill": "本波核心爽点（可与上波相同或升级）",
|
||||
"reply": "给用户看的一句说明"
|
||||
}`;
|
||||
}
|
||||
@@ -0,0 +1,251 @@
|
||||
import { randomUUID } from "node:crypto";
|
||||
import type { Blackboard } from "../blackboard/blackboard.js";
|
||||
import {
|
||||
DEFAULT_BUFFER_TARGET_CHARS,
|
||||
LISTEN_BUFFER_TARGET_TAG,
|
||||
LISTEN_CONSUMED_CHARS_TAG,
|
||||
LISTEN_OUTLINE_TAG,
|
||||
LISTEN_PHASE_TAG,
|
||||
LISTEN_SECTIONS_TAG,
|
||||
LISTEN_STRUCTURE_TAG,
|
||||
LISTEN_SUMMARIES_TAG,
|
||||
LISTEN_THRILL_TAG,
|
||||
LISTEN_WORLD_TAG,
|
||||
MAX_SUMMARIES_IN_CONTEXT,
|
||||
type ListenPhase,
|
||||
type ListenSerialState,
|
||||
type ListenSerialView,
|
||||
type OutlineItem,
|
||||
type SectionItem,
|
||||
} from "./types.js";
|
||||
|
||||
function parseJson<T>(raw: string | undefined, fallback: T): T {
|
||||
if (!raw?.trim()) return fallback;
|
||||
try {
|
||||
return JSON.parse(raw) as T;
|
||||
} catch {
|
||||
return fallback;
|
||||
}
|
||||
}
|
||||
|
||||
function parsePhase(raw: string | undefined): ListenPhase {
|
||||
const v = (raw ?? "").trim();
|
||||
if (
|
||||
v === "intake" ||
|
||||
v === "awaiting_confirm" ||
|
||||
v === "running" ||
|
||||
v === "soft_stop"
|
||||
) {
|
||||
return v;
|
||||
}
|
||||
return "intake";
|
||||
}
|
||||
|
||||
export function readListenState(board: Blackboard): ListenSerialState {
|
||||
const bufferRaw = board.getContentByTag(LISTEN_BUFFER_TARGET_TAG);
|
||||
const bufferTargetChars = Math.max(
|
||||
500,
|
||||
Number.parseInt(bufferRaw ?? "", 10) || DEFAULT_BUFFER_TARGET_CHARS,
|
||||
);
|
||||
const consumedRaw = board.getContentByTag(LISTEN_CONSUMED_CHARS_TAG);
|
||||
const consumedChars = Math.max(0, Number.parseInt(consumedRaw ?? "", 10) || 0);
|
||||
return {
|
||||
phase: parsePhase(board.getContentByTag(LISTEN_PHASE_TAG)),
|
||||
thrill: (board.getContentByTag(LISTEN_THRILL_TAG) ?? "").trim(),
|
||||
world: (board.getContentByTag(LISTEN_WORLD_TAG) ?? "").trim(),
|
||||
structure: (board.getContentByTag(LISTEN_STRUCTURE_TAG) ?? "").trim(),
|
||||
outline: parseJson<OutlineItem[]>(board.getContentByTag(LISTEN_OUTLINE_TAG), []),
|
||||
sections: parseJson<SectionItem[]>(
|
||||
board.getContentByTag(LISTEN_SECTIONS_TAG),
|
||||
[],
|
||||
),
|
||||
summaries: parseJson<string[]>(
|
||||
board.getContentByTag(LISTEN_SUMMARIES_TAG),
|
||||
[],
|
||||
),
|
||||
bufferTargetChars,
|
||||
consumedChars,
|
||||
};
|
||||
}
|
||||
|
||||
function writeTag(board: Blackboard, tag: string, content: string): void {
|
||||
board.write({ tag, content, source: "agent" });
|
||||
}
|
||||
|
||||
export function writeListenPhase(board: Blackboard, phase: ListenPhase): void {
|
||||
writeTag(board, LISTEN_PHASE_TAG, phase);
|
||||
}
|
||||
|
||||
export function writeListenBrief(
|
||||
board: Blackboard,
|
||||
brief: { thrill: string; world: string; structure: string },
|
||||
): void {
|
||||
writeTag(board, LISTEN_THRILL_TAG, brief.thrill.trim());
|
||||
writeTag(board, LISTEN_WORLD_TAG, brief.world.trim());
|
||||
writeTag(board, LISTEN_STRUCTURE_TAG, brief.structure.trim());
|
||||
}
|
||||
|
||||
export function writeListenOutline(board: Blackboard, outline: OutlineItem[]): void {
|
||||
writeTag(board, LISTEN_OUTLINE_TAG, JSON.stringify(outline));
|
||||
}
|
||||
|
||||
export function writeListenSections(board: Blackboard, sections: SectionItem[]): void {
|
||||
writeTag(board, LISTEN_SECTIONS_TAG, JSON.stringify(sections));
|
||||
}
|
||||
|
||||
export function writeListenSummaries(board: Blackboard, summaries: string[]): void {
|
||||
const trimmed = summaries
|
||||
.map((s) => s.trim())
|
||||
.filter(Boolean)
|
||||
.slice(-MAX_SUMMARIES_IN_CONTEXT);
|
||||
writeTag(board, LISTEN_SUMMARIES_TAG, JSON.stringify(trimmed));
|
||||
}
|
||||
|
||||
export function writeConsumedChars(board: Blackboard, n: number): void {
|
||||
writeTag(board, LISTEN_CONSUMED_CHARS_TAG, String(Math.max(0, Math.floor(n))));
|
||||
}
|
||||
|
||||
export function ensureBufferTarget(board: Blackboard): number {
|
||||
const existing = board.getContentByTag(LISTEN_BUFFER_TARGET_TAG);
|
||||
if (!existing?.trim()) {
|
||||
writeTag(board, LISTEN_BUFFER_TARGET_TAG, String(DEFAULT_BUFFER_TARGET_CHARS));
|
||||
return DEFAULT_BUFFER_TARGET_CHARS;
|
||||
}
|
||||
return Math.max(
|
||||
500,
|
||||
Number.parseInt(existing, 10) || DEFAULT_BUFFER_TARGET_CHARS,
|
||||
);
|
||||
}
|
||||
|
||||
export function unreadChars(state: ListenSerialState): number {
|
||||
const total = state.sections.reduce((n, s) => n + s.text.length, 0);
|
||||
return Math.max(0, total - state.consumedChars);
|
||||
}
|
||||
|
||||
export function toListenView(state: ListenSerialState): ListenSerialView {
|
||||
return {
|
||||
phase: state.phase,
|
||||
thrill: state.thrill,
|
||||
world: state.world,
|
||||
structure: state.structure,
|
||||
outline: state.outline.map((o) => ({
|
||||
id: o.id,
|
||||
title: o.title,
|
||||
status: o.status,
|
||||
payoff: o.payoff,
|
||||
})),
|
||||
sections: state.sections.map((s) => ({
|
||||
id: s.id,
|
||||
outlineId: s.outlineId,
|
||||
title: s.title,
|
||||
text: s.text,
|
||||
summary: s.summary,
|
||||
consumedChars: s.consumedChars,
|
||||
})),
|
||||
unreadChars: unreadChars(state),
|
||||
bufferTargetChars: state.bufferTargetChars,
|
||||
consumedChars: state.consumedChars,
|
||||
};
|
||||
}
|
||||
|
||||
/** 未写完的大纲项 */
|
||||
export function nextPendingOutline(state: ListenSerialState): OutlineItem | undefined {
|
||||
return state.outline.find((o) => o.status === "pending");
|
||||
}
|
||||
|
||||
export function makeOutlineItems(
|
||||
titles: string[],
|
||||
payoffTitle?: string,
|
||||
): OutlineItem[] {
|
||||
const payoff = (payoffTitle ?? titles[titles.length - 1] ?? "").trim();
|
||||
return titles
|
||||
.map((t) => t.trim())
|
||||
.filter(Boolean)
|
||||
.map((title) => ({
|
||||
id: randomUUID(),
|
||||
title,
|
||||
status: "pending" as const,
|
||||
payoff: payoff ? title === payoff : false,
|
||||
}));
|
||||
}
|
||||
|
||||
/** 插嘴：丢掉未消费正文尾部，并把对应大纲改回 pending / 用新标题替换未完成段 */
|
||||
export function applyRedirectCut(
|
||||
state: ListenSerialState,
|
||||
newPendingTitles: string[],
|
||||
): Pick<ListenSerialState, "outline" | "sections" | "summaries" | "consumedChars"> {
|
||||
let keptChars = 0;
|
||||
const keptSections: SectionItem[] = [];
|
||||
for (const sec of state.sections) {
|
||||
const remaining = sec.text.length - sec.consumedChars;
|
||||
if (sec.consumedChars <= 0 && remaining === sec.text.length) {
|
||||
// 整节未消费：丢弃
|
||||
continue;
|
||||
}
|
||||
if (sec.consumedChars >= sec.text.length) {
|
||||
keptSections.push(sec);
|
||||
keptChars += sec.text.length;
|
||||
continue;
|
||||
}
|
||||
// 部分消费：截断到已消费
|
||||
const cutText = sec.text.slice(0, sec.consumedChars);
|
||||
keptSections.push({
|
||||
...sec,
|
||||
text: cutText,
|
||||
summary: sec.summary,
|
||||
consumedChars: cutText.length,
|
||||
});
|
||||
keptChars += cutText.length;
|
||||
}
|
||||
|
||||
const doneOutlineIds = new Set(
|
||||
keptSections.map((s) => s.outlineId).filter(Boolean),
|
||||
);
|
||||
const keptOutline = state.outline
|
||||
.filter((o) => o.status === "done" && doneOutlineIds.has(o.id))
|
||||
.map((o) => ({ ...o, status: "done" as const }));
|
||||
|
||||
const newItems = makeOutlineItems(newPendingTitles);
|
||||
// 最后一节标为兑现
|
||||
if (newItems.length) {
|
||||
for (const item of newItems) item.payoff = false;
|
||||
newItems[newItems.length - 1]!.payoff = true;
|
||||
}
|
||||
|
||||
const summaries = keptSections
|
||||
.map((s) => s.summary)
|
||||
.filter(Boolean)
|
||||
.slice(-MAX_SUMMARIES_IN_CONTEXT);
|
||||
|
||||
return {
|
||||
outline: [...keptOutline, ...newItems],
|
||||
sections: keptSections,
|
||||
summaries,
|
||||
consumedChars: keptChars,
|
||||
};
|
||||
}
|
||||
|
||||
export function markOutlineDone(
|
||||
outline: OutlineItem[],
|
||||
outlineId: string,
|
||||
): OutlineItem[] {
|
||||
return outline.map((o) =>
|
||||
o.id === outlineId ? { ...o, status: "done" as const } : o,
|
||||
);
|
||||
}
|
||||
|
||||
export function isConfirmText(text: string): boolean {
|
||||
const t = text.trim();
|
||||
if (!t) return false;
|
||||
return /^(确认|就这样|开始|可以|好的|没问题|开写|写吧)([！!。．\s]*)$/u.test(t);
|
||||
}
|
||||
|
||||
export function isNextWaveText(text: string): boolean {
|
||||
const t = text.trim();
|
||||
return /下一波|再来一波|继续下一|开下一波/u.test(t);
|
||||
}
|
||||
|
||||
export function isEndText(text: string): boolean {
|
||||
const t = text.trim();
|
||||
return /^(收工|结束|完结|停)([！!。．\s]*)$/u.test(t);
|
||||
}
|
||||
@@ -0,0 +1,343 @@
|
||||
import { randomUUID } from "node:crypto";
|
||||
import type { Blackboard } from "../blackboard/blackboard.js";
|
||||
import type { LlmProvider } from "../llm/client.js";
|
||||
import {
|
||||
llmIntake,
|
||||
llmNextWave,
|
||||
llmOutline,
|
||||
llmRedirect,
|
||||
llmSection,
|
||||
} from "./generate.js";
|
||||
import {
|
||||
applyRedirectCut,
|
||||
ensureBufferTarget,
|
||||
isConfirmText,
|
||||
isEndText,
|
||||
isNextWaveText,
|
||||
makeOutlineItems,
|
||||
markOutlineDone,
|
||||
nextPendingOutline,
|
||||
readListenState,
|
||||
unreadChars,
|
||||
writeConsumedChars,
|
||||
writeListenBrief,
|
||||
writeListenOutline,
|
||||
writeListenPhase,
|
||||
writeListenSections,
|
||||
writeListenSummaries,
|
||||
} from "./state.js";
|
||||
import { MAX_SECTIONS_PER_BURST, type ListenSerialState } from "./types.js";
|
||||
|
||||
export type ListenChatTurn = { role: "user" | "assistant"; text: string };
|
||||
|
||||
export type ListenSerialTurnResult = {
|
||||
reply: string;
|
||||
/** 本轮是否进入/保持软停 */
|
||||
softStop?: boolean;
|
||||
};
|
||||
|
||||
function persistState(
|
||||
board: Blackboard,
|
||||
state: ListenSerialState,
|
||||
patch: Partial<ListenSerialState>,
|
||||
): ListenSerialState {
|
||||
const next: ListenSerialState = { ...state, ...patch };
|
||||
if (patch.phase) writeListenPhase(board, patch.phase);
|
||||
if (patch.thrill != null || patch.world != null || patch.structure != null) {
|
||||
writeListenBrief(board, {
|
||||
thrill: next.thrill,
|
||||
world: next.world,
|
||||
structure: next.structure,
|
||||
});
|
||||
}
|
||||
if (patch.outline) writeListenOutline(board, next.outline);
|
||||
if (patch.sections) writeListenSections(board, next.sections);
|
||||
if (patch.summaries) writeListenSummaries(board, next.summaries);
|
||||
if (patch.consumedChars != null) writeConsumedChars(board, next.consumedChars);
|
||||
return next;
|
||||
}
|
||||
|
||||
async function fillBuffer(
|
||||
llm: LlmProvider,
|
||||
board: Blackboard,
|
||||
state: ListenSerialState,
|
||||
): Promise<{ state: ListenSerialState; notes: string[]; softStop: boolean }> {
|
||||
let cur = state;
|
||||
const notes: string[] = [];
|
||||
let softStop = false;
|
||||
let wrote = 0;
|
||||
|
||||
while (
|
||||
unreadChars(cur) < cur.bufferTargetChars &&
|
||||
wrote < MAX_SECTIONS_PER_BURST
|
||||
) {
|
||||
const pending = nextPendingOutline(cur);
|
||||
if (!pending) {
|
||||
softStop = true;
|
||||
break;
|
||||
}
|
||||
const section = await llmSection(llm, cur, pending.title);
|
||||
if (!section.text) {
|
||||
notes.push(`「${pending.title}」正文为空，已跳过。`);
|
||||
cur = persistState(board, cur, {
|
||||
outline: markOutlineDone(cur.outline, pending.id),
|
||||
});
|
||||
wrote += 1;
|
||||
continue;
|
||||
}
|
||||
const item = {
|
||||
id: randomUUID(),
|
||||
outlineId: pending.id,
|
||||
title: pending.title,
|
||||
text: section.text,
|
||||
summary: section.summary || pending.title,
|
||||
consumedChars: 0,
|
||||
};
|
||||
const summaries = [...cur.summaries, item.summary];
|
||||
let outline = markOutlineDone(cur.outline, pending.id);
|
||||
cur = persistState(board, cur, {
|
||||
outline,
|
||||
sections: [...cur.sections, item],
|
||||
summaries,
|
||||
phase: "running",
|
||||
});
|
||||
notes.push(`已写：${pending.title}（${section.text.length} 字）`);
|
||||
wrote += 1;
|
||||
|
||||
if (pending.payoff || section.fulfilled) {
|
||||
// 兑现后把剩余 pending 标 cut，进入软停（缓冲里已有兑现节可听完）
|
||||
outline = cur.outline.map((o) =>
|
||||
o.status === "pending" ? { ...o, status: "cut" as const } : o,
|
||||
);
|
||||
cur = persistState(board, cur, { outline, phase: "soft_stop" });
|
||||
softStop = true;
|
||||
notes.push("本段爽点已兑现，听完缓冲后可「下一波」或「收工」。");
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
if (!softStop && !nextPendingOutline(cur) && unreadChars(cur) > 0) {
|
||||
cur = persistState(board, cur, { phase: "soft_stop" });
|
||||
softStop = true;
|
||||
}
|
||||
|
||||
return { state: cur, notes, softStop };
|
||||
}
|
||||
|
||||
function formatUnreadLine(state: ListenSerialState): string {
|
||||
const n = unreadChars(state);
|
||||
return `目前约有 ${n} 字未读/未听（目标缓冲 ${state.bufferTargetChars} 字）。`;
|
||||
}
|
||||
|
||||
function formatOutlineLine(state: ListenSerialState): string {
|
||||
if (!state.outline.length) return "";
|
||||
return (
|
||||
"大纲：\n" +
|
||||
state.outline
|
||||
.map((o) => {
|
||||
const mark =
|
||||
o.status === "done" ? "✓" : o.status === "cut" ? "×" : "·";
|
||||
const pay = o.payoff ? "（兑现）" : "";
|
||||
return `${mark} ${o.title}${pay}`;
|
||||
})
|
||||
.join("\n")
|
||||
);
|
||||
}
|
||||
|
||||
/**
|
||||
* 听书连载单轮：开书 / 确认 / 补货 / 插嘴 / 软停续波。
|
||||
*/
|
||||
export async function runListenSerialTurn(params: {
|
||||
llm: LlmProvider;
|
||||
board: Blackboard;
|
||||
dialogue: ListenChatTurn[];
|
||||
userText: string;
|
||||
}): Promise<ListenSerialTurnResult> {
|
||||
const { llm, board, dialogue, userText } = params;
|
||||
ensureBufferTarget(board);
|
||||
let state = readListenState(board);
|
||||
const text = userText.trim();
|
||||
|
||||
// —— 软停：收工 / 下一波 ——
|
||||
if (state.phase === "soft_stop") {
|
||||
if (isEndText(text)) {
|
||||
return {
|
||||
reply: "本局收工。想再开可以新建作品，或再说「下一波」。",
|
||||
softStop: true,
|
||||
};
|
||||
}
|
||||
if (isNextWaveText(text) || /继续|再来/u.test(text)) {
|
||||
const wave = await llmNextWave(llm, state);
|
||||
if (wave.thrill) {
|
||||
state = persistState(board, state, { thrill: wave.thrill });
|
||||
}
|
||||
const outline = makeOutlineItems(wave.titles, wave.payoffTitle);
|
||||
state = persistState(board, state, {
|
||||
outline,
|
||||
phase: "running",
|
||||
});
|
||||
const filled = await fillBuffer(llm, board, state);
|
||||
state = filled.state;
|
||||
const parts = [
|
||||
wave.reply,
|
||||
formatOutlineLine(state),
|
||||
filled.notes.join("\n"),
|
||||
formatUnreadLine(state),
|
||||
].filter(Boolean);
|
||||
if (filled.softStop) {
|
||||
parts.push("本波已兑现。可输入「下一波」或「收工」。");
|
||||
}
|
||||
return { reply: parts.join("\n\n"), softStop: filled.softStop };
|
||||
}
|
||||
// 软停时其他话当作插嘴改向再开波
|
||||
}
|
||||
|
||||
// —— 等待确认 ——
|
||||
if (state.phase === "awaiting_confirm") {
|
||||
if (!isConfirmText(text) && text && !/确认|开始|开写/u.test(text)) {
|
||||
// 用户继续改需求 → 回到 intake 再提案
|
||||
const intake = await llmIntake(llm, dialogue);
|
||||
if (intake.type === "ask") {
|
||||
state = persistState(board, state, { phase: "intake" });
|
||||
return { reply: intake.reply };
|
||||
}
|
||||
state = persistState(board, state, {
|
||||
phase: "awaiting_confirm",
|
||||
thrill: intake.thrill,
|
||||
world: intake.world,
|
||||
structure: intake.structure,
|
||||
});
|
||||
return {
|
||||
reply: `${intake.reply}\n\n【核心爽点】${intake.thrill}\n【一句世界】${intake.world}\n【结构】${intake.structure}\n\n回复「确认」开始生成大纲与正文。`,
|
||||
};
|
||||
}
|
||||
|
||||
const outlined = await llmOutline(llm, state);
|
||||
const outline = makeOutlineItems(outlined.titles, outlined.payoffTitle);
|
||||
state = persistState(board, state, { outline, phase: "running" });
|
||||
const filled = await fillBuffer(llm, board, state);
|
||||
state = filled.state;
|
||||
return {
|
||||
reply: [
|
||||
outlined.reply,
|
||||
formatOutlineLine(state),
|
||||
filled.notes.join("\n"),
|
||||
formatUnreadLine(state),
|
||||
filled.softStop
|
||||
? "本段已兑现。可「下一波」或「收工」。"
|
||||
: "不满意某桥段就直接说一句意见，我会改后面的走向。",
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n\n"),
|
||||
softStop: filled.softStop,
|
||||
};
|
||||
}
|
||||
|
||||
// —— 连载中：空输入/继续 = 补货；有意见 = 改向 ——
|
||||
if (state.phase === "running" || state.phase === "soft_stop") {
|
||||
const isContinue =
|
||||
!text ||
|
||||
/^(继续|接着写|补货|下一节)([！!。．\s]*)$/u.test(text);
|
||||
|
||||
if (!isContinue && text) {
|
||||
const redirect = await llmRedirect(llm, state, text);
|
||||
const cut = applyRedirectCut(state, redirect.titles);
|
||||
// 重新标记 payoff
|
||||
const outline = cut.outline.map((o, i, arr) => ({
|
||||
...o,
|
||||
payoff:
|
||||
o.status === "pending" &&
|
||||
(o.title === redirect.payoffTitle || i === arr.length - 1),
|
||||
}));
|
||||
// 只让最后一个 pending 为 payoff
|
||||
let lastPending = -1;
|
||||
outline.forEach((o, i) => {
|
||||
if (o.status === "pending") lastPending = i;
|
||||
if (o.status === "pending") o.payoff = false;
|
||||
});
|
||||
if (lastPending >= 0) outline[lastPending]!.payoff = true;
|
||||
|
||||
state = persistState(board, state, {
|
||||
...cut,
|
||||
outline,
|
||||
phase: "running",
|
||||
...(redirect.thrillOverride
|
||||
? { thrill: redirect.thrillOverride }
|
||||
: {}),
|
||||
});
|
||||
const filled = await fillBuffer(llm, board, state);
|
||||
state = filled.state;
|
||||
return {
|
||||
reply: [
|
||||
redirect.reply,
|
||||
formatOutlineLine(state),
|
||||
filled.notes.join("\n"),
|
||||
formatUnreadLine(state),
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n\n"),
|
||||
softStop: filled.softStop,
|
||||
};
|
||||
}
|
||||
|
||||
const filled = await fillBuffer(llm, board, state);
|
||||
state = filled.state;
|
||||
if (!filled.notes.length && unreadChars(state) >= state.bufferTargetChars) {
|
||||
return {
|
||||
reply: `${formatUnreadLine(state)}\n缓冲仍充足。听/读一些后再说「继续」，或直接一句意见改向。`,
|
||||
softStop: state.phase === "soft_stop",
|
||||
};
|
||||
}
|
||||
return {
|
||||
reply: [
|
||||
filled.notes.join("\n") || "已尝试补货。",
|
||||
formatOutlineLine(state),
|
||||
formatUnreadLine(state),
|
||||
filled.softStop ? "本段已兑现。可「下一波」或「收工」。" : "",
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("\n\n"),
|
||||
softStop: filled.softStop,
|
||||
};
|
||||
}
|
||||
|
||||
// —— intake ——
|
||||
const intake = await llmIntake(llm, dialogue);
|
||||
if (intake.type === "ask") {
|
||||
state = persistState(board, state, { phase: "intake" });
|
||||
return { reply: intake.reply };
|
||||
}
|
||||
if (!intake.thrill || !intake.world) {
|
||||
return {
|
||||
reply: intake.reply || "还需要核心爽点和一句世界，请再补一句。",
|
||||
};
|
||||
}
|
||||
state = persistState(board, state, {
|
||||
phase: "awaiting_confirm",
|
||||
thrill: intake.thrill,
|
||||
world: intake.world,
|
||||
structure: intake.structure || "三段式：铺垫冲突 → 对抗升级 → 爽点兑现",
|
||||
});
|
||||
return {
|
||||
reply: `${intake.reply}\n\n【核心爽点】${intake.thrill}\n【一句世界】${intake.world}\n【结构】${state.structure}\n\n回复「确认」开始生成大纲与正文。`,
|
||||
};
|
||||
}
|
||||
|
||||
/** 前端标记已读：推进消费指针 */
|
||||
export function markListenConsumed(
|
||||
board: Blackboard,
|
||||
absoluteConsumedChars: number,
|
||||
): ListenSerialState {
|
||||
const state = readListenState(board);
|
||||
const total = state.sections.reduce((n, s) => n + s.text.length, 0);
|
||||
const next = Math.max(0, Math.min(total, Math.floor(absoluteConsumedChars)));
|
||||
let remaining = next;
|
||||
const sections = state.sections.map((sec) => {
|
||||
const take = Math.min(sec.text.length, Math.max(0, remaining));
|
||||
remaining -= take;
|
||||
return { ...sec, consumedChars: take };
|
||||
});
|
||||
writeListenSections(board, sections);
|
||||
writeConsumedChars(board, next);
|
||||
return { ...state, sections, consumedChars: next };
|
||||
}
|
||||
@@ -0,0 +1,81 @@
|
||||
/** 听书连载 · 类型与常量 */
|
||||
|
||||
export const LISTEN_SERIAL_MODE_VALUE = "listen_serial";
|
||||
|
||||
export const LISTEN_PHASE_TAG = "听书.阶段";
|
||||
export const LISTEN_THRILL_TAG = "听书.核心爽点";
|
||||
export const LISTEN_WORLD_TAG = "听书.一句世界";
|
||||
export const LISTEN_STRUCTURE_TAG = "听书.结构";
|
||||
export const LISTEN_OUTLINE_TAG = "听书.大纲";
|
||||
export const LISTEN_SECTIONS_TAG = "听书.正文节";
|
||||
export const LISTEN_SUMMARIES_TAG = "听书.小总结";
|
||||
export const LISTEN_BUFFER_TARGET_TAG = "听书.缓冲目标字数";
|
||||
export const LISTEN_CONSUMED_CHARS_TAG = "听书.已消费字数";
|
||||
|
||||
/** 默认未读/未听缓冲目标（字）——实现默认值，可后续调 */
|
||||
export const DEFAULT_BUFFER_TARGET_CHARS = 4000;
|
||||
/** 单次用户触发最多连写几节，避免一次请求过长 */
|
||||
export const MAX_SECTIONS_PER_BURST = 3;
|
||||
/** 进模型的滚动小总结条数上限 */
|
||||
export const MAX_SUMMARIES_IN_CONTEXT = 8;
|
||||
|
||||
export type ListenPhase =
|
||||
| "intake"
|
||||
| "awaiting_confirm"
|
||||
| "running"
|
||||
| "soft_stop";
|
||||
|
||||
export type OutlineItem = {
|
||||
id: string;
|
||||
title: string;
|
||||
/** pending=未写；done=已有正文；cut=被插嘴砍掉 */
|
||||
status: "pending" | "done" | "cut";
|
||||
/** 本段爽点兑现节点 */
|
||||
payoff?: boolean;
|
||||
};
|
||||
|
||||
export type SectionItem = {
|
||||
id: string;
|
||||
outlineId: string;
|
||||
title: string;
|
||||
text: string;
|
||||
summary: string;
|
||||
/** 用户已听/读过的字数（相对本节）；改向时整节未消费可丢 */
|
||||
consumedChars: number;
|
||||
};
|
||||
|
||||
export type ListenSerialState = {
|
||||
phase: ListenPhase;
|
||||
thrill: string;
|
||||
world: string;
|
||||
structure: string;
|
||||
outline: OutlineItem[];
|
||||
sections: SectionItem[];
|
||||
summaries: string[];
|
||||
bufferTargetChars: number;
|
||||
/** 全局已消费正文字数（用于 UI「已听」） */
|
||||
consumedChars: number;
|
||||
};
|
||||
|
||||
export type ListenSerialView = {
|
||||
phase: ListenPhase;
|
||||
thrill: string;
|
||||
world: string;
|
||||
structure: string;
|
||||
outline: Array<{ id: string; title: string; status: string; payoff?: boolean }>;
|
||||
sections: Array<{
|
||||
id: string;
|
||||
outlineId: string;
|
||||
title: string;
|
||||
text: string;
|
||||
summary: string;
|
||||
consumedChars: number;
|
||||
}>;
|
||||
unreadChars: number;
|
||||
bufferTargetChars: number;
|
||||
consumedChars: number;
|
||||
};
|
||||
|
||||
export function isListenSerialModeValue(raw: string | undefined | null): boolean {
|
||||
return (raw ?? "").trim() === LISTEN_SERIAL_MODE_VALUE;
|
||||
}
|
||||
+140
-21
@@ -57,8 +57,33 @@ export type TokenUsage = {
|
||||
cachedTokens?: number;
|
||||
/** Prompt tokens not served from cache (DeepSeek prompt_cache_miss_tokens) */
|
||||
cacheMissTokens?: number;
|
||||
/** 上游若单独返回缓存写入 token（如 cache_creation_input_tokens） */
|
||||
cacheWriteTokens?: number;
|
||||
};
|
||||
|
||||
/** 用量行上的调度信息。不包含提示词或产物正文。 */
|
||||
export type DictateUsageStep = "select" | "fill" | "fill-retry" | "reply" | "play";
|
||||
|
||||
export type UsageDetail = {
|
||||
turnId?: string;
|
||||
step?: DictateUsageStep;
|
||||
retryIndex?: number;
|
||||
segmentChars?: {
|
||||
rules: number;
|
||||
history: number;
|
||||
products: number;
|
||||
injected: number;
|
||||
tail: number;
|
||||
};
|
||||
plannedTags?: string[];
|
||||
succeededTags?: string[];
|
||||
failedTags?: string[];
|
||||
finish?: string;
|
||||
};
|
||||
|
||||
/** complete 返回后由统计层写入 recordId，供补记成功或失败的 tag。 */
|
||||
export type UsageSink = { recordId?: string };
|
||||
|
||||
export type CompleteResult = {
|
||||
content: string;
|
||||
/** 推理模型思维链（如 DeepSeek reasoner 的 reasoning_content） */
|
||||
@@ -84,6 +109,14 @@ export type CompleteOptions = {
|
||||
label?: string;
|
||||
/** 取消进行中的 LLM 请求（停止并重试） */
|
||||
signal?: AbortSignal;
|
||||
/**
|
||||
* 本局会话 id。仅 OpenRouter 会写成请求体 `session_id`，用来粘住同一上游，
|
||||
* 避免开头 message 一变就换缓存节点。其它渠道不发送。
|
||||
*/
|
||||
sessionId?: string;
|
||||
/** 统计补充。不进上游请求体。 */
|
||||
usageDetail?: UsageDetail;
|
||||
usageSink?: UsageSink;
|
||||
};
|
||||
|
||||
export type ToolChoice =
|
||||
@@ -132,6 +165,33 @@ export type LlmProvider = {
|
||||
};
|
||||
|
||||
const JSON_OBJECT_HINT = "请用 JSON 对象回复。";
|
||||
const OPENROUTER_SESSION_ID_MAX = 256;
|
||||
|
||||
/** OpenRouter 用 `session_id` 做粘滞路由；其它 OpenAI 兼容渠道不认识这个字段。 */
|
||||
export function isOpenRouterBaseUrl(baseUrl: string): boolean {
|
||||
const trimmed = baseUrl.trim();
|
||||
if (!trimmed) return false;
|
||||
try {
|
||||
const withProtocol = /^https?:\/\//i.test(trimmed)
|
||||
? trimmed
|
||||
: `https://${trimmed}`;
|
||||
const host = new URL(withProtocol).hostname.toLowerCase();
|
||||
return host === "openrouter.ai" || host.endsWith(".openrouter.ai");
|
||||
} catch {
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
/** 合法则返回要写入请求体的 session_id；非 OpenRouter 或空 id 则不写。 */
|
||||
export function openRouterSessionId(
|
||||
baseUrl: string,
|
||||
sessionId: string | undefined,
|
||||
): string | undefined {
|
||||
if (!isOpenRouterBaseUrl(baseUrl)) return undefined;
|
||||
const id = sessionId?.trim();
|
||||
if (!id) return undefined;
|
||||
return id.slice(0, OPENROUTER_SESSION_ID_MAX);
|
||||
}
|
||||
|
||||
function messagesMentionJson(messages: ChatMessage[]): boolean {
|
||||
for (const message of messages) {
|
||||
@@ -210,6 +270,9 @@ export function buildRequestBody(
|
||||
body.stream_options = { include_usage: true };
|
||||
}
|
||||
|
||||
const sessionId = openRouterSessionId(config.baseUrl, options?.sessionId);
|
||||
if (sessionId) body.session_id = sessionId;
|
||||
|
||||
return body;
|
||||
}
|
||||
|
||||
@@ -232,6 +295,20 @@ function readCacheMissTokens(u: Record<string, unknown>): number | undefined {
|
||||
return readFiniteNumber(u.prompt_cache_miss_tokens ?? u.promptCacheMissTokens);
|
||||
}
|
||||
|
||||
function readCacheWriteTokens(u: Record<string, unknown>): number | undefined {
|
||||
const details = u.prompt_tokens_details ?? u.promptTokensDetails;
|
||||
if (details && typeof details === "object") {
|
||||
const row = details as Record<string, unknown>;
|
||||
const nested = readFiniteNumber(row.cache_write_tokens ?? row.cacheWriteTokens);
|
||||
if (nested != null) return nested;
|
||||
}
|
||||
return readFiniteNumber(
|
||||
u.cache_creation_input_tokens ??
|
||||
u.cache_write_input_tokens ??
|
||||
u.cacheWriteTokens,
|
||||
);
|
||||
}
|
||||
|
||||
/** Parse OpenAI-compatible usage object, including provider-specific cache fields. */
|
||||
export function parseUsage(raw: unknown): TokenUsage | undefined {
|
||||
if (!raw || typeof raw !== "object") return undefined;
|
||||
@@ -243,6 +320,7 @@ export function parseUsage(raw: unknown): TokenUsage | undefined {
|
||||
|
||||
const cachedTokens = readCachedTokens(u);
|
||||
const cacheMissTokens = readCacheMissTokens(u);
|
||||
const cacheWriteTokens = readCacheWriteTokens(u);
|
||||
|
||||
return {
|
||||
promptTokens: Number.isFinite(prompt) ? prompt : 0,
|
||||
@@ -253,22 +331,66 @@ export function parseUsage(raw: unknown): TokenUsage | undefined {
|
||||
(Number.isFinite(completion) ? completion : 0),
|
||||
...(cachedTokens != null ? { cachedTokens } : {}),
|
||||
...(cacheMissTokens != null ? { cacheMissTokens } : {}),
|
||||
...(cacheWriteTokens != null ? { cacheWriteTokens } : {}),
|
||||
};
|
||||
}
|
||||
|
||||
function extractMessageParts(message: Record<string, unknown> | undefined): {
|
||||
function stringifyToolArguments(value: unknown): string {
|
||||
if (typeof value === "string" && value.trim()) return value;
|
||||
if (value && typeof value === "object") return JSON.stringify(value);
|
||||
return "{}";
|
||||
}
|
||||
|
||||
function toolCallFromRow(row: Record<string, unknown>, index: number): ParsedToolCall | null {
|
||||
const fn = row.function;
|
||||
const nested = fn && typeof fn === "object" ? (fn as Record<string, unknown>) : null;
|
||||
const source = nested ?? row;
|
||||
const name = typeof source.name === "string" ? source.name.trim() : "";
|
||||
if (!name) return null;
|
||||
const rawId = typeof row.id === "string" ? row.id.trim() : "";
|
||||
return {
|
||||
id: rawId || `call-${index}`,
|
||||
name,
|
||||
arguments: stringifyToolArguments(
|
||||
source.arguments ?? source.args ?? row.arguments ?? row.args ?? row.input,
|
||||
),
|
||||
};
|
||||
}
|
||||
|
||||
function embeddedContent(raw: unknown): { text: string | null; calls: ParsedToolCall[] } {
|
||||
if (typeof raw === "string") return { text: raw.trim() || null, calls: [] };
|
||||
if (!Array.isArray(raw)) return { text: raw == null ? null : "", calls: [] };
|
||||
const texts: string[] = [];
|
||||
const calls: ParsedToolCall[] = [];
|
||||
raw.forEach((part, index) => {
|
||||
if (!part || typeof part !== "object") return;
|
||||
const row = part as Record<string, unknown>;
|
||||
if (typeof row.text === "string" && row.text.trim()) texts.push(row.text.trim());
|
||||
const fn = row.functionCall ?? row.function_call;
|
||||
if (row.type === "tool_use" || (fn && typeof fn === "object")) {
|
||||
const nested = fn && typeof fn === "object" ? (fn as Record<string, unknown>) : row;
|
||||
const parsed = toolCallFromRow(
|
||||
{
|
||||
id: row.id,
|
||||
name: nested.name ?? row.name,
|
||||
arguments: nested.args ?? nested.arguments ?? row.input ?? row.arguments,
|
||||
},
|
||||
index,
|
||||
);
|
||||
if (parsed) calls.push(parsed);
|
||||
}
|
||||
});
|
||||
return { text: texts.join("\n").trim() || null, calls };
|
||||
}
|
||||
|
||||
/** 从一条 assistant message 取出正文和工具调用。对象参数、缺 id、content 片段里的调用都收。 */
|
||||
export function extractMessageParts(message: Record<string, unknown> | undefined): {
|
||||
content: string | null;
|
||||
reasoning?: string;
|
||||
toolCalls: ParsedToolCall[];
|
||||
} {
|
||||
if (!message) return { content: "", toolCalls: [] };
|
||||
const rawContent = message.content;
|
||||
const content =
|
||||
typeof rawContent === "string"
|
||||
? rawContent.trim() || null
|
||||
: rawContent == null
|
||||
? null
|
||||
: "";
|
||||
const embedded = embeddedContent(message.content);
|
||||
const reasoningRaw =
|
||||
(typeof message.reasoning_content === "string" &&
|
||||
message.reasoning_content.trim()) ||
|
||||
@@ -280,21 +402,18 @@ function extractMessageParts(message: Record<string, unknown> | undefined): {
|
||||
const toolCalls: ParsedToolCall[] = [];
|
||||
const rawCalls = message.tool_calls;
|
||||
if (Array.isArray(rawCalls)) {
|
||||
for (const call of rawCalls) {
|
||||
if (!call || typeof call !== "object") continue;
|
||||
const c = call as Record<string, unknown>;
|
||||
const fn = c.function;
|
||||
if (!fn || typeof fn !== "object") continue;
|
||||
const f = fn as Record<string, unknown>;
|
||||
const name = typeof f.name === "string" ? f.name : "";
|
||||
const id = typeof c.id === "string" ? c.id : "";
|
||||
const args =
|
||||
typeof f.arguments === "string" ? f.arguments : "{}";
|
||||
if (name && id) {
|
||||
toolCalls.push({ id, name, arguments: args });
|
||||
}
|
||||
rawCalls.forEach((call, index) => {
|
||||
if (!call || typeof call !== "object") return;
|
||||
const parsed = toolCallFromRow(call as Record<string, unknown>, index);
|
||||
if (parsed) toolCalls.push(parsed);
|
||||
});
|
||||
}
|
||||
for (const call of embedded.calls) {
|
||||
if (!toolCalls.some((item) => item.id === call.id && item.name === call.name)) {
|
||||
toolCalls.push(call);
|
||||
}
|
||||
}
|
||||
const content = embedded.text;
|
||||
|
||||
if (toolCalls.length > 0) {
|
||||
return { content, reasoning: reasoning || undefined, toolCalls };
|
||||
|
||||
@@ -0,0 +1,181 @@
|
||||
import { createHash } from "node:crypto";
|
||||
import type { ToolDefinition } from "./client.js";
|
||||
|
||||
/** 一段请求的位置标签。只存标签、字数和哈希，不存正文。 */
|
||||
export type ContextSegment = {
|
||||
tag: string;
|
||||
chars: number;
|
||||
hash: string;
|
||||
};
|
||||
|
||||
/** 挂在一条回复上的各次请求轮廓，选题和填写都保留。 */
|
||||
export type RequestOutline = {
|
||||
at: string;
|
||||
caller: string;
|
||||
step?: string;
|
||||
turnId?: string;
|
||||
model?: string;
|
||||
promptTokens?: number;
|
||||
completionTokens?: number;
|
||||
cachedTokens?: number;
|
||||
cacheMissTokens?: number;
|
||||
cacheWriteTokens?: number;
|
||||
segments: ContextSegment[];
|
||||
};
|
||||
|
||||
/** 和同会话上一条请求比，前缀从哪里开始对不上。 */
|
||||
export type CacheBreakInfo = {
|
||||
/** 本窗口里没有更早的同会话请求 */
|
||||
cold: boolean;
|
||||
/** -1：本条在上一请求的前缀里整段对上 */
|
||||
breakAt: number;
|
||||
tag?: string;
|
||||
/** 断点之前、和上一请求逐段相同的段数 */
|
||||
matched: number;
|
||||
};
|
||||
|
||||
const SPLIT =
|
||||
/(?:^|\n)(#{2,4} [^\n]+|最后一轮用户输入|【用户】|【助手】)/g;
|
||||
|
||||
function hashText(text: string): string {
|
||||
return createHash("sha256").update(text).digest("hex").slice(0, 12);
|
||||
}
|
||||
|
||||
function segment(tag: string, text: string): ContextSegment {
|
||||
return { tag, chars: text.length, hash: hashText(text) };
|
||||
}
|
||||
|
||||
function headingTag(title: string): string {
|
||||
const beforeOrder = title.split("〔")[0]?.trim() ?? title;
|
||||
const beforeParen = beforeOrder.split(/（|\(/)[0]?.trim() ?? beforeOrder;
|
||||
const beforeAt = beforeParen.split(" @ ")[0]?.trim() ?? beforeParen;
|
||||
const tag = (beforeAt || title).trim();
|
||||
return tag.length > 48 ? `${tag.slice(0, 48)}…` : tag;
|
||||
}
|
||||
|
||||
function collectMarks(content: string): Array<{ index: number; tag: string }> {
|
||||
const marks: Array<{ index: number; tag: string }> = [];
|
||||
let section = "";
|
||||
for (const match of content.matchAll(SPLIT)) {
|
||||
const raw = match[1] ?? "";
|
||||
const start = (match.index ?? 0) + (match[0].startsWith("\n") ? 1 : 0);
|
||||
if (raw.startsWith("#")) {
|
||||
const level = /^#+/.exec(raw)?.[0].length ?? 2;
|
||||
const title = raw.replace(/^#+\s*/, "");
|
||||
const tag = headingTag(title);
|
||||
if (level === 2) section = tag;
|
||||
marks.push({ index: start, tag });
|
||||
continue;
|
||||
}
|
||||
if (raw === "最后一轮用户输入" && section.startsWith("本轮")) {
|
||||
marks.push({ index: start, tag: "最后一轮用户输入" });
|
||||
continue;
|
||||
}
|
||||
if (
|
||||
(raw === "【用户】" || raw === "【助手】") &&
|
||||
section.startsWith("更早的对话")
|
||||
) {
|
||||
marks.push({
|
||||
index: start,
|
||||
tag: raw === "【用户】" ? "更早·用户" : "更早·助手",
|
||||
});
|
||||
}
|
||||
}
|
||||
return marks;
|
||||
}
|
||||
|
||||
/** 把一块正文切成有序标签。前缀里的标题与更早对话轮次分开，方便对上断点。 */
|
||||
export function outlineBlock(fallbackTag: string, content: string): ContextSegment[] {
|
||||
if (!content) return [];
|
||||
const marks = collectMarks(content);
|
||||
if (marks.length === 0) return [segment(fallbackTag, content)];
|
||||
const out: ContextSegment[] = [];
|
||||
if (marks[0]!.index > 0) {
|
||||
const head = content.slice(0, marks[0]!.index);
|
||||
if (head.trim()) out.push(segment(fallbackTag, head));
|
||||
}
|
||||
for (let i = 0; i < marks.length; i++) {
|
||||
const end = i + 1 < marks.length ? marks[i + 1]!.index : content.length;
|
||||
out.push(segment(marks[i]!.tag, content.slice(marks[i]!.index, end)));
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
function roleTag(role: string): string {
|
||||
if (role === "system") return "规则";
|
||||
if (role === "user") return "用户";
|
||||
if (role === "assistant") return "助手";
|
||||
if (role === "tool") return "工具结果";
|
||||
return role || "消息";
|
||||
}
|
||||
|
||||
/** 按实际送出的消息切段。工具定义若有，放在正文之前。 */
|
||||
export function outlineRequest(
|
||||
messages: Array<{ role: string; content?: string | null }>,
|
||||
tools?: ToolDefinition[],
|
||||
): ContextSegment[] {
|
||||
const out: ContextSegment[] = [];
|
||||
if (tools?.length) {
|
||||
out.push(segment("工具", JSON.stringify(tools)));
|
||||
}
|
||||
for (const message of messages) {
|
||||
const content = typeof message.content === "string" ? message.content : "";
|
||||
out.push(...outlineBlock(roleTag(message.role), content));
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
/**
|
||||
* 返回下一请求里第一段对不上的下标。
|
||||
* 下一请求比上一请求更短、但已有段全部相同，视为整段接上（-1）。
|
||||
*/
|
||||
export function firstCacheBreak(
|
||||
prev: Array<Pick<ContextSegment, "tag" | "hash">>,
|
||||
next: Array<Pick<ContextSegment, "tag" | "hash">>,
|
||||
): number {
|
||||
const shared = Math.min(prev.length, next.length);
|
||||
for (let i = 0; i < shared; i++) {
|
||||
if (prev[i]!.tag !== next[i]!.tag || prev[i]!.hash !== next[i]!.hash) return i;
|
||||
}
|
||||
if (next.length > prev.length) return prev.length;
|
||||
return -1;
|
||||
}
|
||||
|
||||
export function annotateCacheBreaks<
|
||||
T extends { id: string; at: string; sessionId?: string; segments?: ContextSegment[] },
|
||||
>(records: T[]): Map<string, CacheBreakInfo> {
|
||||
const bySession = new Map<string, T[]>();
|
||||
for (const record of records) {
|
||||
const key = record.sessionId?.trim() || `__none__:${record.id}`;
|
||||
const list = bySession.get(key) ?? [];
|
||||
list.push(record);
|
||||
bySession.set(key, list);
|
||||
}
|
||||
const marks = new Map<string, CacheBreakInfo>();
|
||||
for (const list of bySession.values()) {
|
||||
list.sort((a, b) => a.at.localeCompare(b.at) || a.id.localeCompare(b.id));
|
||||
let prev: ContextSegment[] | undefined;
|
||||
for (const record of list) {
|
||||
const segments = record.segments ?? [];
|
||||
if (!prev) {
|
||||
marks.set(record.id, {
|
||||
cold: true,
|
||||
breakAt: segments.length ? 0 : -1,
|
||||
tag: segments[0]?.tag,
|
||||
matched: 0,
|
||||
});
|
||||
prev = segments;
|
||||
continue;
|
||||
}
|
||||
const breakAt = firstCacheBreak(prev, segments);
|
||||
marks.set(record.id, {
|
||||
cold: false,
|
||||
breakAt,
|
||||
tag: breakAt >= 0 ? segments[breakAt]?.tag : undefined,
|
||||
matched: breakAt < 0 ? segments.length : breakAt,
|
||||
});
|
||||
prev = segments;
|
||||
}
|
||||
}
|
||||
return marks;
|
||||
}
|
||||
@@ -29,6 +29,9 @@ function applyToolCallDelta(
|
||||
const f = fn as Record<string, unknown>;
|
||||
if (typeof f.name === "string") entry.name = f.name;
|
||||
if (typeof f.arguments === "string") entry.arguments += f.arguments;
|
||||
else if (f.arguments && typeof f.arguments === "object" && !entry.arguments) {
|
||||
entry.arguments = JSON.stringify(f.arguments);
|
||||
}
|
||||
}
|
||||
acc.set(index, entry);
|
||||
}
|
||||
@@ -36,10 +39,10 @@ function applyToolCallDelta(
|
||||
|
||||
function toolCallsFromAccumulator(acc: ToolCallAccumulator): ParsedToolCall[] {
|
||||
const out: ParsedToolCall[] = [];
|
||||
for (const [, entry] of [...acc.entries()].sort((a, b) => a[0] - b[0])) {
|
||||
for (const [index, entry] of [...acc.entries()].sort((a, b) => a[0] - b[0])) {
|
||||
const name = entry.name?.trim();
|
||||
const id = entry.id?.trim();
|
||||
if (!name || !id) continue;
|
||||
if (!name) continue;
|
||||
const id = entry.id?.trim() || `call-${index}`;
|
||||
out.push({ id, name, arguments: entry.arguments || "{}" });
|
||||
}
|
||||
return out;
|
||||
|
||||
+103
-19
@@ -6,7 +6,15 @@ import type {
|
||||
LlmProvider,
|
||||
StreamCallbacks,
|
||||
TokenUsage,
|
||||
ToolDefinition,
|
||||
UsageDetail,
|
||||
UsageSink,
|
||||
} from "./client.js";
|
||||
import {
|
||||
outlineRequest,
|
||||
type ContextSegment,
|
||||
type RequestOutline,
|
||||
} from "./context-outline.js";
|
||||
import {
|
||||
recordTokenUsage,
|
||||
toMessageTokenUsage,
|
||||
@@ -28,6 +36,9 @@ export type LlmTrackingContext = {
|
||||
pendingReasoning?: string;
|
||||
/** 全量请求上下文；挂到下一条「结果向」系统消息 */
|
||||
pendingContextTrace?: LlmContextTrace;
|
||||
/** 本条回复之前的各次请求轮廓；查看时选题和填写都在 */
|
||||
pendingOutlines?: RequestOutline[];
|
||||
lastSegments?: ContextSegment[];
|
||||
};
|
||||
|
||||
function capturePendingTrace(
|
||||
@@ -41,12 +52,16 @@ function capturePendingTrace(
|
||||
caller: string | undefined,
|
||||
model?: string,
|
||||
generation?: CompleteOptions["generation"],
|
||||
tools?: ToolDefinition[],
|
||||
): void {
|
||||
const segments = outlineRequest(messages, tools);
|
||||
ctx.lastSegments = segments;
|
||||
ctx.pendingContextTrace = buildContextTrace({
|
||||
caller: caller ?? "unknown",
|
||||
messages,
|
||||
model,
|
||||
generation,
|
||||
segments,
|
||||
});
|
||||
}
|
||||
|
||||
@@ -55,9 +70,13 @@ function recordUsage(
|
||||
caller: string | undefined,
|
||||
model: string | undefined,
|
||||
usage: TokenUsage | undefined,
|
||||
meta?: { detail?: UsageDetail; latencyMs?: number; sink?: UsageSink },
|
||||
): void {
|
||||
if (!usage) return;
|
||||
const record = recordTokenUsage({
|
||||
const segments = ctx.lastSegments;
|
||||
ctx.lastSegments = undefined;
|
||||
const detail = meta?.detail;
|
||||
const record = usage
|
||||
? recordTokenUsage({
|
||||
sessionId: ctx.sessionId,
|
||||
bookId: ctx.bookId,
|
||||
bookTitle: ctx.bookTitle,
|
||||
@@ -69,8 +88,36 @@ function recordUsage(
|
||||
totalTokens: usage.totalTokens,
|
||||
cachedTokens: usage.cachedTokens,
|
||||
cacheMissTokens: usage.cacheMissTokens,
|
||||
});
|
||||
ctx.pendingUsage = toMessageTokenUsage(record);
|
||||
cacheWriteTokens: usage.cacheWriteTokens,
|
||||
turnId: detail?.turnId,
|
||||
step: detail?.step,
|
||||
retryIndex: detail?.retryIndex,
|
||||
segmentChars: detail?.segmentChars,
|
||||
plannedTags: detail?.plannedTags,
|
||||
succeededTags: detail?.succeededTags,
|
||||
failedTags: detail?.failedTags,
|
||||
finish: detail?.finish,
|
||||
latencyMs: meta?.latencyMs,
|
||||
segments,
|
||||
})
|
||||
: undefined;
|
||||
if (record && meta?.sink) meta.sink.recordId = record.id;
|
||||
if (record) ctx.pendingUsage = toMessageTokenUsage(record);
|
||||
if (!segments?.length && !record) return;
|
||||
const outline: RequestOutline = {
|
||||
at: record?.at ?? new Date().toISOString(),
|
||||
caller: record?.caller ?? caller ?? "unknown",
|
||||
step: detail?.step,
|
||||
turnId: detail?.turnId,
|
||||
model: record?.model ?? model,
|
||||
promptTokens: record?.promptTokens,
|
||||
completionTokens: record?.completionTokens,
|
||||
cachedTokens: record?.cachedTokens,
|
||||
cacheMissTokens: record?.cacheMissTokens,
|
||||
cacheWriteTokens: record?.cacheWriteTokens,
|
||||
segments: segments ?? [],
|
||||
};
|
||||
ctx.pendingOutlines = [...(ctx.pendingOutlines ?? []), outline];
|
||||
}
|
||||
|
||||
export class TokenTrackingProvider implements LlmProvider {
|
||||
@@ -79,6 +126,14 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
private readonly getContext: () => LlmTrackingContext,
|
||||
) {}
|
||||
|
||||
/** 调用方未指定时，用本局会话 id，供 OpenRouter 粘滞路由。 */
|
||||
private withSession<T extends CompleteOptions>(options?: T): T {
|
||||
const sessionId = this.getContext().sessionId?.trim();
|
||||
if (!sessionId) return (options ?? {}) as T;
|
||||
if (options?.sessionId?.trim()) return options;
|
||||
return { ...options, sessionId } as T;
|
||||
}
|
||||
|
||||
async complete(
|
||||
messages: Parameters<LlmProvider["complete"]>[0],
|
||||
options?: CompleteOptions,
|
||||
@@ -91,7 +146,8 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
messages,
|
||||
},
|
||||
async () => {
|
||||
const result = await this.inner.complete(messages, options);
|
||||
const started = Date.now();
|
||||
const result = await this.inner.complete(messages, this.withSession(options));
|
||||
const ctx = this.getContext();
|
||||
capturePendingTrace(
|
||||
ctx,
|
||||
@@ -100,7 +156,11 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
result.model,
|
||||
options?.generation,
|
||||
);
|
||||
recordUsage(ctx, options?.caller, result.model, result.usage);
|
||||
recordUsage(ctx, options?.caller, result.model, result.usage, {
|
||||
detail: options?.usageDetail,
|
||||
latencyMs: Date.now() - started,
|
||||
sink: options?.usageSink,
|
||||
});
|
||||
if (result.reasoning?.trim()) {
|
||||
ctx.pendingReasoning = result.reasoning.trim();
|
||||
}
|
||||
@@ -130,7 +190,8 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
},
|
||||
async () => {
|
||||
let reasoningBuf = "";
|
||||
const result = await inner.completeStream!(messages, options, {
|
||||
const started = Date.now();
|
||||
const result = await inner.completeStream!(messages, this.withSession(options), {
|
||||
onReasoningDelta: (delta) => {
|
||||
reasoningBuf += delta;
|
||||
const ctx = this.getContext();
|
||||
@@ -147,7 +208,11 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
result.model,
|
||||
options?.generation,
|
||||
);
|
||||
recordUsage(ctx, options?.caller, result.model, result.usage);
|
||||
recordUsage(ctx, options?.caller, result.model, result.usage, {
|
||||
detail: options?.usageDetail,
|
||||
latencyMs: Date.now() - started,
|
||||
sink: options?.usageSink,
|
||||
});
|
||||
if (result.reasoning?.trim()) {
|
||||
ctx.pendingReasoning = result.reasoning.trim();
|
||||
}
|
||||
@@ -169,7 +234,11 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
tools: options.tools,
|
||||
},
|
||||
async () => {
|
||||
const result = await this.inner.completeWithTools(messages, options);
|
||||
const started = Date.now();
|
||||
const result = await this.inner.completeWithTools(
|
||||
messages,
|
||||
this.withSession(options),
|
||||
);
|
||||
const ctx = this.getContext();
|
||||
capturePendingTrace(
|
||||
ctx,
|
||||
@@ -177,8 +246,13 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
options.caller,
|
||||
result.model,
|
||||
options.generation,
|
||||
options.tools,
|
||||
);
|
||||
recordUsage(ctx, options.caller, result.model, result.usage);
|
||||
recordUsage(ctx, options.caller, result.model, result.usage, {
|
||||
detail: options.usageDetail,
|
||||
latencyMs: Date.now() - started,
|
||||
sink: options.usageSink,
|
||||
});
|
||||
if (result.reasoning?.trim()) {
|
||||
ctx.pendingReasoning = result.reasoning.trim();
|
||||
}
|
||||
@@ -206,15 +280,20 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
},
|
||||
async () => {
|
||||
let reasoningBuf = "";
|
||||
const result = await inner.completeWithToolsStream!(messages, options, {
|
||||
onReasoningDelta: (delta) => {
|
||||
reasoningBuf += delta;
|
||||
const ctx = this.getContext();
|
||||
ctx.pendingReasoning = reasoningBuf;
|
||||
callbacks.onReasoningDelta?.(delta);
|
||||
const started = Date.now();
|
||||
const result = await inner.completeWithToolsStream!(
|
||||
messages,
|
||||
this.withSession(options),
|
||||
{
|
||||
onReasoningDelta: (delta) => {
|
||||
reasoningBuf += delta;
|
||||
const ctx = this.getContext();
|
||||
ctx.pendingReasoning = reasoningBuf;
|
||||
callbacks.onReasoningDelta?.(delta);
|
||||
},
|
||||
onContentDelta: callbacks.onContentDelta,
|
||||
},
|
||||
onContentDelta: callbacks.onContentDelta,
|
||||
});
|
||||
);
|
||||
const ctx = this.getContext();
|
||||
capturePendingTrace(
|
||||
ctx,
|
||||
@@ -222,8 +301,13 @@ export class TokenTrackingProvider implements LlmProvider {
|
||||
options.caller,
|
||||
result.model,
|
||||
options.generation,
|
||||
options.tools,
|
||||
);
|
||||
recordUsage(ctx, options.caller, result.model, result.usage);
|
||||
recordUsage(ctx, options.caller, result.model, result.usage, {
|
||||
detail: options.usageDetail,
|
||||
latencyMs: Date.now() - started,
|
||||
sink: options.usageSink,
|
||||
});
|
||||
if (result.reasoning?.trim()) {
|
||||
ctx.pendingReasoning = result.reasoning.trim();
|
||||
}
|
||||
|
||||
@@ -98,6 +98,8 @@ const TOOL_LABEL: Record<string, string> = {
|
||||
review_blackboard: "审黑板",
|
||||
finish: "结束",
|
||||
insert: "写入产物",
|
||||
reply_user: "回复用户",
|
||||
select_nodes: "选定节点",
|
||||
write_product: "写入产物",
|
||||
delete: "删除产物",
|
||||
declare_variable: "声明变量",
|
||||
@@ -214,6 +216,7 @@ export function labelCaller(caller: string | undefined): string {
|
||||
if (who === "main_agent" || who.startsWith("main_agent:")) return "总管";
|
||||
if (who === "intake_extract") return "需求抽取";
|
||||
if (who === "dictate_agent") return "对话落盘";
|
||||
if (who.startsWith("listen_serial_")) return "听书连载";
|
||||
if (who === "preset-probe") return "预设探测";
|
||||
if (who.startsWith("model-compare:")) {
|
||||
return `模型对比 ${who.slice("model-compare:".length)}`;
|
||||
|
||||
@@ -75,18 +75,24 @@ export function protagonistFragmentFromPersona(
|
||||
);
|
||||
}
|
||||
|
||||
/** 黑板上尚无该 tag 的主角设定时写入（可指定开场同槽 tag）。 */
|
||||
/**
|
||||
* 黑板上尚无该 tag 的主角设定时写入（可指定开场同槽 tag）。
|
||||
* overwrite：开场自带的同槽名片随开场改稿同步。
|
||||
*/
|
||||
export function ensureProtagonistFromPersona(
|
||||
board: Blackboard,
|
||||
persona: OpeningUserPersona | null | undefined,
|
||||
tag: string = PROTAGONIST_TAG,
|
||||
opts: { overwrite?: boolean } = {},
|
||||
): boolean {
|
||||
if (!persona?.name?.trim()) return false;
|
||||
const writeTag = tag.trim() || PROTAGONIST_TAG;
|
||||
if (board.getContentByTag(writeTag)?.trim()) return false;
|
||||
const existing = board.getContentByTag(writeTag)?.trim();
|
||||
const content = protagonistFragmentFromPersona(persona);
|
||||
if (existing && (!opts.overwrite || existing === content.trim())) return false;
|
||||
board.write({
|
||||
tag: writeTag,
|
||||
content: protagonistFragmentFromPersona(persona),
|
||||
content,
|
||||
source: "runtime",
|
||||
});
|
||||
return true;
|
||||
|
||||
+110
-21
@@ -140,6 +140,10 @@ import {
|
||||
isDictateModeValue,
|
||||
} from "../dictate/types.js";
|
||||
import { applyDictatePlayBind } from "../dictate/play-bind.js";
|
||||
import {
|
||||
applyMaintenanceStrategyToPlaySlots,
|
||||
resolveMaintenanceStrategyFromBoard,
|
||||
} from "../skills/maintenance-strategy.js";
|
||||
import {
|
||||
dictateProductFamily,
|
||||
mergeGenerationRulesArtifacts,
|
||||
@@ -148,6 +152,11 @@ import {
|
||||
VARIABLE_CATALOG_TAG,
|
||||
seedVariablesFromCatalog,
|
||||
} from "../skills/variable-catalog.js";
|
||||
import {
|
||||
VARIABLE_DESIGN_TAG,
|
||||
syncVariableDesignToRuntime,
|
||||
writeVariableDesignSync,
|
||||
} from "../skills/variable-design-compile.js";
|
||||
import {
|
||||
VALUE_MAP_TAG,
|
||||
reprojectValueMaps,
|
||||
@@ -164,6 +173,11 @@ import {
|
||||
parseShellAdaptationFromReplyFormat,
|
||||
presentOutputFromFallback,
|
||||
} from "../skills/present-packet.js";
|
||||
import { parseReplyDataContract } from "../skills/reply-contract.js";
|
||||
import {
|
||||
mergePresentWithProjections,
|
||||
projectPresentBlocks,
|
||||
} from "../skills/present-project.js";
|
||||
import { debugLog, labelAction, logStateChange, sessionSnap } from "../log.js";
|
||||
import {
|
||||
executeChance,
|
||||
@@ -2049,21 +2063,45 @@ export class PhaseRuntime {
|
||||
// 偏好/文风改由节点产物（设计.用户需求 / 设计.叙事指南*）挂载，不再全量自动摊库
|
||||
if (this.isDictateIntakeMode()) {
|
||||
const bound = applyDictatePlayBind(this.blackboard);
|
||||
if (bound.boundTags.length) {
|
||||
if (!bound.skipped && bound.boundTags.length) {
|
||||
this.onMessage(
|
||||
`[对话落盘] 已自动挂载 ${bound.boundTags.length} 个产物到游玩上下文：${bound.boundTags.join("、")}`,
|
||||
);
|
||||
} else {
|
||||
} else if (!bound.skipped) {
|
||||
this.onMessage("[对话落盘] 已生成默认游玩规格（尚无产物可挂载）");
|
||||
}
|
||||
const strategyRaw = this.blackboard.getContentByTag("设计.维护策略")?.trim();
|
||||
if (strategyRaw) {
|
||||
try {
|
||||
const s = JSON.parse(strategyRaw) as {
|
||||
advice?: string;
|
||||
auditor?: boolean;
|
||||
inline?: boolean;
|
||||
};
|
||||
if (s.auditor) {
|
||||
this.onMessage(
|
||||
`[维护策略] 旁观副 LLM 填表（play_slots.auditor=true）`,
|
||||
);
|
||||
} else if (s.advice && s.advice !== "none") {
|
||||
this.onMessage(`[维护策略] ${s.advice} → 旁观未开`);
|
||||
}
|
||||
} catch {
|
||||
/* ignore */
|
||||
}
|
||||
}
|
||||
this.seedVariablesAndReproject("opening-setup");
|
||||
} else {
|
||||
const finalTag = this.blackboard.getContentByTag(WORKER_SET_FINAL_TAG)?.trim();
|
||||
const draft = this.blackboard.getContentByTag(WORKER_SET_DRAFT_TAG)?.trim();
|
||||
if (!finalTag) {
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(this.blackboard);
|
||||
const play_slots = applyMaintenanceStrategyToPlaySlots(
|
||||
defaultPlaySlots(),
|
||||
strategy,
|
||||
);
|
||||
const spec = draft?.trim()
|
||||
? draft
|
||||
: JSON.stringify({ play_slots: defaultPlaySlots() });
|
||||
: JSON.stringify({ play_slots }, null, 2);
|
||||
this.writeWorkerTagContent(WORKER_SET_FINAL_TAG, spec, "opening-setup");
|
||||
}
|
||||
}
|
||||
@@ -2077,36 +2115,84 @@ export class PhaseRuntime {
|
||||
};
|
||||
}
|
||||
|
||||
/** 目录初值 → 变量.当前；再按 设计.变量映射 重写投影 tag */
|
||||
/** 变量设计 → 目录 → 变量.当前；再按 设计.变量映射 重写投影 tag */
|
||||
private seedVariablesAndReproject(actor: string): void {
|
||||
const seeded = seedVariablesFromCatalog({
|
||||
const designSync = syncVariableDesignToRuntime({
|
||||
designRaw: this.blackboard.getContentByTag(VARIABLE_DESIGN_TAG),
|
||||
catalogRaw: this.blackboard.getContentByTag(VARIABLE_CATALOG_TAG),
|
||||
currentRaw: this.blackboard.getContentByTag(OPENING_CURRENT_VARS_TAG),
|
||||
initialRaw: this.blackboard.getContentByTag(OPENING_INITIAL_VARS_TAG),
|
||||
actor: `system:${actor}`,
|
||||
});
|
||||
if (seeded.initial) {
|
||||
this.blackboard.write({
|
||||
tag: OPENING_INITIAL_VARS_TAG,
|
||||
content: seeded.initial,
|
||||
source: actor,
|
||||
if (designSync.catalog || designSync.current || designSync.initial) {
|
||||
writeVariableDesignSync(designSync, (tag, content) => {
|
||||
this.blackboard.write({ tag, content, source: actor });
|
||||
});
|
||||
}
|
||||
if (seeded.current) {
|
||||
this.blackboard.write({
|
||||
tag: OPENING_CURRENT_VARS_TAG,
|
||||
content: seeded.current,
|
||||
source: actor,
|
||||
});
|
||||
if (seeded.seededKeys.length) {
|
||||
if (designSync.added.length) {
|
||||
this.onMessage(
|
||||
`[变量] 已种初值：${seeded.seededKeys.join("、")}`,
|
||||
`[变量] 已从变量设计同步目录：${designSync.added.join("、")}`,
|
||||
);
|
||||
}
|
||||
if (designSync.current && designSync.seededKeys.length) {
|
||||
this.onMessage(
|
||||
`[变量] 已种初值：${designSync.seededKeys.join("、")}`,
|
||||
);
|
||||
}
|
||||
} else {
|
||||
const seeded = seedVariablesFromCatalog({
|
||||
catalogRaw: this.blackboard.getContentByTag(VARIABLE_CATALOG_TAG),
|
||||
currentRaw: this.blackboard.getContentByTag(OPENING_CURRENT_VARS_TAG),
|
||||
initialRaw: this.blackboard.getContentByTag(OPENING_INITIAL_VARS_TAG),
|
||||
actor: `system:${actor}`,
|
||||
});
|
||||
if (seeded.initial) {
|
||||
this.blackboard.write({
|
||||
tag: OPENING_INITIAL_VARS_TAG,
|
||||
content: seeded.initial,
|
||||
source: actor,
|
||||
});
|
||||
}
|
||||
if (seeded.current) {
|
||||
this.blackboard.write({
|
||||
tag: OPENING_CURRENT_VARS_TAG,
|
||||
content: seeded.current,
|
||||
source: actor,
|
||||
});
|
||||
if (seeded.seededKeys.length) {
|
||||
this.onMessage(
|
||||
`[变量] 已种初值：${seeded.seededKeys.join("、")}`,
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
this.applyValueMapProjections(actor);
|
||||
}
|
||||
|
||||
/** 把 program_projection 块从黑板灌进 present 正文 */
|
||||
private enrichPresentWithProjections(content: string): string {
|
||||
const trimmed = content.trim();
|
||||
// 纯 Markdown 终稿保持原文；投影由视图 presentationProjection 灌进壳
|
||||
if (!trimmed.startsWith("{")) return content;
|
||||
const replyFormatRaw =
|
||||
this.blackboard.getContentByTag("设计.正文组成") ??
|
||||
this.blackboard.getContentByTag("设计.0层正文组成") ??
|
||||
this.blackboard.getContentByTag("设计.回复格式");
|
||||
const contract = parseReplyDataContract(replyFormatRaw);
|
||||
if (!contract?.programBlocks.length) return content;
|
||||
const projections = projectPresentBlocks({
|
||||
contract,
|
||||
statusBarRaw:
|
||||
this.blackboard.getContentByTag("设计.监控栏") ??
|
||||
this.blackboard.getContentByTag("设计.回复呈现"),
|
||||
currentVarsRaw: this.blackboard.getContentByTag(OPENING_CURRENT_VARS_TAG),
|
||||
topologyRaw: this.blackboard.getContentByTag("运行.拓扑"),
|
||||
readTag: (tag) => this.blackboard.getContentByTag(tag),
|
||||
});
|
||||
if (!Object.keys(projections).length) return content;
|
||||
const shell =
|
||||
parseShellAdaptationFromReplyFormat(replyFormatRaw)?.shell_id ?? "prose";
|
||||
return mergePresentWithProjections(content, projections, shell);
|
||||
}
|
||||
|
||||
/** 按当前真值持续投影映射 tag（非 once） */
|
||||
private applyValueMapProjections(workerId: string): void {
|
||||
if (!this.blackboard.getContentByTag(VALUE_MAP_TAG)?.trim()) return;
|
||||
@@ -2589,10 +2675,13 @@ export class PhaseRuntime {
|
||||
): string {
|
||||
const tableTags = new Set(["运行.初始变量", "变量.当前"]);
|
||||
let toWrite = content;
|
||||
if (tag === "输出.用户展示" || tag === "输出.开场白") {
|
||||
toWrite = this.enrichPresentWithProjections(toWrite);
|
||||
}
|
||||
let prevDoc: TableDoc | null = null;
|
||||
let nextDoc: TableDoc | null = null;
|
||||
if (tableTags.has(tag)) {
|
||||
const patch = parseTableDoc(content);
|
||||
const patch = parseTableDoc(toWrite);
|
||||
if (patch) {
|
||||
prevDoc = parseTableDoc(this.blackboard.getContentByTag(tag));
|
||||
const { doc, skipped } = mergeTableCells({
|
||||
|
||||
@@ -26,6 +26,7 @@ export type AgentMessageKind =
|
||||
| "orchestrator_prompt"
|
||||
| "orchestrator_assessment"
|
||||
| "dictate_reply"
|
||||
| "listen_reply"
|
||||
| "agent_tool"
|
||||
| "worker_running"
|
||||
| "worker_output"
|
||||
|
||||
+83
-33
@@ -1,5 +1,6 @@
|
||||
import type { IncomingMessage, ServerResponse } from "node:http";
|
||||
import { listSkills } from "../skills/loader.js";
|
||||
import type { ModuleCatalogEntry } from "../skills/creation-flow.js";
|
||||
import { createBook, deleteBook, duplicateBook, getBook, listBooks, updateBook } from "../book/store.js";
|
||||
import { sessionManager } from "./session-manager.js";
|
||||
|
||||
@@ -39,17 +40,64 @@ function moduleStatusFromSections(
|
||||
return depth >= 2 ? "ready" : "partial";
|
||||
}
|
||||
|
||||
type ModuleBrowseRow = {
|
||||
id: string;
|
||||
name: string;
|
||||
declaration: string;
|
||||
artifact: string;
|
||||
hasPrompt: boolean;
|
||||
sectionsPresent: string[];
|
||||
status: ModuleStatus;
|
||||
group: string;
|
||||
presentMode: "message" | "zero_layer" | null;
|
||||
layer: "intermediate" | "final" | null;
|
||||
repeatable: boolean;
|
||||
closer: boolean;
|
||||
priorArtifact: boolean;
|
||||
programSide: boolean;
|
||||
libraries: string[];
|
||||
land: string;
|
||||
};
|
||||
|
||||
async function moduleBrowseFields(
|
||||
entry: ModuleCatalogEntry,
|
||||
): Promise<
|
||||
Omit<
|
||||
ModuleBrowseRow,
|
||||
| "id"
|
||||
| "name"
|
||||
| "declaration"
|
||||
| "artifact"
|
||||
| "hasPrompt"
|
||||
| "sectionsPresent"
|
||||
| "status"
|
||||
>
|
||||
> {
|
||||
const { dictateModuleLandHint, isDictateProgramModule } = await import(
|
||||
"../skills/creation-flow.js"
|
||||
);
|
||||
const { moduleBrowseGroup, modulePresentationMode } = await import(
|
||||
"../dictate/creation-plan.js"
|
||||
);
|
||||
const { getLibrary } = await import("../libraries/registry.js");
|
||||
return {
|
||||
group: moduleBrowseGroup(entry),
|
||||
presentMode: modulePresentationMode(entry.id),
|
||||
layer: entry.layer ?? null,
|
||||
repeatable: Boolean(entry.repeatable),
|
||||
closer: Boolean(entry.closer),
|
||||
priorArtifact: entry.kind === "prior-artifact",
|
||||
programSide: isDictateProgramModule(entry),
|
||||
libraries: (entry.libraries ?? []).map(
|
||||
(id) => getLibrary(id)?.label ?? id,
|
||||
),
|
||||
land: dictateModuleLandHint(entry),
|
||||
};
|
||||
}
|
||||
|
||||
async function loadModulesPayload(skillId: string): Promise<{
|
||||
skillPackId: string;
|
||||
modules: Array<{
|
||||
id: string;
|
||||
name: string;
|
||||
declaration: string;
|
||||
artifact: string;
|
||||
hasPrompt: boolean;
|
||||
sectionsPresent: string[];
|
||||
status: ModuleStatus;
|
||||
}>;
|
||||
groups: Array<{ id: string; label: string; hint: string }>;
|
||||
modules: ModuleBrowseRow[];
|
||||
}> {
|
||||
const { loadSkill } = await import("../skills/loader.js");
|
||||
const {
|
||||
@@ -59,13 +107,14 @@ async function loadModulesPayload(skillId: string): Promise<{
|
||||
catalogModulesForIntake,
|
||||
MODULE_SECTION_IDS,
|
||||
} = await import("../skills/creation-flow.js");
|
||||
const { MODULE_BROWSE_GROUPS } = await import("../dictate/creation-plan.js");
|
||||
const skill = await loadSkill(skillId);
|
||||
if (!skill.skillPackRoot) {
|
||||
return { skillPackId: skillId, modules: [] };
|
||||
return { groups: [...MODULE_BROWSE_GROUPS], modules: [] };
|
||||
}
|
||||
const catalog = await loadModuleCatalog(skill.skillPackRoot);
|
||||
const modules = [];
|
||||
for (const m of catalogModulesForIntake(catalog ?? { modules: [] }, "recipe")) {
|
||||
const modules: ModuleBrowseRow[] = [];
|
||||
for (const m of catalogModulesForIntake(catalog ?? { modules: [] }, "dictate")) {
|
||||
const prompt = await loadModulePrompt(skill.skillPackRoot, m.id);
|
||||
const sections = prompt
|
||||
? parseModulePromptSections(prompt)
|
||||
@@ -83,27 +132,23 @@ async function loadModulesPayload(skillId: string): Promise<{
|
||||
hasPrompt,
|
||||
sectionsPresent: [...sectionsPresent],
|
||||
status: moduleStatusFromSections(hasPrompt, sectionsPresent, taskBody),
|
||||
...(await moduleBrowseFields(m)),
|
||||
});
|
||||
}
|
||||
return { skillPackId: skillId, modules };
|
||||
return { groups: [...MODULE_BROWSE_GROUPS], modules };
|
||||
}
|
||||
|
||||
async function loadModuleDetailPayload(
|
||||
skillId: string,
|
||||
moduleId: string,
|
||||
): Promise<{
|
||||
skillPackId: string;
|
||||
id: string;
|
||||
name: string;
|
||||
declaration: string;
|
||||
artifact: string;
|
||||
hasPrompt: boolean;
|
||||
sectionsPresent: string[];
|
||||
status: ModuleStatus;
|
||||
sections: Record<string, string>;
|
||||
meta: Record<string, unknown> | null;
|
||||
raw: string;
|
||||
} | null> {
|
||||
): Promise<
|
||||
| (ModuleBrowseRow & {
|
||||
sections: Record<string, string>;
|
||||
meta: Record<string, unknown> | null;
|
||||
raw: string;
|
||||
})
|
||||
| null
|
||||
> {
|
||||
const { loadSkill } = await import("../skills/loader.js");
|
||||
const {
|
||||
loadModuleCatalog,
|
||||
@@ -141,7 +186,6 @@ async function loadModuleDetailPayload(
|
||||
}
|
||||
}
|
||||
return {
|
||||
skillPackId: skillId,
|
||||
id: entry.id,
|
||||
name: entry.name,
|
||||
declaration: entry.declaration,
|
||||
@@ -153,6 +197,7 @@ async function loadModuleDetailPayload(
|
||||
sectionsPresent,
|
||||
sections.task,
|
||||
),
|
||||
...(await moduleBrowseFields(entry)),
|
||||
sections,
|
||||
meta,
|
||||
raw: prompt ?? "",
|
||||
@@ -254,7 +299,7 @@ export async function handleBooksApi(
|
||||
return true;
|
||||
}
|
||||
|
||||
/** 默认配方包的技能池（编排备选） */
|
||||
/** 对话落盘节点：按角色分组，不含旧配方包 */
|
||||
if (pathname === "/api/modules" && req.method === "GET") {
|
||||
const { DEFAULT_ORCHESTRATOR_ID } = await import(
|
||||
"../config/default-orchestrator.js"
|
||||
@@ -343,8 +388,8 @@ export async function handleBooksApi(
|
||||
orchestratorId?: string;
|
||||
/** 旧客户端可传；新建作品改由创作对话判断路径 */
|
||||
recipeId?: string;
|
||||
/** 旧客户端兼容；新建作品统一使用 dictate */
|
||||
creationMode?: "recipe" | "dictate";
|
||||
/** 进料方式：对话落盘（默认）| 听书连载 | 旧工序编排 */
|
||||
creationMode?: "recipe" | "dictate" | "listen_serial";
|
||||
};
|
||||
const skills = await listSkills();
|
||||
const requested = body.orchestratorId?.trim();
|
||||
@@ -356,14 +401,19 @@ export async function handleBooksApi(
|
||||
json(res, 400, { error: "没有可用的编排器技能包" });
|
||||
return true;
|
||||
}
|
||||
const book = createBook({ title: body.title });
|
||||
const rawMode = body.creationMode;
|
||||
const creationMode =
|
||||
rawMode === "listen_serial" || rawMode === "recipe" || rawMode === "dictate"
|
||||
? rawMode
|
||||
: "dictate";
|
||||
const book = createBook({ title: body.title, creationMode });
|
||||
updateBook(book.id, {
|
||||
activeSkillId: director.name,
|
||||
activeSkillName: director.description?.split("\n")[0]?.slice(0, 80) || director.name,
|
||||
orchestratorId: director.name,
|
||||
orchestratorName: director.name,
|
||||
creationMode,
|
||||
});
|
||||
const creationMode = "dictate";
|
||||
const session = await sessionManager.createForBook(
|
||||
book.id,
|
||||
director.name,
|
||||
|
||||
@@ -55,6 +55,7 @@ const PHASE_UNIT_LABELS: Record<string, string> = {
|
||||
const CREATION_MODE_LABELS: Record<string, string> = {
|
||||
recipe: "工序编排",
|
||||
dictate: "对话落盘",
|
||||
listen_serial: "听书连载",
|
||||
};
|
||||
|
||||
const SKILL_PACK_LABELS: Record<string, string> = {
|
||||
|
||||
+299
-35
@@ -93,6 +93,7 @@ import {
|
||||
sessionSkillPackId,
|
||||
skillPacksMatch,
|
||||
} from "../book/skill-id.js";
|
||||
import type { RequestOutline } from "../llm/context-outline.js";
|
||||
import {
|
||||
pruneContextTraces,
|
||||
type LlmContextTrace,
|
||||
@@ -102,6 +103,8 @@ import type { BlackboardItem } from "../types/blackboard.js";
|
||||
import { Blackboard } from "../blackboard/blackboard.js";
|
||||
import { parseWorkerSetYaml } from "../skills/worker-set-parse.js";
|
||||
import { parseShellAdaptationFromReplyFormat } from "../skills/present-packet.js";
|
||||
import { parseReplyDataContract } from "../skills/reply-contract.js";
|
||||
import { projectPresentBlocks } from "../skills/present-project.js";
|
||||
import {
|
||||
CREATION_ACCEPTED_CONTENT_TAG,
|
||||
CREATION_ACCEPTED_UNITS_TAG,
|
||||
@@ -182,6 +185,16 @@ import {
|
||||
sortDictateProducts,
|
||||
type DictateProduct,
|
||||
} from "../dictate/index.js";
|
||||
import {
|
||||
LISTEN_SERIAL_MODE_VALUE,
|
||||
isListenSerialModeValue,
|
||||
markListenConsumed,
|
||||
readListenState,
|
||||
runListenSerialTurn,
|
||||
toListenView,
|
||||
type ListenSerialView,
|
||||
} from "../listen-serial/index.js";
|
||||
import { synthesizeSpeech, type TtsSynthesizeResult } from "../tts/index.js";
|
||||
import {
|
||||
formatQuestionAnswersForAi,
|
||||
formatQuestionAnswersForDisplay,
|
||||
@@ -255,6 +268,8 @@ export type ChatMessage = {
|
||||
thinking?: string;
|
||||
/** 该次 LLM 请求实际携带的全量上下文（观察用；按设置只保留最新 N 条） */
|
||||
contextTrace?: LlmContextTrace;
|
||||
/** 本条回复对应的各次请求标签轮廓（选题、填写都在；不含正文） */
|
||||
requestOutlines?: RequestOutline[];
|
||||
/** run 验收后过程讨论折叠（主 feed 隐藏）；创作 messages 不删不藏 */
|
||||
compressed?: boolean;
|
||||
tokenUsage?: MessageTokenUsage;
|
||||
@@ -323,6 +338,14 @@ export type SessionView = {
|
||||
block_labels?: Record<string, string>;
|
||||
empty_states?: Record<string, string>;
|
||||
};
|
||||
/** 正文组成固定前端。游玩和开场把 present.v1 灌进去，替换示例内容。 */
|
||||
presentationFrontend?: { html: string; css: string; js: string };
|
||||
/** 正文组成数据契约里的区。格式化时按 id 填进对应 data-block。 */
|
||||
presentationSlots?: Array<{ id: string; type: string }>;
|
||||
/**
|
||||
* 程序投影块（来自 变量.当前 等）。前端灌 present 时与消息正文合并，随真值刷新。
|
||||
*/
|
||||
presentationProjection?: Record<string, unknown>;
|
||||
/** 总管正在输出的思维链（流式，轮询用） */
|
||||
agentThinking?: string;
|
||||
/** Agent / Worker 流式输出（轮询用） */
|
||||
@@ -333,10 +356,12 @@ export type SessionView = {
|
||||
*/
|
||||
runActive?: boolean;
|
||||
resumed?: boolean;
|
||||
/** 创作进料：recipe=配方节点流；dictate=转述整理直填 */
|
||||
creationMode?: "recipe" | "dictate";
|
||||
/** 创作进料：recipe=配方节点流；dictate=转述整理；listen_serial=听书连载 */
|
||||
creationMode?: "recipe" | "dictate" | "listen_serial";
|
||||
/** 转述式已写入黑板的产物（按相对序） */
|
||||
dictateProducts?: Array<{ tag: string; content: string; order?: number }>;
|
||||
/** 听书连载投影（大纲/正文缓冲/未读字数） */
|
||||
listenSerial?: ListenSerialView;
|
||||
tokenStats?: {
|
||||
sessionTotal: number;
|
||||
sessionCached?: number;
|
||||
@@ -555,8 +580,16 @@ export class SessionManager {
|
||||
this.writeOpeningPersonaChoice(s, true);
|
||||
}
|
||||
|
||||
// 与进游玩一致：用完整文档（含 present.v1），勿用侧栏预览的压扁正文
|
||||
const rawOnBoard = board.getContentByTag(picked.tag)?.trim() ?? "";
|
||||
const body =
|
||||
splitOpeningDocument(
|
||||
rawOnBoard || picked.document || picked.text,
|
||||
).text.trim() ||
|
||||
picked.document?.trim() ||
|
||||
picked.text.trim();
|
||||
const expanded = this.expandDirectivesText(
|
||||
picked.text,
|
||||
body,
|
||||
this.playDirectiveContext(s),
|
||||
);
|
||||
board.write({
|
||||
@@ -566,7 +599,13 @@ export class SessionManager {
|
||||
});
|
||||
const openingAt = s.messages.findIndex((m) => m.title === "开场白");
|
||||
if (openingAt >= 0) {
|
||||
s.messages[openingAt] = this.playOpeningMessage(expanded);
|
||||
// 保留 id，前端才能 patch 呈现壳 forceFill，而不是拆 iframe 后软灌样本槽
|
||||
s.messages[openingAt] = this.playOpeningMessage(
|
||||
expanded,
|
||||
s.messages[openingAt]!.id,
|
||||
);
|
||||
} else {
|
||||
s.messages.unshift(this.playOpeningMessage(expanded));
|
||||
}
|
||||
if (s.bookId) this.persist(s);
|
||||
return this.toView(sessionId);
|
||||
@@ -613,11 +652,13 @@ export class SessionManager {
|
||||
|
||||
const next = applyContextOrderEdit(current!, edit);
|
||||
const serialized = serializeContextOrder(next);
|
||||
const orderMeta = bb.getLatestByTag(CONTEXT_ORDER_TAG)?.metadata;
|
||||
|
||||
bb.write({
|
||||
tag: CONTEXT_ORDER_TAG,
|
||||
content: serialized,
|
||||
source: "user",
|
||||
metadata: orderMeta,
|
||||
});
|
||||
|
||||
if (workerSetRaw) {
|
||||
@@ -631,6 +672,7 @@ export class SessionManager {
|
||||
tag: targetTag,
|
||||
content: mergeContextOrderIntoWorkerSetJson(workerSetRaw, next),
|
||||
source: "user",
|
||||
metadata: bb.getLatestByTag(targetTag)?.metadata,
|
||||
});
|
||||
}
|
||||
} else {
|
||||
@@ -639,6 +681,7 @@ export class SessionManager {
|
||||
tag: CONTEXT_ORDER_TAG,
|
||||
content: serialized,
|
||||
source: "user",
|
||||
metadata: orderMeta,
|
||||
});
|
||||
}
|
||||
|
||||
@@ -670,7 +713,7 @@ export class SessionManager {
|
||||
bookId?: string,
|
||||
preselectSkillId?: string,
|
||||
selectedRecipeId?: string,
|
||||
opts?: { creationMode?: "recipe" | "dictate" },
|
||||
opts?: { creationMode?: "recipe" | "dictate" | "listen_serial" },
|
||||
): Promise<SessionView> {
|
||||
const id = randomUUID();
|
||||
const messages: ChatMessage[] = [];
|
||||
@@ -713,18 +756,22 @@ export class SessionManager {
|
||||
} else {
|
||||
await runtime.start();
|
||||
}
|
||||
const useDictate = opts?.creationMode === "dictate";
|
||||
const creationMode = opts?.creationMode ?? "recipe";
|
||||
const useDictate = creationMode === "dictate";
|
||||
const useListen = creationMode === "listen_serial";
|
||||
await this.refreshRecipeOptions(
|
||||
managed,
|
||||
useDictate ? "dictate" : "recipe",
|
||||
useDictate || useListen ? "dictate" : "recipe",
|
||||
);
|
||||
|
||||
if (selectedRecipeId?.trim()) {
|
||||
await this.applySelectedRecipe(managed, selectedRecipeId.trim(), {
|
||||
seedFlow: !useDictate,
|
||||
seedFlow: !useDictate && !useListen,
|
||||
});
|
||||
}
|
||||
if (useDictate) {
|
||||
if (useListen) {
|
||||
this.enterListenSerialMode(managed);
|
||||
} else if (useDictate) {
|
||||
this.enterDictateMode(managed);
|
||||
} else {
|
||||
await this.maybeKickAgentAfterDemand(runtime);
|
||||
@@ -780,6 +827,145 @@ export class SessionManager {
|
||||
);
|
||||
}
|
||||
|
||||
/** 听书连载：薄上下文连续产文，不走对话落盘工具环 */
|
||||
private enterListenSerialMode(s: ManagedSession): void {
|
||||
const board = s.runtime.getBlackboard();
|
||||
board.write({
|
||||
tag: CREATION_INTAKE_MODE_TAG,
|
||||
content: LISTEN_SERIAL_MODE_VALUE,
|
||||
source: "runtime",
|
||||
});
|
||||
s.runtime.enterDictateIntake(
|
||||
"听书连载：用几句话描述题材与爽点；确认后我会生成大纲并自动补正文。不满意就说一句意见改后面。",
|
||||
);
|
||||
const welcome =
|
||||
"听书连载：说几句你想听的故事（题材、人设、爽点即可）。我会轻回问或给出草案；确认后自动出大纲和正文缓冲。听的过程中随时一句意见，只改后面的走向。";
|
||||
s.messages.push({
|
||||
id: randomUUID(),
|
||||
role: "system",
|
||||
text: welcome,
|
||||
createdAt: new Date().toISOString(),
|
||||
kind: "orchestrator_prompt",
|
||||
actor: "orchestrator",
|
||||
title: "听书连载",
|
||||
body: welcome,
|
||||
});
|
||||
}
|
||||
|
||||
private isListenSerialMode(s: ManagedSession): boolean {
|
||||
return isListenSerialModeValue(
|
||||
s.runtime.getBlackboard().getContentByTag(CREATION_INTAKE_MODE_TAG),
|
||||
);
|
||||
}
|
||||
|
||||
private creationIntakeMode(
|
||||
s: ManagedSession,
|
||||
): "recipe" | "dictate" | "listen_serial" {
|
||||
if (this.isListenSerialMode(s)) return "listen_serial";
|
||||
if (this.isDictateMode(s)) return "dictate";
|
||||
return "recipe";
|
||||
}
|
||||
|
||||
private extractListenDialogue(
|
||||
messages: ChatMessage[],
|
||||
): Array<{ role: "user" | "assistant"; text: string }> {
|
||||
const out: Array<{ role: "user" | "assistant"; text: string }> = [];
|
||||
for (const m of messages) {
|
||||
if (m.kind === "user_input" || m.role === "user") {
|
||||
const t = (m.body ?? m.text ?? "").trim();
|
||||
if (t) out.push({ role: "user", text: t });
|
||||
continue;
|
||||
}
|
||||
if (m.kind === "listen_reply") {
|
||||
const t = (m.body ?? m.text ?? "").trim();
|
||||
if (t) out.push({ role: "assistant", text: t });
|
||||
}
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
private async runListenSerialTurnForSession(
|
||||
s: ManagedSession,
|
||||
userText: string,
|
||||
): Promise<void> {
|
||||
const llm = createDefaultMainAgentLlm(s.trackingRef, () =>
|
||||
this.personaForManaged(s),
|
||||
);
|
||||
const dialogue = this.extractListenDialogue(s.messages);
|
||||
const result = await runListenSerialTurn({
|
||||
llm,
|
||||
board: s.runtime.getBlackboard(),
|
||||
dialogue,
|
||||
userText,
|
||||
});
|
||||
const keep = normalizeContextTraceKeepLatest(
|
||||
loadAppSettings().contextTraceKeepLatest,
|
||||
);
|
||||
const requestOutlines = s.trackingRef.current.pendingOutlines;
|
||||
const pendingTrace = s.trackingRef.current.pendingContextTrace;
|
||||
const tokenUsage = s.trackingRef.current.pendingUsage;
|
||||
s.trackingRef.current.pendingOutlines = undefined;
|
||||
s.trackingRef.current.pendingContextTrace = undefined;
|
||||
s.trackingRef.current.pendingUsage = undefined;
|
||||
s.messages.push({
|
||||
id: randomUUID(),
|
||||
role: "system",
|
||||
text: result.reply,
|
||||
createdAt: new Date().toISOString(),
|
||||
kind: "listen_reply",
|
||||
actor: "orchestrator",
|
||||
title: "听书连载",
|
||||
body: result.reply,
|
||||
...(tokenUsage ? { tokenUsage } : {}),
|
||||
...(keep > 0 && pendingTrace ? { contextTrace: pendingTrace } : {}),
|
||||
...(requestOutlines?.length ? { requestOutlines } : {}),
|
||||
});
|
||||
s.messages = pruneContextTraces(s.messages, keep);
|
||||
s.runtime.enterDictateIntake(
|
||||
result.softStop
|
||||
? "本段已告一段落。输入「下一波」或「收工」，或一句意见改向。"
|
||||
: "继续听/读，或说「继续」补货；不满意直接一句意见。",
|
||||
);
|
||||
}
|
||||
|
||||
/** 听书：标记已消费字数（阅读进度） */
|
||||
markListenSerialConsumed(
|
||||
sessionId: string,
|
||||
consumedChars: number,
|
||||
): SessionView {
|
||||
const s = this.require(sessionId);
|
||||
if (!this.isListenSerialMode(s)) {
|
||||
throw new Error("当前不是听书连载模式");
|
||||
}
|
||||
markListenConsumed(s.runtime.getBlackboard(), consumedChars);
|
||||
if (s.bookId) this.persist(s);
|
||||
return this.toView(sessionId);
|
||||
}
|
||||
|
||||
/** 听书：按节合成语音（默认 Edge） */
|
||||
async synthesizeListenSection(
|
||||
sessionId: string,
|
||||
sectionId: string,
|
||||
): Promise<TtsSynthesizeResult & { sectionId: string; title: string }> {
|
||||
const s = this.require(sessionId);
|
||||
if (!this.isListenSerialMode(s)) {
|
||||
throw new Error("当前不是听书连载模式");
|
||||
}
|
||||
const id = sectionId.trim();
|
||||
if (!id) throw new Error("需要 sectionId");
|
||||
const state = readListenState(s.runtime.getBlackboard());
|
||||
const section = state.sections.find((x) => x.id === id);
|
||||
if (!section?.text?.trim()) {
|
||||
throw new Error("未找到该节正文");
|
||||
}
|
||||
const spoken = `【${section.title}】\n${section.text}`.trim();
|
||||
const result = await synthesizeSpeech({
|
||||
text: spoken,
|
||||
cacheKey: `${sessionId}:${id}`,
|
||||
});
|
||||
return { ...result, sectionId: id, title: section.title };
|
||||
}
|
||||
|
||||
private listDictateProducts(s: ManagedSession): DictateProduct[] {
|
||||
const board = s.runtime.getBlackboard();
|
||||
const products: DictateProduct[] = [];
|
||||
@@ -897,11 +1083,10 @@ export class SessionManager {
|
||||
if (dictateProductFamily(tag) === OPENING_PRODUCT_FAMILY) {
|
||||
const split = splitOpeningDocument(content);
|
||||
if (split.persona) {
|
||||
ensureProtagonistFromPersona(
|
||||
board,
|
||||
split.persona,
|
||||
protagonistTagForOpeningSlot(tag),
|
||||
);
|
||||
const personaTag = protagonistTagForOpeningSlot(tag);
|
||||
ensureProtagonistFromPersona(board, split.persona, personaTag, {
|
||||
overwrite: personaTag !== PROTAGONIST_TAG,
|
||||
});
|
||||
}
|
||||
// 仅有用户角色 meta、无叙事：抬到主角设定，不污染开场白
|
||||
if (split.persona && !split.text.trim()) {
|
||||
@@ -1009,6 +1194,15 @@ export class SessionManager {
|
||||
s.messages.length,
|
||||
this.captureCheckpoint(s),
|
||||
);
|
||||
const keep = normalizeContextTraceKeepLatest(
|
||||
loadAppSettings().contextTraceKeepLatest,
|
||||
);
|
||||
const requestOutlines = s.trackingRef.current.pendingOutlines;
|
||||
const pendingTrace = s.trackingRef.current.pendingContextTrace;
|
||||
const tokenUsage = s.trackingRef.current.pendingUsage;
|
||||
s.trackingRef.current.pendingOutlines = undefined;
|
||||
s.trackingRef.current.pendingContextTrace = undefined;
|
||||
s.trackingRef.current.pendingUsage = undefined;
|
||||
s.messages.push({
|
||||
id: randomUUID(),
|
||||
role: "system",
|
||||
@@ -1019,7 +1213,11 @@ export class SessionManager {
|
||||
title: "转述整理",
|
||||
body: result.reply,
|
||||
...(thinking ? { thinking } : {}),
|
||||
...(tokenUsage ? { tokenUsage } : {}),
|
||||
...(keep > 0 && pendingTrace ? { contextTrace: pendingTrace } : {}),
|
||||
...(requestOutlines?.length ? { requestOutlines } : {}),
|
||||
});
|
||||
s.messages = pruneContextTraces(s.messages, keep);
|
||||
s.runtime.enterDictateIntake();
|
||||
}
|
||||
|
||||
@@ -1494,6 +1692,7 @@ export class SessionManager {
|
||||
// 转述聊天常夹确认/建议，不能当开场白
|
||||
if (
|
||||
kind === "dictate_reply" ||
|
||||
kind === "listen_reply" ||
|
||||
kind === "agent_tool" ||
|
||||
kind === "orchestrator_thinking" ||
|
||||
kind === "worker_questions"
|
||||
@@ -1830,9 +2029,9 @@ export class SessionManager {
|
||||
);
|
||||
}
|
||||
|
||||
private playOpeningMessage(text: string): ChatMessage {
|
||||
private playOpeningMessage(text: string, keepId?: string): ChatMessage {
|
||||
return {
|
||||
id: randomUUID(),
|
||||
id: keepId || randomUUID(),
|
||||
role: "system",
|
||||
text,
|
||||
createdAt: new Date().toISOString(),
|
||||
@@ -1917,6 +2116,29 @@ export class SessionManager {
|
||||
return count;
|
||||
}
|
||||
|
||||
/**
|
||||
* 切换/改了全局用户角色：游玩层开场按新 @玩家 重展开，并改用全局角色卡。
|
||||
*/
|
||||
rebindPlayOpeningsForActivePersona(): number {
|
||||
let count = 0;
|
||||
for (const [sessionId, s] of this.sessions) {
|
||||
if (!isPlayLayerActive(s.runtime.getSession().slots)) continue;
|
||||
if (inferLifecycleStage(s.runtime.getSession()) !== "play") continue;
|
||||
const aux = buildPlayAuxView(
|
||||
s.runtime.getBlackboard(),
|
||||
s.runtime.getSession().slots,
|
||||
);
|
||||
if (!aux.openings.length) continue;
|
||||
try {
|
||||
this.selectPlayOpening(sessionId, aux.selectedOpeningIndex, false);
|
||||
count += 1;
|
||||
} catch {
|
||||
/* 单局失败不挡其它局 */
|
||||
}
|
||||
}
|
||||
return count;
|
||||
}
|
||||
|
||||
async sendMessage(
|
||||
id: string,
|
||||
text: string,
|
||||
@@ -1983,13 +2205,15 @@ export class SessionManager {
|
||||
const inPlay =
|
||||
isPlayLayerActive(sess.slots) ||
|
||||
inferLifecycleStage(sess) === "play";
|
||||
if (
|
||||
this.isDictateMode(s) &&
|
||||
reason?.kind === "input" &&
|
||||
!inPlay
|
||||
) {
|
||||
await this.runDictateTurnForSession(s, trimmed || text);
|
||||
return;
|
||||
if (reason?.kind === "input" && !inPlay) {
|
||||
if (this.isListenSerialMode(s)) {
|
||||
await this.runListenSerialTurnForSession(s, trimmed || text);
|
||||
return;
|
||||
}
|
||||
if (this.isDictateMode(s)) {
|
||||
await this.runDictateTurnForSession(s, trimmed || text);
|
||||
return;
|
||||
}
|
||||
}
|
||||
}
|
||||
if (reason?.kind === "approve_step") {
|
||||
@@ -2237,13 +2461,15 @@ export class SessionManager {
|
||||
const inPlay =
|
||||
isPlayLayerActive(sess.slots) ||
|
||||
inferLifecycleStage(sess) === "play";
|
||||
if (
|
||||
this.isDictateMode(s) &&
|
||||
reason?.kind === "input" &&
|
||||
!inPlay
|
||||
) {
|
||||
await this.runDictateTurnForSession(s, trimmed);
|
||||
return;
|
||||
if (reason?.kind === "input" && !inPlay) {
|
||||
if (this.isListenSerialMode(s)) {
|
||||
await this.runListenSerialTurnForSession(s, trimmed);
|
||||
return;
|
||||
}
|
||||
if (this.isDictateMode(s)) {
|
||||
await this.runDictateTurnForSession(s, trimmed);
|
||||
return;
|
||||
}
|
||||
}
|
||||
await s.runtime.submitInput(trimmed);
|
||||
});
|
||||
@@ -3100,6 +3326,7 @@ export class SessionManager {
|
||||
undefined;
|
||||
const pendingTrace = managed.trackingRef.current.pendingContextTrace;
|
||||
let contextTrace: LlmContextTrace | undefined;
|
||||
let requestOutlines: RequestOutline[] | undefined;
|
||||
if (
|
||||
pendingTrace &&
|
||||
classified &&
|
||||
@@ -3112,10 +3339,19 @@ export class SessionManager {
|
||||
contextTrace = pendingTrace;
|
||||
}
|
||||
managed.trackingRef.current.pendingContextTrace = undefined;
|
||||
requestOutlines = managed.trackingRef.current.pendingOutlines;
|
||||
managed.trackingRef.current.pendingOutlines = undefined;
|
||||
}
|
||||
|
||||
managed.messages.push(
|
||||
this.msg("system", text, usage, thinking, contextTrace),
|
||||
this.msg(
|
||||
"system",
|
||||
text,
|
||||
usage,
|
||||
thinking,
|
||||
contextTrace,
|
||||
requestOutlines?.length ? requestOutlines : undefined,
|
||||
),
|
||||
);
|
||||
managed.messages = pruneContextTraces(
|
||||
managed.messages,
|
||||
@@ -3405,6 +3641,7 @@ export class SessionManager {
|
||||
tokenUsage?: MessageTokenUsage,
|
||||
thinking?: string,
|
||||
contextTrace?: LlmContextTrace,
|
||||
requestOutlines?: RequestOutline[],
|
||||
): ChatMessage {
|
||||
const base = {
|
||||
id: randomUUID(),
|
||||
@@ -3414,6 +3651,7 @@ export class SessionManager {
|
||||
...(tokenUsage ? { tokenUsage } : {}),
|
||||
...(thinking ? { thinking } : {}),
|
||||
...(contextTrace ? { contextTrace } : {}),
|
||||
...(requestOutlines?.length ? { requestOutlines } : {}),
|
||||
};
|
||||
if (role === "user") {
|
||||
return {
|
||||
@@ -3812,10 +4050,12 @@ export class SessionManager {
|
||||
lifecycleStage,
|
||||
{ workerSetYaml },
|
||||
);
|
||||
const shellAdapt = parseShellAdaptationFromReplyFormat(
|
||||
const replyFormatRaw =
|
||||
s.runtime.getBlackboard().getContentByTag("设计.正文组成") ??
|
||||
s.runtime.getBlackboard().getContentByTag("设计.回复格式"),
|
||||
);
|
||||
s.runtime.getBlackboard().getContentByTag("设计.0层正文组成") ??
|
||||
s.runtime.getBlackboard().getContentByTag("设计.回复格式");
|
||||
const shellAdapt = parseShellAdaptationFromReplyFormat(replyFormatRaw);
|
||||
const replyContract = parseReplyDataContract(replyFormatRaw);
|
||||
|
||||
return {
|
||||
id,
|
||||
@@ -3831,10 +4071,13 @@ export class SessionManager {
|
||||
actions,
|
||||
pipeline: buildPipeline(session),
|
||||
lifecycleStage,
|
||||
creationMode: this.isDictateMode(s) ? "dictate" : "recipe",
|
||||
creationMode: this.creationIntakeMode(s),
|
||||
dictateProducts: this.isDictateMode(s)
|
||||
? this.listDictateProducts(s)
|
||||
: undefined,
|
||||
listenSerial: this.isListenSerialMode(s)
|
||||
? toListenView(readListenState(s.runtime.getBlackboard()))
|
||||
: undefined,
|
||||
playLayerActive: isPlayLayerActive(session.slots),
|
||||
playReady: canEnterPlay(session),
|
||||
hasProduct: Boolean(
|
||||
@@ -3895,6 +4138,27 @@ export class SessionManager {
|
||||
...shellAdapt.tweaks,
|
||||
}
|
||||
: undefined,
|
||||
presentationFrontend: replyContract?.frontend,
|
||||
presentationSlots: replyContract
|
||||
? [...replyContract.modelBlocks, ...replyContract.programBlocks].map((block) => ({
|
||||
id: block.id,
|
||||
type: block.type,
|
||||
}))
|
||||
: undefined,
|
||||
presentationProjection: (() => {
|
||||
if (!replyContract?.programBlocks.length) return undefined;
|
||||
const board = s.runtime.getBlackboard();
|
||||
const projected = projectPresentBlocks({
|
||||
contract: replyContract,
|
||||
statusBarRaw:
|
||||
board.getContentByTag("设计.监控栏") ??
|
||||
board.getContentByTag("设计.回复呈现"),
|
||||
currentVarsRaw: board.getContentByTag("变量.当前"),
|
||||
topologyRaw: board.getContentByTag("运行.拓扑"),
|
||||
readTag: (tag) => board.getContentByTag(tag),
|
||||
});
|
||||
return Object.keys(projected).length ? projected : undefined;
|
||||
})(),
|
||||
liveStream: this.buildLiveStreamView(id),
|
||||
agentThinking: this.agentThinkingLive.get(id),
|
||||
runActive: Boolean(s.runAbort),
|
||||
|
||||
@@ -613,8 +613,13 @@ export async function handleSettingsApi(
|
||||
activate: body.activate,
|
||||
creationDefault: body.creationDefault,
|
||||
});
|
||||
const reboundSessions =
|
||||
body.activate === false
|
||||
? 0
|
||||
: sessionManager.rebindPlayOpeningsForActivePersona();
|
||||
json(res, 201, {
|
||||
...personasPayload({ persona }),
|
||||
reboundSessions,
|
||||
});
|
||||
} catch (e) {
|
||||
json(res, 400, {
|
||||
@@ -634,8 +639,11 @@ export async function handleSettingsApi(
|
||||
if (activate && req.method === "POST") {
|
||||
try {
|
||||
const persona = setActivePersonaId(id);
|
||||
const reboundSessions =
|
||||
sessionManager.rebindPlayOpeningsForActivePersona();
|
||||
json(res, 200, {
|
||||
...personasPayload({ persona, activeId: persona.id }),
|
||||
reboundSessions,
|
||||
});
|
||||
} catch (e) {
|
||||
json(res, 404, {
|
||||
@@ -653,8 +661,13 @@ export async function handleSettingsApi(
|
||||
creationDefault?: boolean;
|
||||
};
|
||||
const persona = updatePersona(id, body);
|
||||
const reboundSessions =
|
||||
getActivePersona()?.id === persona.id
|
||||
? sessionManager.rebindPlayOpeningsForActivePersona()
|
||||
: 0;
|
||||
json(res, 200, {
|
||||
...personasPayload({ persona }),
|
||||
reboundSessions,
|
||||
});
|
||||
} catch (e) {
|
||||
const msg = e instanceof Error ? e.message : String(e);
|
||||
@@ -666,8 +679,11 @@ export async function handleSettingsApi(
|
||||
if (req.method === "DELETE") {
|
||||
try {
|
||||
deletePersona(id);
|
||||
const reboundSessions =
|
||||
sessionManager.rebindPlayOpeningsForActivePersona();
|
||||
json(res, 200, {
|
||||
...personasPayload({ ok: true }),
|
||||
reboundSessions,
|
||||
});
|
||||
} catch (e) {
|
||||
const msg = e instanceof Error ? e.message : String(e);
|
||||
|
||||
@@ -14,6 +14,7 @@ import {
|
||||
SUGGESTED_PRICING_DOCS,
|
||||
} from "../stats/pricing.js";
|
||||
import { attachRecordCost, buildCostReport } from "../stats/pricing-cost.js";
|
||||
import { annotateCacheBreaks } from "../llm/context-outline.js";
|
||||
|
||||
async function readBody(req: IncomingMessage): Promise<string> {
|
||||
const chunks: Buffer[] = [];
|
||||
@@ -77,6 +78,7 @@ export async function handleStatsApi(
|
||||
const records = readRecords({ ...query, limit: query.limit ?? 100 });
|
||||
const catalog = resolveRateCatalog();
|
||||
const cost = buildCostReport(summaryRecords, catalog);
|
||||
const breaks = annotateCacheBreaks(summaryRecords);
|
||||
json(res, 200, {
|
||||
summary: {
|
||||
...summarizeRecords(summaryRecords),
|
||||
@@ -92,6 +94,7 @@ export async function handleStatsApi(
|
||||
...r,
|
||||
estimatedCost: priced.cost,
|
||||
unmatchedRate: priced.unmatched,
|
||||
cacheBreak: breaks.get(r.id) ?? null,
|
||||
};
|
||||
}),
|
||||
});
|
||||
|
||||
@@ -230,6 +230,54 @@ const server = createServer(async (req, res) => {
|
||||
return;
|
||||
}
|
||||
|
||||
if (req.method === "POST" && sub === "/listen/consume") {
|
||||
const body = JSON.parse(await readBody(req)) as {
|
||||
consumedChars?: number;
|
||||
};
|
||||
const n = Number(body.consumedChars);
|
||||
if (!Number.isFinite(n) || n < 0) {
|
||||
json(res, 400, { error: "需要 consumedChars ≥ 0" });
|
||||
return;
|
||||
}
|
||||
try {
|
||||
const view = sessionManager.markListenSerialConsumed(sessionId, n);
|
||||
json(res, 200, view);
|
||||
} catch (err) {
|
||||
json(res, 400, {
|
||||
error: err instanceof Error ? err.message : "进度更新失败",
|
||||
});
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
if (req.method === "GET" && sub === "/listen/tts") {
|
||||
const sectionId = (url.searchParams.get("sectionId") ?? "").trim();
|
||||
if (!sectionId) {
|
||||
json(res, 400, { error: "需要 sectionId" });
|
||||
return;
|
||||
}
|
||||
try {
|
||||
const audio = await sessionManager.synthesizeListenSection(
|
||||
sessionId,
|
||||
sectionId,
|
||||
);
|
||||
res.writeHead(200, {
|
||||
"Content-Type": audio.contentType,
|
||||
"Content-Length": audio.audio.length,
|
||||
"Cache-Control": audio.cached ? "private, max-age=86400" : "no-store",
|
||||
"X-TTS-Provider": audio.provider,
|
||||
"X-TTS-Voice": audio.voice,
|
||||
"X-TTS-Cached": audio.cached ? "1" : "0",
|
||||
});
|
||||
res.end(audio.audio);
|
||||
} catch (err) {
|
||||
json(res, 400, {
|
||||
error: err instanceof Error ? err.message : "语音合成失败",
|
||||
});
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
if (req.method === "POST" && sub === "/answers") {
|
||||
const body = JSON.parse(await readBody(req)) as {
|
||||
answers?: Array<{
|
||||
|
||||
@@ -1423,6 +1423,14 @@ const DICTATE_PROGRAM_SIDE_MODULE_IDS = new Set([
|
||||
"refine",
|
||||
]);
|
||||
|
||||
/** 落档程序或目录标了 auto 的节点：创作对话不 insert */
|
||||
export function isDictateProgramModule(module: {
|
||||
id: string;
|
||||
auto?: boolean;
|
||||
}): boolean {
|
||||
return Boolean(module.auto) || DICTATE_PROGRAM_SIDE_MODULE_IDS.has(module.id);
|
||||
}
|
||||
|
||||
/**
|
||||
* 注入对话落盘：能力何时落盘（meta when/when_not）+ insert 目标 tag。
|
||||
* 不写 DAG / design-step 话术；不塞执行全文。
|
||||
@@ -1430,7 +1438,7 @@ const DICTATE_PROGRAM_SIDE_MODULE_IDS = new Set([
|
||||
export function formatModuleCatalogForDictate(catalog: ModuleCatalog): string {
|
||||
const dictateModules = catalogModulesForIntake(catalog, "dictate");
|
||||
const lines = dictateModules.map((m) => {
|
||||
const programSide = DICTATE_PROGRAM_SIDE_MODULE_IDS.has(m.id) || m.auto;
|
||||
const programSide = isDictateProgramModule(m);
|
||||
const flags = [
|
||||
m.repeatable ? "〔可反复〕" : "",
|
||||
m.closer ? "〔收口〕" : "",
|
||||
@@ -1439,7 +1447,7 @@ export function formatModuleCatalogForDictate(catalog: ModuleCatalog): string {
|
||||
]
|
||||
.filter(Boolean)
|
||||
.join("");
|
||||
const land = dictateLandHint(m);
|
||||
const land = dictateModuleLandHint(m);
|
||||
const parts: string[] = [
|
||||
`- ${m.name}${flags}：${m.declaration}`,
|
||||
` 落盘：${land}`,
|
||||
@@ -1455,7 +1463,7 @@ export function formatModuleCatalogForDictate(catalog: ModuleCatalog): string {
|
||||
"「必须」和「有必要」里用户还没提到的：本轮不落盘，按已有材料给 2～3 个选项询问，选定后再写。含糊或未到时机的下游同样先问，不先写。",
|
||||
"同轮写入的每份产物各回 reply_module，可见回复综合成一次。",
|
||||
"标〔落档程序〕的：创作对话一般不 insert，开玩/落档时由程序处理。",
|
||||
"标〔收口〕的开场：写开场前，列出仍未落的「必须」和「有必要」，问用户这些还没设计要补哪几项或明确不要。用户这句已经说清的再写入。普通模式使用回复呈现/正文组成，0 层模式使用对应 0 层节点，不得混用。用户确认可以收口后，程序依次编译信息可见范围、游玩执行方式、上下文排序与运行配置，再写所选模式的开场产物。",
|
||||
"标〔收口〕的开场：写开场前，列出仍未落的「必须」和「有必要」，问用户这些还没设计要补哪几项或明确不要。用户这句已经说清的再写入。普通模式使用回复呈现/正文组成，0 层模式使用对应 0 层节点，不得混用。用户确认可以收口后，程序依次按有无变量真值固定开旁观填表、编译信息可见范围、游玩执行方式、上下文排序与运行配置，再写所选模式的开场产物。",
|
||||
"标〔可反复〕的（生成规则、具体实例、开场白、拓扑等）：每条单独落盘，用「基名#唯一id」（生成规则用规则名，具体实例用名称，开场白用场景短码）。同完整 tag 才覆盖，禁止反复写基名盖掉别的条。开场全部保留；游玩时左右切换，选中的一条作为 0 层。",
|
||||
"能力执行全文不进聊天；规格只走 toolcall。insert 成功后按返回的 reply_module 写可见回复（见系统提示「对用户可见回复」）。",
|
||||
lines.join("\n"),
|
||||
@@ -1464,12 +1472,15 @@ export function formatModuleCatalogForDictate(catalog: ModuleCatalog): string {
|
||||
return head;
|
||||
}
|
||||
|
||||
function dictateLandHint(m: ModuleCatalogEntry): string {
|
||||
if (DICTATE_PROGRAM_SIDE_MODULE_IDS.has(m.id) || m.auto) {
|
||||
export function dictateModuleLandHint(m: ModuleCatalogEntry): string {
|
||||
if (isDictateProgramModule(m)) {
|
||||
if (m.id === "worker-spec") {
|
||||
return "落档/开玩程序处理：有变量真值时固定开旁观填表，创作期一般不 insert";
|
||||
}
|
||||
return "落档/开玩程序处理；创作期一般不 insert";
|
||||
}
|
||||
if (m.id === "variable-design") {
|
||||
return "declare_variable / declare_map；长文规则可另 insert 「设计.变量设计与更新规则」";
|
||||
return "insert 「设计.变量设计与更新规则」（有真值时程序固定旁观副 LLM 填表，并同步目录与真表）；或用 declare_variable / declare_map 逐条钉目录";
|
||||
}
|
||||
if (m.id === "opening-setup") {
|
||||
return "insert 「设计.开场白#场景短码」——可增殖：每条开场各占一 tag；同完整 tag 才覆盖。游玩时左右切换，选中的一条作为 0 层";
|
||||
|
||||
@@ -0,0 +1,199 @@
|
||||
/**
|
||||
* 有变量真值时固定旁观副 LLM 填表（auditor_bundle）。
|
||||
* 旁观上下文刻意保持小：上一份表 + 创作期维护准则 + 本轮主世界正文；
|
||||
* 不再按维护复杂度.建议在内联 / 旁观间切换，避免主世界 tool / state_ops 反复探测。
|
||||
* 变量设计里的维护复杂度可作旁观备注，不驱动开槽。
|
||||
*/
|
||||
import type { Blackboard } from "../blackboard/blackboard.js";
|
||||
import { extractJsonObjectText } from "./worker-set-parse.js";
|
||||
import type { PlaySlotsConfig } from "./play-slots.js";
|
||||
import { VARIABLE_DESIGN_TAG } from "./variable-design-compile.js";
|
||||
|
||||
export const MAINTENANCE_STRATEGY_SCHEMA = "maintenance-strategy.v1" as const;
|
||||
export const MAINTENANCE_STRATEGY_TAG = "设计.维护策略";
|
||||
|
||||
export type MaintenanceAdvice =
|
||||
| "none"
|
||||
| "inline_tool"
|
||||
| "inline_structured"
|
||||
| "auditor_bundle";
|
||||
|
||||
export type MaintenanceStrategyDoc = {
|
||||
schema: typeof MAINTENANCE_STRATEGY_SCHEMA;
|
||||
advice: MaintenanceAdvice;
|
||||
/** 是否启用旁观维护槽 */
|
||||
auditor: boolean;
|
||||
/** 是否走主世界内联维护（已弃用开槽；固定旁观后恒为 false） */
|
||||
inline: boolean;
|
||||
source: string;
|
||||
依据?: string[];
|
||||
};
|
||||
|
||||
function asString(v: unknown): string | undefined {
|
||||
if (typeof v !== "string") return undefined;
|
||||
const t = v.trim();
|
||||
return t || undefined;
|
||||
}
|
||||
|
||||
function asObject(v: unknown): Record<string, unknown> | undefined {
|
||||
if (!v || typeof v !== "object" || Array.isArray(v)) return undefined;
|
||||
return v as Record<string, unknown>;
|
||||
}
|
||||
|
||||
export function normalizeMaintenanceAdvice(raw: unknown): MaintenanceAdvice {
|
||||
const s = asString(raw)?.toLowerCase() ?? "";
|
||||
if (
|
||||
s === "auditor_bundle" ||
|
||||
s === "auditor" ||
|
||||
s === "旁观" ||
|
||||
s === "旁观维护" ||
|
||||
s === "bundle"
|
||||
) {
|
||||
return "auditor_bundle";
|
||||
}
|
||||
if (
|
||||
s === "inline_structured" ||
|
||||
s === "structured" ||
|
||||
s === "内联结构化" ||
|
||||
s === "state_ops"
|
||||
) {
|
||||
return "inline_structured";
|
||||
}
|
||||
if (
|
||||
s === "inline_tool" ||
|
||||
s === "inline" ||
|
||||
s === "tool" ||
|
||||
s === "内联" ||
|
||||
s === "内联工具"
|
||||
) {
|
||||
return "inline_tool";
|
||||
}
|
||||
if (s === "none" || s === "无" || s === "不维护") return "none";
|
||||
return "none";
|
||||
}
|
||||
|
||||
function designHasTruths(raw: string | null | undefined): boolean {
|
||||
if (!raw?.trim()) return false;
|
||||
const text = extractJsonObjectText(raw) ?? raw.trim();
|
||||
let doc: unknown;
|
||||
try {
|
||||
doc = JSON.parse(text);
|
||||
} catch {
|
||||
return false;
|
||||
}
|
||||
const root = asObject(doc);
|
||||
if (!root) return false;
|
||||
const body = asObject(root.正文) ?? root;
|
||||
const truths = body.真值 ?? body.fields ?? body.变量;
|
||||
return Array.isArray(truths) && truths.length > 0;
|
||||
}
|
||||
|
||||
/** 解析变量设计产物中的维护复杂度（仅作备注；开槽不读此建议） */
|
||||
export function parseMaintenanceAdviceFromVariableDesign(
|
||||
raw: string | null | undefined,
|
||||
): { advice: MaintenanceAdvice; 依据: string[] } | null {
|
||||
if (!raw?.trim()) return null;
|
||||
const text = extractJsonObjectText(raw) ?? raw.trim();
|
||||
let doc: unknown;
|
||||
try {
|
||||
doc = JSON.parse(text);
|
||||
} catch {
|
||||
return null;
|
||||
}
|
||||
const root = asObject(doc);
|
||||
if (!root) return null;
|
||||
const body = asObject(root.正文) ?? root;
|
||||
const box = asObject(body.维护复杂度) ?? asObject(body.maintenance);
|
||||
if (!box) {
|
||||
if (designHasTruths(raw)) {
|
||||
return {
|
||||
advice: "auditor_bundle",
|
||||
依据: ["变量设计有真值；程序固定旁观副 LLM 填表"],
|
||||
};
|
||||
}
|
||||
return { advice: "none", 依据: [] };
|
||||
}
|
||||
const advice = normalizeMaintenanceAdvice(box.建议 ?? box.advice ?? box.strategy);
|
||||
const reasonsRaw = box.依据 ?? box.reasons ?? box.why;
|
||||
const 依据 = Array.isArray(reasonsRaw)
|
||||
? reasonsRaw.map((r) => asString(r)).filter((r): r is string => Boolean(r))
|
||||
: asString(reasonsRaw)
|
||||
? [asString(reasonsRaw)!]
|
||||
: [];
|
||||
return { advice, 依据 };
|
||||
}
|
||||
|
||||
export function strategyFromAdvice(
|
||||
advice: MaintenanceAdvice,
|
||||
依据: string[] = [],
|
||||
source = VARIABLE_DESIGN_TAG,
|
||||
): MaintenanceStrategyDoc {
|
||||
// 开槽层：仅 auditor_bundle 开旁观；inline_* 在 resolve 阶段会被抬成旁观
|
||||
const auditor = advice === "auditor_bundle";
|
||||
const inline = advice === "inline_tool" || advice === "inline_structured";
|
||||
return {
|
||||
schema: MAINTENANCE_STRATEGY_SCHEMA,
|
||||
advice,
|
||||
auditor,
|
||||
inline,
|
||||
source,
|
||||
...(依据.length ? { 依据 } : {}),
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* 有真值 → 固定旁观填表；无真值 → 不启旁观。
|
||||
* 设计里的 inline_* 建议不关闭旁观，只记入依据。
|
||||
*/
|
||||
export function resolveMaintenanceStrategyFromBoard(
|
||||
board: Blackboard,
|
||||
): MaintenanceStrategyDoc {
|
||||
const designRaw = board.getContentByTag(VARIABLE_DESIGN_TAG);
|
||||
if (!designHasTruths(designRaw)) {
|
||||
return strategyFromAdvice(
|
||||
"none",
|
||||
["无变量真值，不启旁观填表"],
|
||||
VARIABLE_DESIGN_TAG,
|
||||
);
|
||||
}
|
||||
const parsed = parseMaintenanceAdviceFromVariableDesign(designRaw);
|
||||
const 依据 = [
|
||||
"程序固定旁观副 LLM 填表（上一份表 + 维护准则 + 本轮主世界正文）",
|
||||
...(parsed?.依据 ?? []),
|
||||
];
|
||||
if (
|
||||
parsed &&
|
||||
parsed.advice !== "auditor_bundle" &&
|
||||
parsed.advice !== "none"
|
||||
) {
|
||||
依据.push(`设计原建议 ${parsed.advice}，已忽略内联开槽`);
|
||||
}
|
||||
return strategyFromAdvice("auditor_bundle", 依据, VARIABLE_DESIGN_TAG);
|
||||
}
|
||||
|
||||
/** 按维护策略覆盖 play_slots.auditor（不改 gm/narrator 等） */
|
||||
export function applyMaintenanceStrategyToPlaySlots(
|
||||
slots: PlaySlotsConfig,
|
||||
strategy: MaintenanceStrategyDoc,
|
||||
): PlaySlotsConfig {
|
||||
return {
|
||||
...slots,
|
||||
auditor: strategy.auditor,
|
||||
};
|
||||
}
|
||||
|
||||
export function serializeMaintenanceStrategy(doc: MaintenanceStrategyDoc): string {
|
||||
return JSON.stringify(doc, null, 2);
|
||||
}
|
||||
|
||||
export function writeMaintenanceStrategy(
|
||||
board: Blackboard,
|
||||
strategy: MaintenanceStrategyDoc,
|
||||
source = "system:maintenance-strategy",
|
||||
): void {
|
||||
board.write({
|
||||
tag: MAINTENANCE_STRATEGY_TAG,
|
||||
content: serializeMaintenanceStrategy(strategy),
|
||||
source,
|
||||
});
|
||||
}
|
||||
@@ -347,13 +347,18 @@ export function parseOpeningUserPersona(
|
||||
if (!text) return null;
|
||||
try {
|
||||
const doc = JSON.parse(text) as unknown;
|
||||
const row =
|
||||
doc && typeof doc === "object" && !Array.isArray(doc)
|
||||
? (doc as Record<string, unknown>)
|
||||
: undefined;
|
||||
const meta =
|
||||
row?.meta && typeof row.meta === "object" && !Array.isArray(row.meta)
|
||||
? (row.meta as Record<string, unknown>)
|
||||
: undefined;
|
||||
const fromJson =
|
||||
parseUserPersonaField(doc) ||
|
||||
parseUserPersonaField(
|
||||
doc && typeof doc === "object" && !Array.isArray(doc)
|
||||
? (doc as Record<string, unknown>).用户角色
|
||||
: undefined,
|
||||
);
|
||||
parseUserPersonaField(row?.用户角色) ||
|
||||
parseUserPersonaField(meta?.用户角色);
|
||||
if (fromJson) return fromJson;
|
||||
} catch {
|
||||
/* not json */
|
||||
|
||||
+63
-31
@@ -45,7 +45,10 @@ export type PlayAuxOpening = {
|
||||
/** 产物 tag：设计.开场白 或 设计.开场白#槽位；opening-setup 候选为假想 tag */
|
||||
tag: string;
|
||||
slot: string;
|
||||
/** 侧栏预览用：present.v1 已压成可读正文 */
|
||||
text: string;
|
||||
/** 选用时落盘 / 灌呈现壳：保留完整 present.v1 或散文原文 */
|
||||
document: string;
|
||||
persona: OpeningUserPersona | null;
|
||||
};
|
||||
|
||||
@@ -67,34 +70,15 @@ function selectedOpeningIndex(slots: Record<string, unknown> | undefined): numbe
|
||||
return 0;
|
||||
}
|
||||
|
||||
/** 黑板上全部可选用开场（工序候选优先；否则对话落盘 设计.开场白[#槽]） */
|
||||
export function collectOpenings(board: Blackboard): PlayAuxOpening[] {
|
||||
const payload = parseOpeningSealPayload(
|
||||
board.getContentByTag(OPENING_SETUP_ARTIFACT_TAG),
|
||||
0,
|
||||
);
|
||||
if (payload?.candidates.length) {
|
||||
return payload.candidates.map((text, index) => {
|
||||
const one = parseOpeningSealPayload(
|
||||
board.getContentByTag(OPENING_SETUP_ARTIFACT_TAG),
|
||||
index,
|
||||
);
|
||||
const tag = `${OPENING_SETUP_ARTIFACT_TAG}#${index + 1}`;
|
||||
return {
|
||||
index,
|
||||
tag,
|
||||
slot: String(index + 1),
|
||||
text: presentReadablePlain(text) || text,
|
||||
persona: one?.persona ?? payload.persona,
|
||||
};
|
||||
});
|
||||
}
|
||||
type OpeningProductRow = {
|
||||
tag: string;
|
||||
content: string;
|
||||
order: number;
|
||||
slot: string;
|
||||
};
|
||||
|
||||
const rows: Array<{
|
||||
tag: string;
|
||||
content: string;
|
||||
order: number;
|
||||
}> = [];
|
||||
function listOpeningProductRows(board: Blackboard): OpeningProductRow[] {
|
||||
const rows: OpeningProductRow[] = [];
|
||||
for (const e of board.listTagIndex()) {
|
||||
const family = dictateProductFamily(e.tag);
|
||||
if (
|
||||
@@ -111,26 +95,74 @@ export function collectOpenings(board: Blackboard): PlayAuxOpening[] {
|
||||
typeof rawOrder === "number" && Number.isFinite(rawOrder)
|
||||
? rawOrder
|
||||
: effectiveDictateOrder({ tag: e.tag });
|
||||
rows.push({ tag: e.tag, content, order });
|
||||
rows.push({
|
||||
tag: e.tag,
|
||||
content,
|
||||
order,
|
||||
slot: dictateProductSlot(e.tag),
|
||||
});
|
||||
}
|
||||
rows.sort((a, b) => {
|
||||
if (a.order !== b.order) return a.order - b.order;
|
||||
return a.tag.localeCompare(b.tag, "zh");
|
||||
});
|
||||
return rows;
|
||||
}
|
||||
|
||||
function openingsFromRows(
|
||||
board: Blackboard,
|
||||
rows: OpeningProductRow[],
|
||||
): PlayAuxOpening[] {
|
||||
return rows.map((row, index) => {
|
||||
const split = splitOpeningDocument(row.content);
|
||||
const text = split.text.trim() || row.content;
|
||||
const document = split.text.trim() || row.content;
|
||||
return {
|
||||
index,
|
||||
tag: row.tag,
|
||||
slot: dictateProductSlot(row.tag),
|
||||
text: presentReadablePlain(text) || text,
|
||||
slot: row.slot,
|
||||
document,
|
||||
text: presentReadablePlain(document) || document,
|
||||
persona: personaBoundToOpening(board, row.tag, row.content),
|
||||
};
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* 可切换开场。已有 `设计.开场白#场景短码` 时只认这些带槽位的卡，
|
||||
* 裸 `设计.开场白` 与整包 `设计.开场白与开场变量` 不进切换列表。
|
||||
* 还没有槽位卡时，才回退旧整包候选或裸基名。
|
||||
*/
|
||||
export function collectOpenings(board: Blackboard): PlayAuxOpening[] {
|
||||
const rows = listOpeningProductRows(board);
|
||||
const slotted = rows.filter((row) => row.slot);
|
||||
if (slotted.length) return openingsFromRows(board, slotted);
|
||||
|
||||
const payload = parseOpeningSealPayload(
|
||||
board.getContentByTag(OPENING_SETUP_ARTIFACT_TAG),
|
||||
0,
|
||||
);
|
||||
if (payload?.candidates.length) {
|
||||
return payload.candidates.map((text, index) => {
|
||||
const one = parseOpeningSealPayload(
|
||||
board.getContentByTag(OPENING_SETUP_ARTIFACT_TAG),
|
||||
index,
|
||||
);
|
||||
const tag = `${OPENING_SETUP_ARTIFACT_TAG}#${index + 1}`;
|
||||
const document = text.trim();
|
||||
return {
|
||||
index,
|
||||
tag,
|
||||
slot: String(index + 1),
|
||||
document,
|
||||
text: presentReadablePlain(document) || document,
|
||||
persona: one?.persona ?? payload.persona,
|
||||
};
|
||||
});
|
||||
}
|
||||
|
||||
return openingsFromRows(board, rows);
|
||||
}
|
||||
|
||||
export function buildPlayAuxView(
|
||||
board: Blackboard,
|
||||
slots?: Record<string, unknown>,
|
||||
|
||||
@@ -40,9 +40,11 @@ export type ShellAdaptation = {
|
||||
tweaks: PresentShellTweaks;
|
||||
};
|
||||
|
||||
export type PresentBlocks = Partial<
|
||||
Record<PresentRegionId, string | Record<string, unknown> | unknown[]>
|
||||
>;
|
||||
export type PresentBlockValue = string | Record<string, unknown> | unknown[];
|
||||
|
||||
/** 已知壳区域以外的键来自本局正文组成数据契约，必须原样保留。 */
|
||||
export type PresentBlocks = Partial<Record<PresentRegionId, PresentBlockValue>> &
|
||||
Record<string, PresentBlockValue | undefined>;
|
||||
|
||||
export type PresentPacket = {
|
||||
schema: typeof PRESENT_SCHEMA;
|
||||
@@ -112,25 +114,29 @@ export function normalizePresentShellId(
|
||||
return fallback;
|
||||
}
|
||||
|
||||
function asPresentBlockValue(value: unknown): PresentBlockValue | undefined {
|
||||
if (typeof value === "string") return value;
|
||||
if (Array.isArray(value)) return value;
|
||||
if (value && typeof value === "object") return value as Record<string, unknown>;
|
||||
return undefined;
|
||||
}
|
||||
|
||||
function normalizeBlocks(raw: unknown): PresentBlocks {
|
||||
if (!raw || typeof raw !== "object" || Array.isArray(raw)) return {};
|
||||
const row = raw as Record<string, unknown>;
|
||||
const out: PresentBlocks = {};
|
||||
const keys: PresentRegionId[] = [
|
||||
"monitor",
|
||||
"header",
|
||||
"body",
|
||||
"footer",
|
||||
"aside",
|
||||
"hidden",
|
||||
];
|
||||
for (const k of keys) {
|
||||
if (row[k] !== undefined) out[k] = row[k] as PresentBlocks[typeof k];
|
||||
for (const [key, value] of Object.entries(row)) {
|
||||
if (key === "正文" || key === "监控") continue;
|
||||
const block = asPresentBlockValue(value);
|
||||
if (block !== undefined) out[key] = block;
|
||||
}
|
||||
if (out.body === undefined && row.正文 !== undefined) {
|
||||
const body = asPresentBlockValue(row.正文);
|
||||
if (body !== undefined) out.body = body;
|
||||
}
|
||||
// 兼容中文键
|
||||
if (out.body === undefined && row.正文 !== undefined) out.body = row.正文 as string;
|
||||
if (out.monitor === undefined && row.监控 !== undefined) {
|
||||
out.monitor = row.监控 as PresentBlocks["monitor"];
|
||||
const monitor = asPresentBlockValue(row.监控);
|
||||
if (monitor !== undefined) out.monitor = monitor;
|
||||
}
|
||||
if (typeof out.body === "string") {
|
||||
out.body = stripPresentSourceFences(out.body);
|
||||
|
||||
@@ -0,0 +1,351 @@
|
||||
/**
|
||||
* 程序投影：把 变量.当前 / 拓扑等权威数据灌进 present.v1 的 program_projection 区。
|
||||
* 模型只填 model_generated；本模块在写盘与视图层补齐投影块。
|
||||
*/
|
||||
import { parseTableDoc } from "../blackboard/table-cells.js";
|
||||
import { tableValueMap } from "../blackboard/table-side-effects.js";
|
||||
import { extractJsonObjectText } from "./worker-set-parse.js";
|
||||
import type { ReplyContractBlock, ReplyDataContract } from "./reply-contract.js";
|
||||
import {
|
||||
parsePresentPacket,
|
||||
stringifyPresentPacket,
|
||||
type PresentBlockValue,
|
||||
type PresentShellId,
|
||||
} from "./present-packet.js";
|
||||
|
||||
export type ProjectionBinding = {
|
||||
id: string;
|
||||
type: string;
|
||||
ref?: string;
|
||||
fields?: string[];
|
||||
};
|
||||
|
||||
function asString(v: unknown): string | undefined {
|
||||
if (typeof v !== "string") return undefined;
|
||||
const t = v.trim();
|
||||
return t || undefined;
|
||||
}
|
||||
|
||||
function asObject(v: unknown): Record<string, unknown> | undefined {
|
||||
if (!v || typeof v !== "object" || Array.isArray(v)) return undefined;
|
||||
return v as Record<string, unknown>;
|
||||
}
|
||||
|
||||
/** 从回复呈现（设计.监控栏）取出程序投影绑定 */
|
||||
export function parseStatusBarProjectionBindings(
|
||||
raw: string | null | undefined,
|
||||
): ProjectionBinding[] {
|
||||
if (!raw?.trim()) return [];
|
||||
const text = extractJsonObjectText(raw) ?? raw.trim();
|
||||
let doc: unknown;
|
||||
try {
|
||||
doc = JSON.parse(text);
|
||||
} catch {
|
||||
return [];
|
||||
}
|
||||
const root = asObject(doc);
|
||||
if (!root) return [];
|
||||
const body = asObject(root.正文) ?? root;
|
||||
const regions = body.语义区域 ?? body.regions ?? body.areas;
|
||||
if (!Array.isArray(regions)) return [];
|
||||
const out: ProjectionBinding[] = [];
|
||||
for (const item of regions) {
|
||||
const row = asObject(item);
|
||||
if (!row) continue;
|
||||
const id = asString(row.id);
|
||||
if (!id) continue;
|
||||
const source = asObject(row.来源) ?? asObject(row.source);
|
||||
const mode =
|
||||
asString(source?.方式) ??
|
||||
asString(source?.mode) ??
|
||||
asString(source?.source) ??
|
||||
asString(row.来源);
|
||||
if (
|
||||
mode &&
|
||||
mode !== "program_projection" &&
|
||||
mode !== "程序投影" &&
|
||||
mode !== "projection"
|
||||
) {
|
||||
continue;
|
||||
}
|
||||
// 无方式时：若标了 ref 也当作可投影
|
||||
const ref =
|
||||
asString(source?.ref) ??
|
||||
asString(source?.tag) ??
|
||||
asString(row.ref) ??
|
||||
asString(row.权威ref);
|
||||
const fieldsRaw = source?.字段 ?? source?.fields ?? row.字段 ?? row.fields;
|
||||
const fields = Array.isArray(fieldsRaw)
|
||||
? fieldsRaw
|
||||
.map((f) => asString(f))
|
||||
.filter((f): f is string => Boolean(f))
|
||||
: undefined;
|
||||
out.push({
|
||||
id,
|
||||
type:
|
||||
asString(row.内容类型) ??
|
||||
asString(row.type) ??
|
||||
asString(row.类型) ??
|
||||
"kv",
|
||||
...(ref ? { ref } : {}),
|
||||
...(fields?.length ? { fields } : {}),
|
||||
});
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
function pickValues(
|
||||
all: Map<string, unknown>,
|
||||
fields?: string[],
|
||||
): Record<string, unknown> {
|
||||
const out: Record<string, unknown> = {};
|
||||
if (fields?.length) {
|
||||
for (const key of fields) {
|
||||
if (all.has(key)) out[key] = all.get(key);
|
||||
}
|
||||
return out;
|
||||
}
|
||||
for (const [key, value] of all) out[key] = value;
|
||||
return out;
|
||||
}
|
||||
|
||||
/** 表单元格常把对象/数组存成 JSON 字符串；投影前解开，避免壳只能看到一坨字。 */
|
||||
export function coerceJsonValue(value: unknown): unknown {
|
||||
if (typeof value === "string") {
|
||||
const t = value.trim();
|
||||
if (
|
||||
(t.startsWith("{") && t.endsWith("}")) ||
|
||||
(t.startsWith("[") && t.endsWith("]"))
|
||||
) {
|
||||
try {
|
||||
return coerceJsonValue(JSON.parse(t));
|
||||
} catch {
|
||||
return value;
|
||||
}
|
||||
}
|
||||
return value;
|
||||
}
|
||||
if (Array.isArray(value)) return value.map(coerceJsonValue);
|
||||
if (value && typeof value === "object") {
|
||||
const out: Record<string, unknown> = {};
|
||||
for (const [key, child] of Object.entries(value as Record<string, unknown>)) {
|
||||
out[key] = coerceJsonValue(child);
|
||||
}
|
||||
return out;
|
||||
}
|
||||
return value;
|
||||
}
|
||||
|
||||
function coerceValueMap(
|
||||
values: Record<string, unknown>,
|
||||
): Record<string, unknown> {
|
||||
const out: Record<string, unknown> = {};
|
||||
for (const [key, value] of Object.entries(values)) {
|
||||
out[key] = coerceJsonValue(value);
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
function isPlainObject(value: unknown): value is Record<string, unknown> {
|
||||
return Boolean(value) && typeof value === "object" && !Array.isArray(value);
|
||||
}
|
||||
|
||||
/** 名字 → 属性对象 的矩阵，压成带 name 列的行数组，方便 table 渲染。 */
|
||||
function matrixMapToRows(
|
||||
map: Record<string, unknown>,
|
||||
nameKey = "name",
|
||||
): Record<string, unknown>[] {
|
||||
return Object.entries(map).map(([name, stats]) => {
|
||||
if (isPlainObject(stats)) return { [nameKey]: name, ...stats };
|
||||
return { [nameKey]: name, value: stats };
|
||||
});
|
||||
}
|
||||
|
||||
function shapeTable(values: Record<string, unknown>): PresentBlockValue {
|
||||
const coerced = coerceValueMap(values);
|
||||
const keys = Object.keys(coerced);
|
||||
if (keys.length === 1) {
|
||||
const source = keys[0]!;
|
||||
const only = coerced[source];
|
||||
if (Array.isArray(only)) {
|
||||
return { source, rows: only };
|
||||
}
|
||||
if (isPlainObject(only)) {
|
||||
const entries = Object.entries(only);
|
||||
if (
|
||||
entries.length &&
|
||||
entries.every(([, child]) => isPlainObject(child))
|
||||
) {
|
||||
return { source, rows: matrixMapToRows(only) };
|
||||
}
|
||||
}
|
||||
}
|
||||
return coerced;
|
||||
}
|
||||
|
||||
function shapeBlock(
|
||||
type: string,
|
||||
values: Record<string, unknown>,
|
||||
): PresentBlockValue | undefined {
|
||||
const keys = Object.keys(values);
|
||||
if (!keys.length) return undefined;
|
||||
if (type === "list") {
|
||||
const coerced = coerceValueMap(values);
|
||||
return Object.keys(coerced).map(
|
||||
(key) => `${key}：${typeof coerced[key] === "string" ? coerced[key] : JSON.stringify(coerced[key])}`,
|
||||
);
|
||||
}
|
||||
if (type === "text" || type === "markdown" || type === "safe_markdown") {
|
||||
const coerced = coerceValueMap(values);
|
||||
return Object.keys(coerced)
|
||||
.map((key) => {
|
||||
const cell = coerced[key];
|
||||
const text =
|
||||
typeof cell === "string" ? cell : JSON.stringify(cell, null, 2);
|
||||
return `【${key}】${text}`;
|
||||
})
|
||||
.join("\n");
|
||||
}
|
||||
if (type === "table") {
|
||||
return shapeTable(values);
|
||||
}
|
||||
// kv / topology / 默认：解开 JSON 字符串后的对象
|
||||
return coerceValueMap(values);
|
||||
}
|
||||
|
||||
function readTagContent(
|
||||
readTag: ((tag: string) => string | null | undefined) | undefined,
|
||||
tag: string,
|
||||
fallbacks: Record<string, string | null | undefined>,
|
||||
): string | null | undefined {
|
||||
if (Object.prototype.hasOwnProperty.call(fallbacks, tag)) {
|
||||
return fallbacks[tag];
|
||||
}
|
||||
return readTag?.(tag);
|
||||
}
|
||||
|
||||
function valuesFromRef(params: {
|
||||
ref: string;
|
||||
fields?: string[];
|
||||
readTag?: (tag: string) => string | null | undefined;
|
||||
currentVarsRaw?: string | null;
|
||||
topologyRaw?: string | null;
|
||||
}): Record<string, unknown> {
|
||||
const ref = params.ref.trim();
|
||||
const fallbacks: Record<string, string | null | undefined> = {
|
||||
"变量.当前": params.currentVarsRaw,
|
||||
"运行.初始变量": params.currentVarsRaw,
|
||||
"运行.拓扑": params.topologyRaw,
|
||||
};
|
||||
const raw = readTagContent(params.readTag, ref, fallbacks);
|
||||
if (ref === "变量.当前" || ref === "运行.初始变量") {
|
||||
const map = tableValueMap(parseTableDoc(raw));
|
||||
return pickValues(map, params.fields);
|
||||
}
|
||||
if (!raw?.trim()) return {};
|
||||
// 表文档
|
||||
const table = parseTableDoc(raw);
|
||||
if (table?.rows.length) {
|
||||
return pickValues(tableValueMap(table), params.fields);
|
||||
}
|
||||
// JSON 对象（拓扑等）
|
||||
try {
|
||||
const text = extractJsonObjectText(raw) ?? raw.trim();
|
||||
const parsed = JSON.parse(text) as unknown;
|
||||
const obj = asObject(parsed);
|
||||
if (obj) {
|
||||
if (params.fields?.length) {
|
||||
const out: Record<string, unknown> = {};
|
||||
for (const key of params.fields) {
|
||||
if (key in obj) out[key] = obj[key];
|
||||
}
|
||||
return out;
|
||||
}
|
||||
return obj;
|
||||
}
|
||||
} catch {
|
||||
/* plain */
|
||||
}
|
||||
if (params.fields?.length === 1) {
|
||||
return { [params.fields[0]!]: raw.trim() };
|
||||
}
|
||||
return { value: raw.trim() };
|
||||
}
|
||||
|
||||
/**
|
||||
* 按正文组成契约 + 回复呈现绑定，从黑板投影 program_projection 块。
|
||||
*/
|
||||
export function projectPresentBlocks(params: {
|
||||
contract: ReplyDataContract;
|
||||
statusBarRaw?: string | null;
|
||||
currentVarsRaw?: string | null;
|
||||
topologyRaw?: string | null;
|
||||
readTag?: (tag: string) => string | null | undefined;
|
||||
}): Record<string, PresentBlockValue> {
|
||||
const bindings = parseStatusBarProjectionBindings(params.statusBarRaw);
|
||||
const byId = new Map(bindings.map((b) => [b.id, b]));
|
||||
const currentMap = tableValueMap(parseTableDoc(params.currentVarsRaw));
|
||||
const out: Record<string, PresentBlockValue> = {};
|
||||
|
||||
for (const block of params.contract.programBlocks) {
|
||||
const bound = byId.get(block.id);
|
||||
const type = bound?.type || block.type || "kv";
|
||||
let values: Record<string, unknown>;
|
||||
if (bound?.ref) {
|
||||
values = valuesFromRef({
|
||||
ref: bound.ref,
|
||||
fields: bound.fields,
|
||||
readTag: params.readTag,
|
||||
currentVarsRaw: params.currentVarsRaw,
|
||||
topologyRaw: params.topologyRaw,
|
||||
});
|
||||
} else if (type === "topology") {
|
||||
values = valuesFromRef({
|
||||
ref: "运行.拓扑",
|
||||
fields: bound?.fields,
|
||||
readTag: params.readTag,
|
||||
currentVarsRaw: params.currentVarsRaw,
|
||||
topologyRaw: params.topologyRaw,
|
||||
});
|
||||
} else {
|
||||
values = pickValues(currentMap, bound?.fields);
|
||||
}
|
||||
const shaped = shapeBlock(type, values);
|
||||
if (shaped !== undefined) out[block.id] = shaped;
|
||||
}
|
||||
return out;
|
||||
}
|
||||
|
||||
/** 仅用契约（无监控栏）时的便捷入口 */
|
||||
export function projectPresentBlocksFromContract(
|
||||
contract: ReplyDataContract,
|
||||
currentVarsRaw: string | null | undefined,
|
||||
): Record<string, PresentBlockValue> {
|
||||
return projectPresentBlocks({ contract, currentVarsRaw });
|
||||
}
|
||||
|
||||
export function mergePresentWithProjections(
|
||||
content: string,
|
||||
projections: Record<string, PresentBlockValue>,
|
||||
fallbackShell: PresentShellId = "prose",
|
||||
): string {
|
||||
if (!Object.keys(projections).length) return content;
|
||||
const view = parsePresentPacket(content, fallbackShell);
|
||||
const blocks = { ...view.packet.blocks, ...projections };
|
||||
return stringifyPresentPacket({
|
||||
...view.packet,
|
||||
blocks,
|
||||
});
|
||||
}
|
||||
|
||||
export function projectionForBlocks(
|
||||
blocks: ReplyContractBlock[],
|
||||
currentVarsRaw: string | null | undefined,
|
||||
): Record<string, PresentBlockValue> {
|
||||
if (!blocks.length) return {};
|
||||
const contract: ReplyDataContract = {
|
||||
modelBlocks: [],
|
||||
programBlocks: blocks,
|
||||
};
|
||||
return projectPresentBlocks({ contract, currentVarsRaw });
|
||||
}
|
||||
@@ -0,0 +1,249 @@
|
||||
/**
|
||||
* 从「设计.正文组成」取出数据契约。
|
||||
* 开场与之后每一楼只填模型生成区；程序把这份 present.v1 灌进已固定的 frontend。
|
||||
*/
|
||||
import { extractJsonObjectText } from "./worker-set-parse.js";
|
||||
|
||||
export type ReplyContractBlock = {
|
||||
id: string;
|
||||
type: string;
|
||||
source: "model_generated" | "program_projection";
|
||||
role?: string;
|
||||
};
|
||||
|
||||
export type ReplyFrontend = {
|
||||
html: string;
|
||||
css: string;
|
||||
js: string;
|
||||
};
|
||||
|
||||
export type ReplyDataContract = {
|
||||
modelBlocks: ReplyContractBlock[];
|
||||
programBlocks: ReplyContractBlock[];
|
||||
frontend?: ReplyFrontend;
|
||||
};
|
||||
|
||||
export function parseReplyDataContract(
|
||||
raw: string | undefined | null,
|
||||
): ReplyDataContract | undefined {
|
||||
const doc = parseObject(raw);
|
||||
if (!doc) return undefined;
|
||||
const body = asObject(doc.正文) ?? doc;
|
||||
const contract = asObject(body.数据契约) ?? asObject(body.data_contract);
|
||||
const blocksRaw = asObject(contract?.blocks);
|
||||
if (!blocksRaw) return undefined;
|
||||
const modelBlocks: ReplyContractBlock[] = [];
|
||||
const programBlocks: ReplyContractBlock[] = [];
|
||||
for (const [id, value] of Object.entries(blocksRaw)) {
|
||||
const row = asObject(value);
|
||||
if (!id.trim() || !row) continue;
|
||||
const block: ReplyContractBlock = {
|
||||
id: id.trim(),
|
||||
type: asString(row.类型 ?? row.type) || "text",
|
||||
source:
|
||||
asString(row.来源 ?? row.source) === "program_projection"
|
||||
? "program_projection"
|
||||
: "model_generated",
|
||||
role: asString(row.职责 ?? row.role),
|
||||
};
|
||||
if (block.source === "program_projection") programBlocks.push(block);
|
||||
else modelBlocks.push(block);
|
||||
}
|
||||
if (!modelBlocks.length && !programBlocks.length) return undefined;
|
||||
const frontend = parseFrontend(asObject(body.frontend));
|
||||
return {
|
||||
modelBlocks,
|
||||
programBlocks,
|
||||
...(frontend ? { frontend } : {}),
|
||||
};
|
||||
}
|
||||
|
||||
export function replyContractSummary(contract: ReplyDataContract): {
|
||||
schema: "present.v1";
|
||||
model_blocks: Array<{ id: string; type: string; role?: string }>;
|
||||
program_blocks: string[];
|
||||
rule: string;
|
||||
} {
|
||||
return {
|
||||
schema: "present.v1",
|
||||
model_blocks: contract.modelBlocks.map((block) => ({
|
||||
id: block.id,
|
||||
type: block.type,
|
||||
...(block.role ? { role: block.role } : {}),
|
||||
})),
|
||||
program_blocks: contract.programBlocks.map((block) => block.id),
|
||||
rule: "开场和之后每一轮只填写 model_blocks。程序把这份 present.v1 灌进已固定的 frontend，替换示例内容。不要重写 frontend，不要填写 program_blocks。",
|
||||
};
|
||||
}
|
||||
|
||||
/** 开场 JSON 里所有 present.v1 的 blocks。缺模型区时返回错误，供 insert 拒绝。 */
|
||||
export function openingContractError(
|
||||
content: string,
|
||||
contract: ReplyDataContract,
|
||||
): string | undefined {
|
||||
if (!contract.modelBlocks.length) return undefined;
|
||||
const maps = collectPresentBlockMaps(content);
|
||||
const need = formatModelBlocks(contract);
|
||||
if (!maps.length) {
|
||||
return `有正文组成时，开场必须给出 present.v1，并填写模型生成区：${need}。程序投影区不要填写。`;
|
||||
}
|
||||
const missing = new Set<string>();
|
||||
for (const blocks of maps) {
|
||||
for (const block of contract.modelBlocks) {
|
||||
if (!Object.prototype.hasOwnProperty.call(blocks, block.id)) {
|
||||
missing.add(block.id);
|
||||
}
|
||||
}
|
||||
}
|
||||
if (!missing.size) return undefined;
|
||||
return `开场的 present.v1 未按正文组成填写模型生成区，缺少：${[...missing].join("、")}。需要：${need}。程序投影区不要填写。`;
|
||||
}
|
||||
|
||||
export function presentSchemaFromContract(contract: ReplyDataContract): {
|
||||
schema: Record<string, unknown>;
|
||||
loose: boolean;
|
||||
} {
|
||||
const properties: Record<string, unknown> = {};
|
||||
const required: string[] = [];
|
||||
let loose = false;
|
||||
for (const block of contract.modelBlocks) {
|
||||
required.push(block.id);
|
||||
const built = schemaForBlock(block);
|
||||
properties[block.id] = built.schema;
|
||||
if (built.loose) loose = true;
|
||||
}
|
||||
return {
|
||||
loose,
|
||||
schema: {
|
||||
type: "object",
|
||||
additionalProperties: false,
|
||||
required: ["schema", "blocks"],
|
||||
properties: {
|
||||
schema: { type: "string", enum: ["present.v1"] },
|
||||
blocks: {
|
||||
type: "object",
|
||||
additionalProperties: false,
|
||||
required,
|
||||
properties,
|
||||
},
|
||||
},
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
export function contractOutputInstruction(contract: ReplyDataContract): string {
|
||||
const lines = contract.modelBlocks.map((block) => {
|
||||
const role = block.role ? `：${block.role}` : "";
|
||||
return `- ${block.id}（${typeHint(block.type)}）${role}`;
|
||||
});
|
||||
const projected = contract.programBlocks.map((block) => block.id);
|
||||
return `
|
||||
|
||||
---
|
||||
|
||||
## 运行时输出协议（按正文组成灌数）
|
||||
|
||||
本局前端已经固定。只输出一个 present.v1 对象，程序把它灌进正文组成，替换示例内容。
|
||||
不要输出 HTML、CSS、JS，不要重写模板，不要再套 shell。
|
||||
|
||||
只填模型生成区：
|
||||
${lines.join("\n")}
|
||||
${projected.length ? `不要填写程序投影区：${projected.join("、")}。\n` : ""}list 用字符串数组，没有内容时给 []。markdown 与 text 用字符串。
|
||||
不要写追问、askUser、自评或「可验收产物」。`;
|
||||
}
|
||||
|
||||
function formatModelBlocks(contract: ReplyDataContract): string {
|
||||
return contract.modelBlocks
|
||||
.map((block) => `${block.id}（${typeHint(block.type)}）`)
|
||||
.join("、");
|
||||
}
|
||||
|
||||
function typeHint(type: string): string {
|
||||
if (type === "list") return "字符串数组";
|
||||
if (type === "markdown" || type === "text") return "字符串";
|
||||
if (type === "kv" || type === "table" || type === "topology") return "对象";
|
||||
return type || "字符串";
|
||||
}
|
||||
|
||||
function schemaForBlock(block: ReplyContractBlock): {
|
||||
schema: Record<string, unknown>;
|
||||
loose: boolean;
|
||||
} {
|
||||
if (block.type === "list") {
|
||||
return {
|
||||
loose: false,
|
||||
schema: { type: "array", items: { type: "string" } },
|
||||
};
|
||||
}
|
||||
if (block.type === "markdown" || block.type === "text") {
|
||||
return { loose: false, schema: { type: "string" } };
|
||||
}
|
||||
return {
|
||||
loose: true,
|
||||
schema: { type: "object", additionalProperties: true },
|
||||
};
|
||||
}
|
||||
|
||||
function collectPresentBlockMaps(content: string): Array<Record<string, unknown>> {
|
||||
const doc = parseObject(content);
|
||||
if (!doc) return [];
|
||||
const maps: Array<Record<string, unknown>> = [];
|
||||
walk(doc, maps, 0);
|
||||
return maps;
|
||||
}
|
||||
|
||||
function walk(node: unknown, maps: Array<Record<string, unknown>>, depth: number): void {
|
||||
if (depth > 8 || node == null) return;
|
||||
if (typeof node === "string") {
|
||||
if (!node.includes("present.v1")) return;
|
||||
const parsed = parseObject(node);
|
||||
if (parsed) walk(parsed, maps, depth + 1);
|
||||
return;
|
||||
}
|
||||
if (Array.isArray(node)) {
|
||||
for (const item of node) walk(item, maps, depth + 1);
|
||||
return;
|
||||
}
|
||||
if (typeof node !== "object") return;
|
||||
const row = node as Record<string, unknown>;
|
||||
const blocks = asObject(row.blocks);
|
||||
if ((row.schema === "present.v1" || row.shell != null) && blocks) {
|
||||
maps.push(blocks);
|
||||
}
|
||||
for (const value of Object.values(row)) walk(value, maps, depth + 1);
|
||||
}
|
||||
|
||||
function parseFrontend(raw: Record<string, unknown> | undefined): ReplyFrontend | undefined {
|
||||
if (!raw) return undefined;
|
||||
const html = asString(raw.html);
|
||||
if (!html || !/<\w+/.test(html)) return undefined;
|
||||
return {
|
||||
html,
|
||||
css: asString(raw.css) ?? "",
|
||||
js: asString(raw.js) ?? "",
|
||||
};
|
||||
}
|
||||
|
||||
function parseObject(raw: string | undefined | null): Record<string, unknown> | undefined {
|
||||
const text = raw?.trim() ?? "";
|
||||
if (!text) return undefined;
|
||||
const jsonText = extractJsonObjectText(text) ?? (text.startsWith("{") ? text : null);
|
||||
if (!jsonText) return undefined;
|
||||
try {
|
||||
const doc = JSON.parse(jsonText) as unknown;
|
||||
return asObject(doc);
|
||||
} catch {
|
||||
return undefined;
|
||||
}
|
||||
}
|
||||
|
||||
function asObject(value: unknown): Record<string, unknown> | undefined {
|
||||
if (!value || typeof value !== "object" || Array.isArray(value)) return undefined;
|
||||
return value as Record<string, unknown>;
|
||||
}
|
||||
|
||||
function asString(value: unknown): string | undefined {
|
||||
if (typeof value !== "string") return undefined;
|
||||
const text = value.trim();
|
||||
return text || undefined;
|
||||
}
|
||||
@@ -0,0 +1,216 @@
|
||||
/**
|
||||
* 从「设计.变量设计与更新规则」编译运行目录，并种进 变量.当前。
|
||||
* 创作产物是合同；设计.变量目录 / 变量.当前 是运行表。
|
||||
*/
|
||||
import { extractJsonObjectText } from "./worker-set-parse.js";
|
||||
import {
|
||||
parseVariableCatalog,
|
||||
seedVariablesFromCatalog,
|
||||
serializeVariableCatalog,
|
||||
upsertVariableField,
|
||||
VARIABLE_CATALOG_TAG,
|
||||
type VariableCatalogDoc,
|
||||
type VariableCatalogField,
|
||||
type VariableFieldType,
|
||||
} from "./variable-catalog.js";
|
||||
|
||||
export const VARIABLE_DESIGN_TAG = "设计.变量设计与更新规则";
|
||||
|
||||
function asString(v: unknown): string | undefined {
|
||||
if (typeof v !== "string") return undefined;
|
||||
const t = v.trim();
|
||||
return t || undefined;
|
||||
}
|
||||
|
||||
function asObject(v: unknown): Record<string, unknown> | undefined {
|
||||
if (!v || typeof v !== "object" || Array.isArray(v)) return undefined;
|
||||
return v as Record<string, unknown>;
|
||||
}
|
||||
|
||||
function parseType(raw: unknown): VariableFieldType {
|
||||
const s = asString(raw)?.toLowerCase();
|
||||
if (s === "number" || s === "数字" || s === "integer" || s === "int") {
|
||||
return "number";
|
||||
}
|
||||
if (s === "boolean" || s === "布尔") return "boolean";
|
||||
if (s === "enum" || s === "枚举") return "enum";
|
||||
return "string";
|
||||
}
|
||||
|
||||
/** 解析变量设计产物里的真值列表 → 目录字段（未写入） */
|
||||
export function catalogFieldsFromVariableDesign(
|
||||
raw: string | null | undefined,
|
||||
): VariableCatalogField[] {
|
||||
if (!raw?.trim()) return [];
|
||||
const text = extractJsonObjectText(raw) ?? raw.trim();
|
||||
let doc: unknown;
|
||||
try {
|
||||
doc = JSON.parse(text);
|
||||
} catch {
|
||||
return [];
|
||||
}
|
||||
const root = asObject(doc);
|
||||
if (!root) return [];
|
||||
const body = asObject(root.正文) ?? root;
|
||||
const truths = body.真值 ?? body.fields ?? body.变量;
|
||||
if (!Array.isArray(truths)) return [];
|
||||
|
||||
const fields: VariableCatalogField[] = [];
|
||||
const seen = new Set<string>();
|
||||
for (const item of truths) {
|
||||
const row = asObject(item);
|
||||
if (!row) continue;
|
||||
const key =
|
||||
asString(row.id) ??
|
||||
asString(row.key) ??
|
||||
asString(row.名) ??
|
||||
asString(row.name) ??
|
||||
asString(row.名称);
|
||||
if (!key || seen.has(key)) continue;
|
||||
seen.add(key);
|
||||
const type = parseType(row.类型 ?? row.type);
|
||||
const initBox = asObject(row.初始化) ?? asObject(row.initial);
|
||||
const initial =
|
||||
initBox?.默认值 ??
|
||||
initBox?.default ??
|
||||
initBox?.value ??
|
||||
row.initial ??
|
||||
row.初值 ??
|
||||
row.默认值;
|
||||
const publicity = asString(row.公开性) ?? asString(row.visibility);
|
||||
const userVisible =
|
||||
publicity === "hidden" ||
|
||||
publicity === "隐藏" ||
|
||||
row.user_visible === false ||
|
||||
row.对用户可见 === false
|
||||
? false
|
||||
: true;
|
||||
const note =
|
||||
asString(row.名称) ??
|
||||
asString(row.用途) ??
|
||||
asString(row.note) ??
|
||||
asString(row.备注);
|
||||
const { doc: one } = upsertVariableField(null, {
|
||||
key,
|
||||
type,
|
||||
initial,
|
||||
user_visible: userVisible,
|
||||
note,
|
||||
});
|
||||
const field = one.fields[0];
|
||||
if (field) fields.push(field);
|
||||
}
|
||||
return fields;
|
||||
}
|
||||
|
||||
/** 把设计稿真值并进已有目录（不删目录里多出来的 declare 字段） */
|
||||
export function mergeCatalogFromVariableDesign(
|
||||
existing: VariableCatalogDoc | null,
|
||||
designRaw: string | null | undefined,
|
||||
): {
|
||||
doc: VariableCatalogDoc;
|
||||
added: string[];
|
||||
updated: string[];
|
||||
} | null {
|
||||
const fromDesign = catalogFieldsFromVariableDesign(designRaw);
|
||||
if (!fromDesign.length) return null;
|
||||
|
||||
let doc: VariableCatalogDoc = existing ?? {
|
||||
schema: "variable-catalog.v1",
|
||||
fields: [],
|
||||
};
|
||||
const added: string[] = [];
|
||||
const updated: string[] = [];
|
||||
for (const field of fromDesign) {
|
||||
const before = doc.fields.find((f) => f.key === field.key);
|
||||
const { doc: next } = upsertVariableField(doc, {
|
||||
key: field.key,
|
||||
type: field.type,
|
||||
initial: field.initial,
|
||||
user_visible: field.user_visible,
|
||||
user_editable: field.user_editable,
|
||||
note: field.note,
|
||||
});
|
||||
doc = next;
|
||||
if (!before) added.push(field.key);
|
||||
else updated.push(field.key);
|
||||
}
|
||||
return { doc, added, updated };
|
||||
}
|
||||
|
||||
export type SyncVariableDesignResult = {
|
||||
/** 有变更时返回新目录 JSON */
|
||||
catalog: string | null;
|
||||
current: string | null;
|
||||
initial: string | null;
|
||||
seededKeys: string[];
|
||||
catalogKeys: string[];
|
||||
added: string[];
|
||||
};
|
||||
|
||||
/**
|
||||
* 设计稿 → 目录 → 种/补 变量.当前 与 运行.初始变量。
|
||||
* 已有格不覆盖；目录缺键才追加。
|
||||
*/
|
||||
export function syncVariableDesignToRuntime(params: {
|
||||
designRaw: string | null | undefined;
|
||||
catalogRaw: string | null | undefined;
|
||||
currentRaw: string | null | undefined;
|
||||
initialRaw: string | null | undefined;
|
||||
}): SyncVariableDesignResult {
|
||||
const empty: SyncVariableDesignResult = {
|
||||
catalog: null,
|
||||
current: null,
|
||||
initial: null,
|
||||
seededKeys: [],
|
||||
catalogKeys: [],
|
||||
added: [],
|
||||
};
|
||||
const existing = parseVariableCatalog(params.catalogRaw);
|
||||
const merged = mergeCatalogFromVariableDesign(existing, params.designRaw);
|
||||
if (!merged) return empty;
|
||||
|
||||
const catalogChanged =
|
||||
!existing ||
|
||||
serializeVariableCatalog(existing) !== serializeVariableCatalog(merged.doc);
|
||||
const catalogRaw = catalogChanged
|
||||
? serializeVariableCatalog(merged.doc)
|
||||
: (params.catalogRaw?.trim() ?? serializeVariableCatalog(merged.doc));
|
||||
|
||||
const seeded = seedVariablesFromCatalog({
|
||||
catalogRaw,
|
||||
currentRaw: params.currentRaw,
|
||||
initialRaw: params.initialRaw,
|
||||
actor: "system:variable-design",
|
||||
});
|
||||
|
||||
return {
|
||||
catalog: catalogChanged ? catalogRaw : null,
|
||||
current: seeded.current,
|
||||
initial: seeded.initial,
|
||||
seededKeys: seeded.seededKeys,
|
||||
catalogKeys: merged.doc.fields.map((f) => f.key),
|
||||
added: merged.added,
|
||||
};
|
||||
}
|
||||
|
||||
/** 把 sync 结果写回黑板；返回实际写入的 tag */
|
||||
export function writeVariableDesignSync(
|
||||
sync: SyncVariableDesignResult,
|
||||
writeTag: (tag: string, content: string) => void,
|
||||
): string[] {
|
||||
const wrote: string[] = [];
|
||||
if (sync.catalog) {
|
||||
writeTag(VARIABLE_CATALOG_TAG, sync.catalog);
|
||||
wrote.push(VARIABLE_CATALOG_TAG);
|
||||
}
|
||||
if (sync.initial) {
|
||||
writeTag("运行.初始变量", sync.initial);
|
||||
wrote.push("运行.初始变量");
|
||||
}
|
||||
if (sync.current) {
|
||||
writeTag("变量.当前", sync.current);
|
||||
wrote.push("变量.当前");
|
||||
}
|
||||
return wrote;
|
||||
}
|
||||
@@ -107,7 +107,7 @@ export const DESIGN_STAGE_SKILL_META: Record<
|
||||
"opening-generator": {
|
||||
label: "开局 · 开场白",
|
||||
purpose:
|
||||
"创作末尾：优先落库设计.开场白与开场变量；否则现写开场；初值与开场同真相。",
|
||||
"创作末尾：每条开场落库设计.开场白#场景短码；没有变量的卡不写初值。",
|
||||
},
|
||||
};
|
||||
|
||||
|
||||
@@ -1,7 +1,8 @@
|
||||
import { appendFileSync, existsSync, mkdirSync, readFileSync, readdirSync } from "node:fs";
|
||||
import { appendFileSync, existsSync, mkdirSync, readFileSync, readdirSync, writeFileSync } from "node:fs";
|
||||
import path from "node:path";
|
||||
import { randomUUID } from "node:crypto";
|
||||
import { ensureUserDataDirs, getUserDataDir } from "../config/user-data-dir.js";
|
||||
import type { ContextSegment } from "../llm/context-outline.js";
|
||||
|
||||
export type TokenUsageRecord = {
|
||||
id: string;
|
||||
@@ -17,7 +18,25 @@ export type TokenUsageRecord = {
|
||||
totalTokens: number;
|
||||
cachedTokens?: number;
|
||||
cacheMissTokens?: number;
|
||||
cacheWriteTokens?: number;
|
||||
messageId?: string;
|
||||
turnId?: string;
|
||||
step?: string;
|
||||
retryIndex?: number;
|
||||
segmentChars?: {
|
||||
rules: number;
|
||||
history: number;
|
||||
products: number;
|
||||
injected: number;
|
||||
tail: number;
|
||||
};
|
||||
/** 有序标签与哈希，不含正文 */
|
||||
segments?: ContextSegment[];
|
||||
plannedTags?: string[];
|
||||
succeededTags?: string[];
|
||||
failedTags?: string[];
|
||||
finish?: string;
|
||||
latencyMs?: number;
|
||||
};
|
||||
|
||||
/** Token usage attached to a chat message in SessionView */
|
||||
@@ -97,7 +116,18 @@ export function recordTokenUsage(
|
||||
totalTokens: partial.totalTokens,
|
||||
cachedTokens: partial.cachedTokens,
|
||||
cacheMissTokens: partial.cacheMissTokens,
|
||||
cacheWriteTokens: partial.cacheWriteTokens,
|
||||
messageId: partial.messageId,
|
||||
turnId: partial.turnId,
|
||||
step: partial.step,
|
||||
retryIndex: partial.retryIndex,
|
||||
segmentChars: partial.segmentChars,
|
||||
segments: partial.segments,
|
||||
plannedTags: partial.plannedTags,
|
||||
succeededTags: partial.succeededTags,
|
||||
failedTags: partial.failedTags,
|
||||
finish: partial.finish,
|
||||
latencyMs: partial.latencyMs,
|
||||
};
|
||||
appendFileSync(dayFile(new Date(record.at)), `${JSON.stringify(record)}\n`, "utf8");
|
||||
return record;
|
||||
@@ -238,6 +268,39 @@ export function getSessionTokenTotals(sessionId: string): {
|
||||
};
|
||||
}
|
||||
|
||||
export function patchTokenUsage(
|
||||
id: string,
|
||||
patch: Partial<
|
||||
Pick<
|
||||
TokenUsageRecord,
|
||||
"plannedTags" | "succeededTags" | "failedTags" | "finish"
|
||||
>
|
||||
>,
|
||||
): void {
|
||||
if (!id) return;
|
||||
const files = listStatFiles();
|
||||
for (let i = files.length - 1; i >= 0; i--) {
|
||||
const file = files[i]!;
|
||||
const lines = readFileSync(file, "utf8").split(/\r?\n/);
|
||||
let changed = false;
|
||||
const next = lines.map((line) => {
|
||||
if (!line.trim()) return line;
|
||||
try {
|
||||
const record = JSON.parse(line) as TokenUsageRecord;
|
||||
if (record.id !== id) return line;
|
||||
changed = true;
|
||||
return JSON.stringify({ ...record, ...patch });
|
||||
} catch {
|
||||
return line;
|
||||
}
|
||||
});
|
||||
if (changed) {
|
||||
writeFileSync(file, next.join("\n"), "utf8");
|
||||
return;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
export function toMessageTokenUsage(record: TokenUsageRecord): MessageTokenUsage {
|
||||
return {
|
||||
totalTokens: record.totalTokens,
|
||||
|
||||
@@ -0,0 +1,42 @@
|
||||
import { createHash } from "node:crypto";
|
||||
import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
|
||||
import path from "node:path";
|
||||
import { getUserDataDir } from "../config/user-data-dir.js";
|
||||
|
||||
function cacheDir(): string {
|
||||
const dir = path.join(getUserDataDir(), "tts-cache");
|
||||
mkdirSync(dir, { recursive: true });
|
||||
return dir;
|
||||
}
|
||||
|
||||
export function ttsCacheKey(parts: {
|
||||
provider: string;
|
||||
voice: string;
|
||||
text: string;
|
||||
extra?: string;
|
||||
}): string {
|
||||
const h = createHash("sha256");
|
||||
h.update(parts.provider);
|
||||
h.update("\0");
|
||||
h.update(parts.voice);
|
||||
h.update("\0");
|
||||
h.update(parts.extra ?? "");
|
||||
h.update("\0");
|
||||
h.update(parts.text);
|
||||
return h.digest("hex").slice(0, 40);
|
||||
}
|
||||
|
||||
export function readTtsCache(key: string, ext: string): Buffer | null {
|
||||
const file = path.join(cacheDir(), `${key}.${ext}`);
|
||||
if (!existsSync(file)) return null;
|
||||
try {
|
||||
return readFileSync(file);
|
||||
} catch {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
export function writeTtsCache(key: string, ext: string, data: Buffer): void {
|
||||
const file = path.join(cacheDir(), `${key}.${ext}`);
|
||||
writeFileSync(file, data);
|
||||
}
|
||||
+109
@@ -0,0 +1,109 @@
|
||||
import { MsEdgeTTS, OUTPUT_FORMAT } from "msedge-tts";
|
||||
import { readTtsCache, ttsCacheKey, writeTtsCache } from "./cache.js";
|
||||
import {
|
||||
DEFAULT_EDGE_VOICE,
|
||||
type TtsSynthesizeParams,
|
||||
type TtsSynthesizeResult,
|
||||
} from "./types.js";
|
||||
|
||||
/** Edge SSML 单段不宜过长；按句切到约此字数 */
|
||||
const CHUNK_TARGET = 1200;
|
||||
|
||||
function escapeXml(text: string): string {
|
||||
return text
|
||||
.replace(/&/g, "&")
|
||||
.replace(/</g, "<")
|
||||
.replace(/>/g, ">")
|
||||
.replace(/"/g, """)
|
||||
.replace(/'/g, "'");
|
||||
}
|
||||
|
||||
function resolveVoice(explicit?: string): string {
|
||||
const fromEnv = process.env.WRITING_AGENT_TTS_VOICE?.trim();
|
||||
return (explicit?.trim() || fromEnv || DEFAULT_EDGE_VOICE).trim();
|
||||
}
|
||||
|
||||
function splitChunks(text: string): string[] {
|
||||
const cleaned = text.replace(/\r\n/g, "\n").trim();
|
||||
if (!cleaned) return [];
|
||||
if (cleaned.length <= CHUNK_TARGET) return [cleaned];
|
||||
|
||||
const parts: string[] = [];
|
||||
let buf = "";
|
||||
const pushBuf = () => {
|
||||
const t = buf.trim();
|
||||
if (t) parts.push(t);
|
||||
buf = "";
|
||||
};
|
||||
|
||||
const sentences = cleaned.split(/(?<=[。！？!?；;\n])/u);
|
||||
for (const s of sentences) {
|
||||
if (!s) continue;
|
||||
if (buf.length + s.length > CHUNK_TARGET && buf.length > 0) {
|
||||
pushBuf();
|
||||
}
|
||||
buf += s;
|
||||
if (buf.length >= CHUNK_TARGET) pushBuf();
|
||||
}
|
||||
pushBuf();
|
||||
return parts.length ? parts : [cleaned.slice(0, CHUNK_TARGET)];
|
||||
}
|
||||
|
||||
async function streamToBuffer(
|
||||
audioStream: NodeJS.ReadableStream,
|
||||
): Promise<Buffer> {
|
||||
const chunks: Buffer[] = [];
|
||||
for await (const chunk of audioStream) {
|
||||
chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
|
||||
}
|
||||
return Buffer.concat(chunks);
|
||||
}
|
||||
|
||||
async function synthesizeChunk(voice: string, text: string): Promise<Buffer> {
|
||||
const tts = new MsEdgeTTS();
|
||||
await tts.setMetadata(voice, OUTPUT_FORMAT.AUDIO_24KHZ_48KBITRATE_MONO_MP3);
|
||||
const { audioStream } = tts.toStream(escapeXml(text));
|
||||
return streamToBuffer(audioStream);
|
||||
}
|
||||
|
||||
/**
|
||||
* Edge 在线神经 TTS（微软朗读通道）。文本会发往微软服务端。
|
||||
*/
|
||||
export async function synthesizeWithEdge(
|
||||
params: TtsSynthesizeParams,
|
||||
): Promise<TtsSynthesizeResult> {
|
||||
const text = params.text.trim();
|
||||
if (!text) throw new Error("TTS 文本为空");
|
||||
const voice = resolveVoice(params.voice);
|
||||
const key = ttsCacheKey({
|
||||
provider: "edge",
|
||||
voice,
|
||||
text,
|
||||
extra: params.cacheKey,
|
||||
});
|
||||
const hit = readTtsCache(key, "mp3");
|
||||
if (hit) {
|
||||
return {
|
||||
audio: hit,
|
||||
contentType: "audio/mpeg",
|
||||
provider: "edge",
|
||||
voice,
|
||||
cached: true,
|
||||
};
|
||||
}
|
||||
|
||||
const chunks = splitChunks(text);
|
||||
const buffers: Buffer[] = [];
|
||||
for (const chunk of chunks) {
|
||||
buffers.push(await synthesizeChunk(voice, chunk));
|
||||
}
|
||||
const audio = Buffer.concat(buffers);
|
||||
writeTtsCache(key, "mp3", audio);
|
||||
return {
|
||||
audio,
|
||||
contentType: "audio/mpeg",
|
||||
provider: "edge",
|
||||
voice,
|
||||
cached: false,
|
||||
};
|
||||
}
|
||||
@@ -0,0 +1,42 @@
|
||||
import { synthesizeWithEdge } from "./edge.js";
|
||||
import {
|
||||
DEFAULT_TTS_PROVIDER,
|
||||
type TtsProviderId,
|
||||
type TtsSynthesizeParams,
|
||||
type TtsSynthesizeResult,
|
||||
} from "./types.js";
|
||||
|
||||
export {
|
||||
DEFAULT_EDGE_VOICE,
|
||||
DEFAULT_TTS_PROVIDER,
|
||||
type TtsProviderId,
|
||||
type TtsSynthesizeParams,
|
||||
type TtsSynthesizeResult,
|
||||
} from "./types.js";
|
||||
|
||||
export function resolveTtsProvider(): TtsProviderId {
|
||||
const raw = (process.env.WRITING_AGENT_TTS_PROVIDER ?? DEFAULT_TTS_PROVIDER)
|
||||
.trim()
|
||||
.toLowerCase();
|
||||
if (raw === "local" || raw === "tencent" || raw === "edge") return raw;
|
||||
return DEFAULT_TTS_PROVIDER;
|
||||
}
|
||||
|
||||
/**
|
||||
* 统一合成入口。当前仅实现 edge；local / tencent 占位。
|
||||
*/
|
||||
export async function synthesizeSpeech(
|
||||
params: TtsSynthesizeParams,
|
||||
): Promise<TtsSynthesizeResult> {
|
||||
const provider = resolveTtsProvider();
|
||||
if (provider === "edge") {
|
||||
return synthesizeWithEdge(params);
|
||||
}
|
||||
if (provider === "local") {
|
||||
throw new Error("本地 TTS 尚未接入；请将 WRITING_AGENT_TTS_PROVIDER=edge");
|
||||
}
|
||||
if (provider === "tencent") {
|
||||
throw new Error("腾讯云 TTS 尚未接入；请将 WRITING_AGENT_TTS_PROVIDER=edge");
|
||||
}
|
||||
return synthesizeWithEdge(params);
|
||||
}
|
||||
@@ -0,0 +1,22 @@
|
||||
/** TTS 提供方（默认 edge；本地/腾讯后接） */
|
||||
|
||||
export type TtsProviderId = "edge" | "local" | "tencent";
|
||||
|
||||
export type TtsSynthesizeParams = {
|
||||
text: string;
|
||||
/** 缓存键额外盐（如 sectionId），便于按节失效 */
|
||||
cacheKey?: string;
|
||||
voice?: string;
|
||||
};
|
||||
|
||||
export type TtsSynthesizeResult = {
|
||||
audio: Buffer;
|
||||
contentType: string;
|
||||
provider: TtsProviderId;
|
||||
voice: string;
|
||||
cached: boolean;
|
||||
};
|
||||
|
||||
export const DEFAULT_TTS_PROVIDER: TtsProviderId = "edge";
|
||||
/** 听书默认中文声线 */
|
||||
export const DEFAULT_EDGE_VOICE = "zh-CN-XiaoxiaoNeural";
|
||||
@@ -39,7 +39,22 @@ export type PersistedChatMessage = {
|
||||
messages: Array<{ role: string; content: string }>;
|
||||
charCount: number;
|
||||
model?: string;
|
||||
segments?: Array<{ tag: string; chars: number; hash: string }>;
|
||||
};
|
||||
/** 各次请求的标签轮廓，不含正文 */
|
||||
requestOutlines?: Array<{
|
||||
at: string;
|
||||
caller: string;
|
||||
step?: string;
|
||||
turnId?: string;
|
||||
model?: string;
|
||||
promptTokens?: number;
|
||||
completionTokens?: number;
|
||||
cachedTokens?: number;
|
||||
cacheMissTokens?: number;
|
||||
cacheWriteTokens?: number;
|
||||
segments: Array<{ tag: string; chars: number; hash: string }>;
|
||||
}>;
|
||||
branchGroupId?: string;
|
||||
branchIndex?: number;
|
||||
branchTotal?: number;
|
||||
|
||||
@@ -1,9 +1,14 @@
|
||||
/** 进料/创作方式：一书一方式，目录按此分夹 */
|
||||
export type BookCreationMode = "recipe" | "dictate" | "listen_serial";
|
||||
|
||||
export type BookProject = {
|
||||
id: string;
|
||||
title: string;
|
||||
/** 用户选定或 agent 确认的 skill 包 */
|
||||
activeSkillId?: string;
|
||||
activeSkillName?: string;
|
||||
/** 新建时选定；缺省视为 dictate（旧书） */
|
||||
creationMode?: BookCreationMode;
|
||||
preview: string;
|
||||
sessionIds: string[];
|
||||
/** 最近一次持久化的 sessionId，用于续作 */
|
||||
@@ -22,6 +27,7 @@ export type BookSummary = Pick<
|
||||
| "title"
|
||||
| "activeSkillId"
|
||||
| "activeSkillName"
|
||||
| "creationMode"
|
||||
| "preview"
|
||||
| "updatedAt"
|
||||
| "orchestratorId"
|
||||
|
||||
@@ -1,3 +1,4 @@
|
||||
import type { ContextSegment } from "../llm/context-outline.js";
|
||||
import type { GenerationParameters } from "./preset.js";
|
||||
|
||||
/** 一次 LLM 调用实际发送的上下文（全量，供右键观察） */
|
||||
@@ -11,6 +12,8 @@ export type LlmContextTrace = {
|
||||
caller: string;
|
||||
createdAt: string;
|
||||
messages: LlmContextMessage[];
|
||||
/** 有序标签，供查看排列和缓存断点；不含正文 */
|
||||
segments?: ContextSegment[];
|
||||
/** 便于列表展示 */
|
||||
charCount: number;
|
||||
model?: string;
|
||||
@@ -57,6 +60,7 @@ export function buildContextTrace(params: {
|
||||
}>;
|
||||
model?: string;
|
||||
generation?: GenerationParameters;
|
||||
segments?: ContextSegment[];
|
||||
}): LlmContextTrace {
|
||||
const messages = snapshotRequestMessages(params.messages);
|
||||
const charCount = messages.reduce((n, m) => n + m.content.length, 0);
|
||||
@@ -64,6 +68,7 @@ export function buildContextTrace(params: {
|
||||
caller: params.caller,
|
||||
createdAt: new Date().toISOString(),
|
||||
messages,
|
||||
...(params.segments?.length ? { segments: params.segments } : {}),
|
||||
charCount,
|
||||
model: params.model,
|
||||
...(params.generation && Object.keys(params.generation).length
|
||||
|
||||
+41
-16
@@ -21,6 +21,12 @@ import {
|
||||
} from "../parse/json-doc.js";
|
||||
import { assembleWorkerContext } from "../skills/context-segments.js";
|
||||
import { isPlayLayerActive } from "../skills/play-turn.js";
|
||||
import { isNarratorEnabled } from "../skills/play-slots.js";
|
||||
import {
|
||||
contractOutputInstruction,
|
||||
parseReplyDataContract,
|
||||
presentSchemaFromContract,
|
||||
} from "../skills/reply-contract.js";
|
||||
import {
|
||||
PRESENT_TAG,
|
||||
PRESENT_JSON_SCHEMA,
|
||||
@@ -829,21 +835,33 @@ export async function runWorkerSkill(params: WorkerRunParams): Promise<WorkerRun
|
||||
isPlayLayerActive(params.slots) && isPlayGmBodyWorker(worker.id);
|
||||
const playPresent =
|
||||
isPlayLayerActive(params.slots) && isPlayPresentWorker(worker.id);
|
||||
const replyFormatRaw =
|
||||
params.blackboard.getContentByTag("设计.正文组成") ??
|
||||
params.blackboard.getContentByTag("设计.0层正文组成") ??
|
||||
params.blackboard.getContentByTag("设计.回复格式");
|
||||
const replyContract = parseReplyDataContract(replyFormatRaw);
|
||||
const narratorOn = isNarratorEnabled(playSlots);
|
||||
const fillByContract =
|
||||
Boolean(replyContract?.frontend && replyContract.modelBlocks.length) &&
|
||||
(playPresent || (playGmBody && !narratorOn));
|
||||
const contractSchema =
|
||||
fillByContract && replyContract
|
||||
? presentSchemaFromContract(replyContract)
|
||||
: undefined;
|
||||
const playShell: PresentShellId | undefined =
|
||||
playPresent || playGmBody
|
||||
? parseShellAdaptationFromReplyFormat(
|
||||
params.blackboard.getContentByTag("设计.正文组成") ??
|
||||
params.blackboard.getContentByTag("设计.回复格式"),
|
||||
)?.shell_id ?? "prose"
|
||||
? parseShellAdaptationFromReplyFormat(replyFormatRaw)?.shell_id ?? "prose"
|
||||
: undefined;
|
||||
|
||||
const systemContent =
|
||||
promptBody +
|
||||
(playGmBody
|
||||
? playGmBodyOutputInstruction()
|
||||
: playPresent && playShell
|
||||
? playPresentOutputInstruction(playShell)
|
||||
: WORKER_OUTPUT_INSTRUCTION) +
|
||||
(contractSchema && replyContract && !useGmHarness
|
||||
? contractOutputInstruction(replyContract)
|
||||
: playGmBody
|
||||
? playGmBodyOutputInstruction()
|
||||
: playPresent && playShell
|
||||
? playPresentOutputInstruction(playShell)
|
||||
: WORKER_OUTPUT_INSTRUCTION) +
|
||||
(useGmHarness ? GM_CHANCE_HARNESS_INSTRUCTION : "");
|
||||
|
||||
// 任务 messages 即可；preset 夹心由 wrapLlmForSession / PresetLlmProvider 统一装配
|
||||
@@ -888,7 +906,7 @@ export async function runWorkerSkill(params: WorkerRunParams): Promise<WorkerRun
|
||||
}
|
||||
: undefined;
|
||||
|
||||
if (playGmBody) {
|
||||
if (playGmBody && !contractSchema) {
|
||||
const completeOpts = {
|
||||
caller: `worker:${worker.id}`,
|
||||
label: worker.name?.trim() || worker.id,
|
||||
@@ -905,12 +923,19 @@ export async function runWorkerSkill(params: WorkerRunParams): Promise<WorkerRun
|
||||
}
|
||||
|
||||
const result = await completeStructured(params.llm, messages, {
|
||||
schema: playPresent ? PRESENT_JSON_SCHEMA : {
|
||||
type: "object",
|
||||
additionalProperties: true,
|
||||
},
|
||||
name: playPresent ? "submit_present_packet" : "submit_worker_result",
|
||||
schemaIsLoose: !playPresent,
|
||||
schema: contractSchema
|
||||
? contractSchema.schema
|
||||
: playPresent
|
||||
? PRESENT_JSON_SCHEMA
|
||||
: {
|
||||
type: "object",
|
||||
additionalProperties: true,
|
||||
},
|
||||
name:
|
||||
contractSchema || playPresent
|
||||
? "submit_present_packet"
|
||||
: "submit_worker_result",
|
||||
schemaIsLoose: contractSchema ? contractSchema.loose : !playPresent,
|
||||
capabilities: resolveActiveProfile()?.capabilities,
|
||||
caller: `worker:${worker.id}`,
|
||||
label: worker.name?.trim() || worker.id,
|
||||
|
||||
@@ -0,0 +1,114 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import {
|
||||
annotateCacheBreaks,
|
||||
firstCacheBreak,
|
||||
outlineRequest,
|
||||
} from "../src/llm/context-outline.js";
|
||||
|
||||
const sample = [
|
||||
"规则开头",
|
||||
"",
|
||||
"## 写入规则",
|
||||
"固定规则",
|
||||
"",
|
||||
"## 更早的对话",
|
||||
"",
|
||||
"【用户】",
|
||||
"第一句",
|
||||
"",
|
||||
"【助手】",
|
||||
"回过",
|
||||
"",
|
||||
"## 当前产物（已写入黑板）",
|
||||
"",
|
||||
"### 设计.氛围 〔order=0〕",
|
||||
"雨",
|
||||
"",
|
||||
"### 设计.角色 〔order=1〕",
|
||||
"甲",
|
||||
"",
|
||||
"## 本轮注入",
|
||||
"",
|
||||
"### topology @ 3",
|
||||
"#### task",
|
||||
"做一件事",
|
||||
"",
|
||||
"## 本轮",
|
||||
"",
|
||||
"方案进度",
|
||||
"",
|
||||
"最后一轮用户输入",
|
||||
"再改一下",
|
||||
].join("\n");
|
||||
|
||||
describe("outlineRequest", () => {
|
||||
it("按标题、对话轮次和最后一轮输入切开，不留下正文", () => {
|
||||
const segments = outlineRequest([
|
||||
{ role: "system", content: sample.slice(0, sample.indexOf("## 更早的对话")) },
|
||||
{ role: "user", content: sample.slice(sample.indexOf("## 更早的对话")) },
|
||||
]);
|
||||
expect(segments.map((item) => item.tag)).toEqual([
|
||||
"规则",
|
||||
"写入规则",
|
||||
"更早的对话",
|
||||
"更早·用户",
|
||||
"更早·助手",
|
||||
"当前产物",
|
||||
"设计.氛围",
|
||||
"设计.角色",
|
||||
"本轮注入",
|
||||
"topology",
|
||||
"task",
|
||||
"本轮",
|
||||
"最后一轮用户输入",
|
||||
]);
|
||||
expect(JSON.stringify(segments)).not.toContain("再改一下");
|
||||
expect(JSON.stringify(segments)).not.toContain("雨");
|
||||
const again = outlineRequest([
|
||||
{ role: "system", content: sample.slice(0, sample.indexOf("## 更早的对话")) },
|
||||
{ role: "user", content: sample.slice(sample.indexOf("## 更早的对话")) },
|
||||
]);
|
||||
expect(again.map((item) => item.hash)).toEqual(segments.map((item) => item.hash));
|
||||
});
|
||||
|
||||
it("工具定义放在正文之前", () => {
|
||||
const segments = outlineRequest(
|
||||
[{ role: "user", content: "你好" }],
|
||||
[
|
||||
{
|
||||
type: "function",
|
||||
function: { name: "insert", description: "写", parameters: {} },
|
||||
},
|
||||
],
|
||||
);
|
||||
expect(segments.map((item) => item.tag)).toEqual(["工具", "用户"]);
|
||||
});
|
||||
});
|
||||
|
||||
describe("firstCacheBreak", () => {
|
||||
it("标出第一处标签或哈希不同的位置", () => {
|
||||
const prev = outlineRequest([{ role: "user", content: sample.slice(sample.indexOf("## 更早的对话")) }]);
|
||||
const changed = sample.replace("雨", "雪");
|
||||
const next = outlineRequest([{ role: "user", content: changed.slice(changed.indexOf("## 更早的对话")) }]);
|
||||
const index = firstCacheBreak(prev, next);
|
||||
expect(next[index]?.tag).toBe("设计.氛围");
|
||||
expect(firstCacheBreak(prev, prev)).toBe(-1);
|
||||
expect(firstCacheBreak(prev, prev.slice(0, 3))).toBe(-1);
|
||||
expect(firstCacheBreak(prev.slice(0, 3), prev)).toBe(3);
|
||||
});
|
||||
});
|
||||
|
||||
describe("annotateCacheBreaks", () => {
|
||||
it("按同会话时间比较，窗口里第一条算冷启动", () => {
|
||||
const a = outlineRequest([{ role: "user", content: "第一" }]);
|
||||
const b = outlineRequest([{ role: "user", content: "第二" }]);
|
||||
const marks = annotateCacheBreaks([
|
||||
{ id: "2", at: "2026-10-02T02:00:00.000Z", sessionId: "s", segments: b },
|
||||
{ id: "1", at: "2026-10-02T01:00:00.000Z", sessionId: "s", segments: a },
|
||||
{ id: "9", at: "2026-10-02T03:00:00.000Z", sessionId: "other", segments: a },
|
||||
]);
|
||||
expect(marks.get("1")).toMatchObject({ cold: true, matched: 0 });
|
||||
expect(marks.get("2")).toMatchObject({ cold: false, breakAt: 0, tag: "用户", matched: 0 });
|
||||
expect(marks.get("9")?.cold).toBe(true);
|
||||
});
|
||||
});
|
||||
@@ -1230,8 +1230,12 @@ modules:
|
||||
expect(sections.blocks.task).toContain("默认只生成 1 条");
|
||||
expect(sections.blocks.task).toContain("@玩家");
|
||||
expect(sections.blocks.task).toContain("随机");
|
||||
expect(sections.blocks.output).toContain('"默认候选id"');
|
||||
expect(sections.blocks.output).toContain('"开场白候选"');
|
||||
expect(module?.artifact).toBe("设计.开场白");
|
||||
expect(sections.blocks.task).toContain("设计.开场白#opening-cycle2");
|
||||
expect(sections.blocks.task).not.toContain("写入 `设计.开场白与开场变量`");
|
||||
expect(sections.blocks.output).toContain("设计.开场白#场景短码");
|
||||
expect(sections.blocks.output).toContain('"开场变量"');
|
||||
expect(sections.blocks.output).not.toContain('"开场白候选"');
|
||||
expect(sections.blocks.output).not.toContain('"开场白全文"');
|
||||
expect(sections.blocks.output).toContain("present.v1");
|
||||
expect(sections.blocks.output).not.toContain('"自评"');
|
||||
|
||||
@@ -2,6 +2,7 @@ import { describe, expect, it } from "vitest";
|
||||
import {
|
||||
CREATION_PLAN_TAG,
|
||||
buildDictateSystemPrompt,
|
||||
formatDictateRoundTail,
|
||||
creationPlanModules,
|
||||
formatCreationPlanCapabilityCatalog,
|
||||
formatCreationPlanProgress,
|
||||
@@ -9,7 +10,9 @@ import {
|
||||
validateCreationPlan,
|
||||
type CreationPlan,
|
||||
} from "../src/dictate/index.js";
|
||||
import { moduleBrowseGroup } from "../src/dictate/creation-plan.js";
|
||||
import {
|
||||
catalogModulesForIntake,
|
||||
formatModuleCatalogForDictate,
|
||||
loadModuleCatalog,
|
||||
} from "../src/skills/creation-flow.js";
|
||||
@@ -65,6 +68,32 @@ describe("creation plan", () => {
|
||||
);
|
||||
});
|
||||
|
||||
it("groups dictate nodes by runtime role and omits the recipe-only merge", async () => {
|
||||
const catalog = await loadModuleCatalog("dialogue/world-simulator");
|
||||
expect(catalog).toBeTruthy();
|
||||
const dictate = catalogModulesForIntake(catalog!, "dictate");
|
||||
const ids = dictate.map((module) => module.id);
|
||||
expect(ids).not.toContain("narrative");
|
||||
expect(moduleBrowseGroup(dictate.find((module) => module.id === "creation-plan")!)).toBe(
|
||||
"plan",
|
||||
);
|
||||
expect(moduleBrowseGroup(dictate.find((module) => module.id === "narrative-guide")!)).toBe(
|
||||
"content",
|
||||
);
|
||||
expect(moduleBrowseGroup(dictate.find((module) => module.id === "reply-format")!)).toBe(
|
||||
"present",
|
||||
);
|
||||
expect(moduleBrowseGroup(dictate.find((module) => module.id === "zero-layer-status")!)).toBe(
|
||||
"present",
|
||||
);
|
||||
expect(moduleBrowseGroup(dictate.find((module) => module.id === "refine")!)).toBe(
|
||||
"program",
|
||||
);
|
||||
expect(new Set(dictate.map((module) => moduleBrowseGroup(module)))).toEqual(
|
||||
new Set(["plan", "content", "present", "program"]),
|
||||
);
|
||||
});
|
||||
|
||||
it("selects exactly one presentation chain", async () => {
|
||||
const catalog = await loadModuleCatalog("dialogue/world-simulator");
|
||||
expect(catalog).toBeTruthy();
|
||||
@@ -160,17 +189,19 @@ describe("creation plan", () => {
|
||||
expect(unusedSection).not.toContain("何时用：");
|
||||
|
||||
const oldFullGuide = formatModuleCatalogForDictate(catalog);
|
||||
const before = buildDictateSystemPrompt({
|
||||
const fixed = buildDictateSystemPrompt();
|
||||
const before = formatDictateRoundTail({
|
||||
capabilityCatalog: shortCatalog,
|
||||
recipeExamples: "- text-generator：文本生成案例",
|
||||
});
|
||||
const after = buildDictateSystemPrompt({ creationPlanState: progress });
|
||||
const after = formatDictateRoundTail({ creationPlanState: progress });
|
||||
expect(fixed).not.toContain("当前决策");
|
||||
expect(before).toContain("尚无本局创作方案");
|
||||
expect(after).toContain("当前决策");
|
||||
expect(after).toContain("收口前的核对清单");
|
||||
expect(after).not.toContain("旧配方案例目录");
|
||||
expect(after.length).toBeLessThan(
|
||||
buildDictateSystemPrompt({ moduleGuide: oldFullGuide }).length,
|
||||
formatDictateRoundTail({ moduleGuide: oldFullGuide }).length,
|
||||
);
|
||||
});
|
||||
|
||||
|
||||
@@ -1,9 +1,11 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { Blackboard } from "../src/blackboard/blackboard.js";
|
||||
import {
|
||||
applyDictatePlayBind,
|
||||
buildDictateContextOrder,
|
||||
bindDictateProductsToPlaySpec,
|
||||
collectDictateBindProducts,
|
||||
DICTATE_PLAY_BIND_INPUT_META,
|
||||
isDictateStyleBindTag,
|
||||
} from "../src/dictate/play-bind.js";
|
||||
import { DICTATE_ORDER_META_KEY } from "../src/dictate/types.js";
|
||||
@@ -215,4 +217,51 @@ describe("dictate play-bind", () => {
|
||||
expect(refs).toContain("上下文.角色态度");
|
||||
expect(refs).not.toContain("设计.变量映射");
|
||||
});
|
||||
|
||||
it("skips rebind on save when mounted products are unchanged", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({ tag: "用户.需求", content: "雨夜", source: "test" });
|
||||
const first = applyDictatePlayBind(board);
|
||||
expect(first.skipped).toBe(false);
|
||||
const orderUpdated = board.getLatestByTag("设计.上下文投影排序")?.updatedAt;
|
||||
const workerUpdated = board.getLatestByTag("设计.worker集")?.updatedAt;
|
||||
expect(board.getLatestByTag("设计.上下文投影排序")?.metadata?.[DICTATE_PLAY_BIND_INPUT_META]).toEqual(
|
||||
board.getLatestByTag("设计.worker集")?.metadata?.[DICTATE_PLAY_BIND_INPUT_META],
|
||||
);
|
||||
|
||||
const second = applyDictatePlayBind(board);
|
||||
expect(second.skipped).toBe(true);
|
||||
expect(second.changed).toBe(false);
|
||||
expect(board.getLatestByTag("设计.上下文投影排序")?.updatedAt).toBe(orderUpdated);
|
||||
expect(board.getLatestByTag("设计.worker集")?.updatedAt).toBe(workerUpdated);
|
||||
});
|
||||
|
||||
it("keeps a hand-edited order until a mounted product changes", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({ tag: "用户.需求", content: "雨夜", source: "test" });
|
||||
applyDictatePlayBind(board);
|
||||
|
||||
const orderItem = board.getLatestByTag("设计.上下文投影排序")!;
|
||||
const doc = JSON.parse(orderItem.content) as { brief?: string };
|
||||
doc.brief = "手改排序";
|
||||
const handEdited = JSON.stringify(doc, null, 2);
|
||||
board.write({
|
||||
tag: "设计.上下文投影排序",
|
||||
content: handEdited,
|
||||
source: "user",
|
||||
metadata: orderItem.metadata,
|
||||
});
|
||||
const kept = applyDictatePlayBind(board);
|
||||
expect(kept.skipped).toBe(true);
|
||||
expect(board.getContentByTag("设计.上下文投影排序")).toBe(handEdited);
|
||||
|
||||
board.write({ tag: "设计.叙事指南", content: "短句", source: "test" });
|
||||
const rebound = applyDictatePlayBind(board);
|
||||
expect(rebound.skipped).toBe(false);
|
||||
expect(rebound.boundTags).toContain("设计.叙事指南");
|
||||
const rewritten = JSON.parse(
|
||||
board.getContentByTag("设计.上下文投影排序") ?? "{}",
|
||||
) as { brief?: string };
|
||||
expect(rewritten.brief).not.toBe("手改排序");
|
||||
});
|
||||
});
|
||||
|
||||
@@ -32,7 +32,8 @@ describe("dictate read_module", () => {
|
||||
content: "{}",
|
||||
self_score: {
|
||||
dims: [
|
||||
{ name: "核心体验", score: 5 },
|
||||
{ name: "美学纲领", score: 5 },
|
||||
{ name: "交互范式", score: 5 },
|
||||
{ name: "背景与规则", score: 5 },
|
||||
],
|
||||
},
|
||||
@@ -82,7 +83,8 @@ describe("dictate read_module", () => {
|
||||
content: "{}",
|
||||
self_score: {
|
||||
dims: [
|
||||
{ name: "核心体验", score: 5 },
|
||||
{ name: "美学纲领", score: 5 },
|
||||
{ name: "交互范式", score: 5 },
|
||||
{ name: "背景与规则", score: 3, gap: "规则因果仍需确认" },
|
||||
],
|
||||
},
|
||||
@@ -205,7 +207,8 @@ describe("dictate read_module", () => {
|
||||
content: "{}",
|
||||
self_score: {
|
||||
dims: [
|
||||
{ name: "核心体验", score: 3 },
|
||||
{ name: "美学纲领", score: 3 },
|
||||
{ name: "交互范式", score: 5 },
|
||||
{ name: "背景与规则", score: 5 },
|
||||
],
|
||||
},
|
||||
@@ -226,7 +229,8 @@ describe("dictate read_module", () => {
|
||||
content: "{}",
|
||||
self_score: {
|
||||
dims: [
|
||||
{ name: "核心体验", score: 3, gap: "发生方式仍偏宽" },
|
||||
{ name: "美学纲领", score: 3, gap: "发生方式仍偏宽" },
|
||||
{ name: "交互范式", score: 5 },
|
||||
{ name: "背景与规则", score: 5 },
|
||||
],
|
||||
},
|
||||
@@ -254,13 +258,114 @@ describe("dictate read_module", () => {
|
||||
layer: "final",
|
||||
selfScore: {
|
||||
dims: expect.arrayContaining([
|
||||
expect.objectContaining({ name: "核心体验", score: 3 }),
|
||||
expect.objectContaining({ name: "美学纲领", score: 3 }),
|
||||
]),
|
||||
},
|
||||
});
|
||||
});
|
||||
});
|
||||
|
||||
describe("opening uses the reply contract", () => {
|
||||
const contract = JSON.stringify({
|
||||
schema: "context-fragment.v1",
|
||||
正文: {
|
||||
数据契约: {
|
||||
blocks: {
|
||||
body: { 类型: "markdown", 来源: "model_generated" },
|
||||
forum_feed: { 类型: "list", 来源: "model_generated" },
|
||||
},
|
||||
},
|
||||
frontend: { html: `<div id="app"></div>`, css: "", js: "" },
|
||||
},
|
||||
});
|
||||
const scores = {
|
||||
dims: [
|
||||
{ name: "开场必要", score: 5 },
|
||||
{ name: "前端符合", score: 5 },
|
||||
{ name: "状态自洽", score: 5 },
|
||||
{ name: "角色占位", score: 5 },
|
||||
{ name: "收口可切换", score: 5 },
|
||||
],
|
||||
};
|
||||
|
||||
it("returns the live contract when the opening module is read", async () => {
|
||||
const { catalog, moduleReadIndex } = await setup();
|
||||
const tags = new Map<string, string>([["设计.正文组成", contract]]);
|
||||
const driver: AgentDriver = {
|
||||
async run(input) {
|
||||
const outcome = await input.handleStep([
|
||||
{
|
||||
id: "r1",
|
||||
name: "read_module",
|
||||
arguments: JSON.stringify({ module_id: "opening-setup" }),
|
||||
},
|
||||
]);
|
||||
if (outcome.kind === "continue") {
|
||||
const payload = JSON.parse(outcome.results[0]!.content);
|
||||
expect(payload.reply_contract.model_blocks.map((block: { id: string }) => block.id)).toEqual([
|
||||
"body",
|
||||
"forum_feed",
|
||||
]);
|
||||
}
|
||||
return { iterations: 1, stop: { kind: "text", content: "已读" } };
|
||||
},
|
||||
};
|
||||
await runDictateTurn({
|
||||
llm: {} as never,
|
||||
dialogue: [],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex,
|
||||
driver,
|
||||
handlers: handlers(tags),
|
||||
});
|
||||
});
|
||||
|
||||
it("rejects an opening that only fills the old body shell", async () => {
|
||||
const { catalog, moduleReadIndex } = await setup();
|
||||
const tags = new Map<string, string>([["设计.正文组成", contract]]);
|
||||
const driver: AgentDriver = {
|
||||
async run(input) {
|
||||
await input.handleStep([
|
||||
{
|
||||
id: "r1",
|
||||
name: "read_module",
|
||||
arguments: JSON.stringify({ module_id: "opening-setup" }),
|
||||
},
|
||||
]);
|
||||
const outcome = await input.handleStep([
|
||||
{
|
||||
id: "i1",
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({
|
||||
position: "设计.开场白",
|
||||
content: JSON.stringify({
|
||||
schema: "present.v1",
|
||||
shell: "prose",
|
||||
blocks: { body: "@玩家推开门。" },
|
||||
}),
|
||||
self_score: scores,
|
||||
}),
|
||||
},
|
||||
]);
|
||||
if (outcome.kind === "continue") {
|
||||
const payload = JSON.parse(outcome.results[0]!.content);
|
||||
expect(payload.error).toContain("forum_feed");
|
||||
}
|
||||
return { iterations: 1, stop: { kind: "text", content: "未写入" } };
|
||||
},
|
||||
};
|
||||
const result = await runDictateTurn({
|
||||
llm: {} as never,
|
||||
dialogue: [],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex,
|
||||
driver,
|
||||
handlers: handlers(tags),
|
||||
});
|
||||
expect(result.wroteTags).toEqual([]);
|
||||
});
|
||||
});
|
||||
|
||||
function handlers(tags: Map<string, string>) {
|
||||
return {
|
||||
listProducts: () =>
|
||||
|
||||
@@ -0,0 +1,869 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import {
|
||||
DICTATE_MODULE_READ_STATE_TAG,
|
||||
buildDictateContext,
|
||||
parseDictateModuleReadState,
|
||||
runDictateTurn,
|
||||
type DictateModuleReadPayload,
|
||||
} from "../src/dictate/index.js";
|
||||
import type { LlmProvider } from "../src/llm/client.js";
|
||||
import type { ModuleCatalog } from "../src/skills/creation-flow.js";
|
||||
|
||||
function scriptedLlm(steps: unknown[]): LlmProvider {
|
||||
let index = 0;
|
||||
return {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools() {
|
||||
const body = steps[index++] ?? {};
|
||||
const select = index === 1;
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, select),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
function toToolCalls(body: unknown, select: boolean) {
|
||||
const row = body && typeof body === "object" ? (body as Record<string, unknown>) : {};
|
||||
if (select) {
|
||||
return [{ id: "select", name: "select_nodes", arguments: JSON.stringify(row) }];
|
||||
}
|
||||
const calls: Array<{ id: string; name: string; arguments: string }> = [];
|
||||
const products = row.products;
|
||||
if (Array.isArray(products)) {
|
||||
products.forEach((item, index) => {
|
||||
const rec = item && typeof item === "object" ? (item as Record<string, unknown>) : {};
|
||||
if ("tag" in rec || "position" in rec || "content" in rec || "正文" in rec) {
|
||||
calls.push({
|
||||
id: `ins-${index}`,
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({
|
||||
position: rec.tag ?? rec.position,
|
||||
content: rec.content ?? rec.正文 ?? "",
|
||||
...(rec.self_score !== undefined ? { self_score: rec.self_score } : {}),
|
||||
}),
|
||||
});
|
||||
return;
|
||||
}
|
||||
Object.entries(rec).forEach(([key, value], piece) => {
|
||||
const content =
|
||||
typeof value === "string"
|
||||
? value
|
||||
: value && typeof value === "object"
|
||||
? String((value as Record<string, unknown>).内容 ?? (value as Record<string, unknown>).正文 ?? (value as Record<string, unknown>).content ?? "")
|
||||
: "";
|
||||
calls.push({
|
||||
id: `ins-${index}-${piece}`,
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({ position: key, content }),
|
||||
});
|
||||
});
|
||||
});
|
||||
} else if (products && typeof products === "object") {
|
||||
Object.entries(products as Record<string, unknown>).forEach(([key, value], index) => {
|
||||
const content =
|
||||
typeof value === "string"
|
||||
? value
|
||||
: value && typeof value === "object"
|
||||
? String((value as Record<string, unknown>).正文 ?? (value as Record<string, unknown>).content ?? "")
|
||||
: "";
|
||||
calls.push({
|
||||
id: `ins-${index}`,
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({ position: key.startsWith("设计.") || key.startsWith("用户.") ? key : `设计.${key}`, content }),
|
||||
});
|
||||
});
|
||||
}
|
||||
for (const [key, value] of Object.entries(row)) {
|
||||
if (key === "products" || key === "reply" || key === "operations") continue;
|
||||
if (typeof value === "string" && (key.startsWith("设计.") || key.startsWith("用户."))) {
|
||||
calls.push({
|
||||
id: `extra-${key}`,
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({ position: key, content: value }),
|
||||
});
|
||||
}
|
||||
}
|
||||
if (Array.isArray(row.operations)) {
|
||||
row.operations.forEach((item, index) => {
|
||||
const rec = item && typeof item === "object" ? (item as Record<string, unknown>) : {};
|
||||
const nested =
|
||||
rec.arguments && typeof rec.arguments === "object"
|
||||
? (rec.arguments as Record<string, unknown>)
|
||||
: rec;
|
||||
const content = String(nested.content ?? "");
|
||||
if (!content.trim()) return;
|
||||
calls.push({
|
||||
id: `op-${index}`,
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({ position: nested.position ?? nested.tag, content }),
|
||||
});
|
||||
});
|
||||
}
|
||||
if (typeof row.reply === "string") {
|
||||
calls.push({
|
||||
id: "reply",
|
||||
name: "reply_user",
|
||||
arguments: JSON.stringify({ text: row.reply }),
|
||||
});
|
||||
}
|
||||
return calls;
|
||||
}
|
||||
|
||||
const catalog = {
|
||||
modules: [
|
||||
{
|
||||
id: "mood",
|
||||
name: "氛围",
|
||||
artifact: "设计.氛围",
|
||||
declaration: "细雨",
|
||||
},
|
||||
],
|
||||
} as ModuleCatalog;
|
||||
|
||||
const modulePayload: DictateModuleReadPayload = {
|
||||
module_id: "mood",
|
||||
name: "氛围",
|
||||
artifact: "设计.氛围",
|
||||
version: "v-test",
|
||||
task: "写氛围",
|
||||
principles: "短句",
|
||||
output: "一段话",
|
||||
score: "",
|
||||
score_dimensions: [],
|
||||
libraries: [],
|
||||
constraints: [],
|
||||
};
|
||||
|
||||
describe("dictate planned round", () => {
|
||||
it("puts earlier dialogue before products and the last user turn at the end", () => {
|
||||
const built = buildDictateContext({
|
||||
systemPrompt: "规则",
|
||||
products: [{ tag: "设计.氛围", content: "旧稿", order: 0 }],
|
||||
dialogue: [
|
||||
{ role: "user", text: "第一句" },
|
||||
{ role: "assistant", text: "记下了" },
|
||||
{ role: "user", text: "改成细雨" },
|
||||
],
|
||||
phase: "fill",
|
||||
injected: "#### task\n写氛围",
|
||||
tail: "方案进度",
|
||||
});
|
||||
const body = built.messages[1]?.content ?? "";
|
||||
expect(body.indexOf("必须调用 reply_user")).toBeLessThan(body.indexOf("第一句"));
|
||||
expect(body.indexOf("第一句")).toBeLessThan(body.indexOf("设计.氛围"));
|
||||
expect(body.indexOf("设计.氛围")).toBeLessThan(body.indexOf("写氛围"));
|
||||
expect(body.indexOf("写氛围")).toBeLessThan(body.indexOf("改成细雨"));
|
||||
expect(body.indexOf("方案进度")).toBeLessThan(body.indexOf("改成细雨"));
|
||||
expect(body.indexOf("写完下列节点后")).toBeLessThan(body.indexOf("写氛围"));
|
||||
expect(built.segmentChars.products).toBeGreaterThan(0);
|
||||
expect(built.segmentChars.rules).toBe("规则".length);
|
||||
|
||||
const select = buildDictateContext({
|
||||
systemPrompt: "规则",
|
||||
products: [{ tag: "设计.氛围", content: "旧稿" }],
|
||||
dialogue: [
|
||||
{ role: "user", text: "第一句" },
|
||||
{ role: "user", text: "改成细雨" },
|
||||
],
|
||||
phase: "select",
|
||||
tail: "方案进度",
|
||||
});
|
||||
const selectBody = select.messages[1]?.content ?? "";
|
||||
expect(selectBody).not.toContain("旧稿");
|
||||
expect(selectBody).toContain("第一句");
|
||||
expect(selectBody.indexOf("第一句")).toBeLessThan(selectBody.indexOf("改成细雨"));
|
||||
});
|
||||
|
||||
it("injects the spec, stamps the read credential, and rewrites only the missing tag", async () => {
|
||||
const products: Array<{ tag: string; content: string }> = [];
|
||||
const tags = new Map<string, string>();
|
||||
const result = await runDictateTurn({
|
||||
llm: scriptedLlm([
|
||||
{
|
||||
clarify: [],
|
||||
inserts: [{ tag: "设计.氛围", module_id: "mood" }],
|
||||
deletes: [],
|
||||
},
|
||||
{ products: [], reply: "还没写" },
|
||||
{
|
||||
products: [{ tag: "设计.氛围", content: "细雨" }],
|
||||
reply: "已写成细雨",
|
||||
},
|
||||
]),
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([
|
||||
["mood", modulePayload],
|
||||
["氛围", modulePayload],
|
||||
]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
const existing = products.find((item) => item.tag === tag);
|
||||
if (existing) existing.content = content;
|
||||
else products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result.wroteTags).toEqual(["设计.氛围"]);
|
||||
expect(products).toEqual([{ tag: "设计.氛围", content: "细雨" }]);
|
||||
expect(result.reply).toContain("已写成细雨");
|
||||
const state = parseDictateModuleReadState(tags.get(DICTATE_MODULE_READ_STATE_TAG));
|
||||
expect(state.modules.mood).toBe("v-test");
|
||||
});
|
||||
|
||||
it("stops after the select step when the plan is a clarification", async () => {
|
||||
let writes = 0;
|
||||
const result = await runDictateTurn({
|
||||
llm: scriptedLlm([
|
||||
{
|
||||
clarify: ["沿用冷色", "改成暖色"],
|
||||
inserts: [],
|
||||
deletes: [],
|
||||
},
|
||||
]),
|
||||
dialogue: [{ role: "user", text: "氛围还没定" }],
|
||||
handlers: {
|
||||
listProducts: () => [],
|
||||
writeProduct: () => {
|
||||
writes += 1;
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: () => undefined,
|
||||
writeTag: () => undefined,
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
expect(writes).toBe(0);
|
||||
expect(result.iterations).toBe(1);
|
||||
expect(result.reply).toContain("沿用冷色");
|
||||
expect(result.reply).toContain("改成暖色");
|
||||
});
|
||||
|
||||
it("says the returned fields when the select plan writes nothing", async () => {
|
||||
const result = await runDictateTurn({
|
||||
llm: scriptedLlm([{ note: "看世界" }]),
|
||||
dialogue: [{ role: "user", text: "超级英雄" }],
|
||||
handlers: {
|
||||
listProducts: () => [],
|
||||
writeProduct: () => undefined,
|
||||
deleteProduct: () => false,
|
||||
readTag: () => undefined,
|
||||
writeTag: () => undefined,
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
expect(result.wroteTags).toEqual([]);
|
||||
expect(result.iterations).toBe(1);
|
||||
expect(result.reply).toContain("note");
|
||||
expect(result.reply).toContain("没有写入产物");
|
||||
});
|
||||
|
||||
it("follows a select result with a fill request that carries specs and products", async () => {
|
||||
const seen: string[] = [];
|
||||
const steps = [
|
||||
{
|
||||
schema: "creation-plan.v1",
|
||||
items: [{ module_id: "mood", tier: "必须", why: "细雨", missing: "", instances: [] }],
|
||||
},
|
||||
{
|
||||
products: [{ tag: "设计.本局创作方案", content: "{\"schema\":\"creation-plan.v1\"}" }],
|
||||
reply: "已写方案",
|
||||
},
|
||||
];
|
||||
let index = 0;
|
||||
const llm: LlmProvider = {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
const body = steps[index++] ?? {};
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, index === 1),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
};
|
||||
const products = [{ tag: "设计.舞台", content: "已有舞台" }];
|
||||
await runDictateTurn({
|
||||
llm,
|
||||
dialogue: [{ role: "user", text: "超级英雄的世界" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([
|
||||
[
|
||||
"creation-plan",
|
||||
{
|
||||
...modulePayload,
|
||||
module_id: "creation-plan",
|
||||
artifact: "设计.本局创作方案",
|
||||
task: "规格任务正文",
|
||||
principles: "规格原则正文",
|
||||
},
|
||||
],
|
||||
]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
const existing = products.find((item) => item.tag === tag);
|
||||
if (existing) existing.content = content;
|
||||
else products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: () => undefined,
|
||||
writeTag: () => undefined,
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen.length).toBeGreaterThanOrEqual(2);
|
||||
expect(seen[0]).not.toContain("已有舞台");
|
||||
expect(seen[0]).not.toContain("规格任务正文");
|
||||
expect(seen[1]).toContain("已有舞台");
|
||||
expect(seen[1]).toContain("规格任务正文");
|
||||
expect(seen[1]).toContain("规格原则正文");
|
||||
expect(seen[1]).toContain("#### output");
|
||||
expect(seen[1]).toContain("设计.本局创作方案");
|
||||
});
|
||||
|
||||
it("treats operations as the selection, injects the node table, and writes every returned item", async () => {
|
||||
const seen: string[] = [];
|
||||
const steps = [
|
||||
{
|
||||
operations: [
|
||||
{ name: "insert", arguments: { position: "设计.氛围", module_id: "mood" } },
|
||||
],
|
||||
reply: "先写氛围",
|
||||
},
|
||||
{
|
||||
operations: [
|
||||
{ position: "设计.氛围", content: "细雨" },
|
||||
{ tag: "设计.旁白", content: "旁白一句" },
|
||||
],
|
||||
reply: "写了两份",
|
||||
},
|
||||
];
|
||||
let index = 0;
|
||||
const llm: LlmProvider = {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
const body = steps[index++] ?? {};
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, index === 1),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
};
|
||||
const products: Array<{ tag: string; content: string }> = [
|
||||
{ tag: "设计.舞台", content: "已有舞台" },
|
||||
];
|
||||
const tags = new Map<string, string>();
|
||||
const result = await runDictateTurn({
|
||||
llm,
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([["mood", modulePayload]]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
const existing = products.find((item) => item.tag === tag);
|
||||
if (existing) existing.content = content;
|
||||
else products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen.length).toBeGreaterThanOrEqual(2);
|
||||
expect(seen[0]).not.toContain("已有舞台");
|
||||
expect(seen[0]).not.toContain("写氛围");
|
||||
expect(seen[1]).toContain("本轮节点表");
|
||||
expect(seen[1]).toContain("写氛围");
|
||||
expect(seen[1]).toContain("已有舞台");
|
||||
expect(result.wroteTags).toEqual(expect.arrayContaining(["设计.氛围", "设计.旁白"]));
|
||||
expect(result.reply).toContain("写了两份");
|
||||
});
|
||||
|
||||
it("persists products named by module or keyed outside the products array", async () => {
|
||||
const resultProducts: Array<{ tag: string; content: string }> = [];
|
||||
const tags = new Map<string, string>();
|
||||
const result = await runDictateTurn({
|
||||
llm: scriptedLlm([
|
||||
{
|
||||
inserts: [{ tag: "设计.氛围", module_id: "mood" }],
|
||||
deletes: [],
|
||||
},
|
||||
{
|
||||
reply: "写了",
|
||||
products: { 氛围: { 正文: "细雨连绵" } },
|
||||
"设计.旁白": "旁白一句",
|
||||
},
|
||||
]),
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([["mood", modulePayload]]),
|
||||
handlers: {
|
||||
listProducts: () => resultProducts,
|
||||
writeProduct: (tag, content) => {
|
||||
resultProducts.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
expect(resultProducts.map((item) => item.tag).sort()).toEqual(["设计.旁白", "设计.氛围"]);
|
||||
expect(resultProducts.find((item) => item.tag === "设计.氛围")?.content).toBe("细雨连绵");
|
||||
expect(result.reply).toContain("写了");
|
||||
expect(result.reply).not.toContain("未给出");
|
||||
});
|
||||
|
||||
it("selects node ids first, then writes them with the user text and the node specs", async () => {
|
||||
const seen: string[] = [];
|
||||
const steps = [
|
||||
{ nodes: [{ module_id: "mood" }] },
|
||||
{
|
||||
products: [{ tag: "设计.氛围", content: "细雨" }],
|
||||
reply: "已写氛围。可能还缺主角设定。",
|
||||
},
|
||||
];
|
||||
let index = 0;
|
||||
const llm: LlmProvider = {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
const body = steps[index++] ?? {};
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, index === 1),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
};
|
||||
const products: Array<{ tag: string; content: string }> = [];
|
||||
const tags = new Map<string, string>();
|
||||
const result = await runDictateTurn({
|
||||
llm,
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([["mood", modulePayload]]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen).toHaveLength(2);
|
||||
expect(seen[0]).toContain("只做选题");
|
||||
expect(seen[0]).toContain("写成细雨");
|
||||
expect(seen[0]).not.toContain("declare_variable");
|
||||
expect(seen[1]).toContain("写成细雨");
|
||||
expect(seen[1]).toContain("写氛围");
|
||||
expect(seen[1]).toContain("第一轮选定的节点：mood");
|
||||
expect(seen[1]).toContain("还可能缺哪些节点");
|
||||
expect(result.wroteTags).toEqual(["设计.氛围"]);
|
||||
expect(result.reply).toContain("可能还缺主角设定");
|
||||
});
|
||||
|
||||
it("maps a user-prefixed tag onto the catalog artifact and reads Chinese content fields", async () => {
|
||||
const narration: DictateModuleReadPayload = {
|
||||
...modulePayload,
|
||||
module_id: "narration",
|
||||
name: "旁白",
|
||||
artifact: "设计.旁白",
|
||||
task: "写旁白",
|
||||
};
|
||||
const products: Array<{ tag: string; content: string }> = [];
|
||||
const tags = new Map<string, string>();
|
||||
const seen: string[] = [];
|
||||
const result = await runDictateTurn({
|
||||
llm: {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
const step = seen.length;
|
||||
const body =
|
||||
step === 1
|
||||
? { nodes: [{ tag: "用户.氛围" }, { module_id: "narration" }] }
|
||||
: {
|
||||
products: [{ "用户.氛围": { 内容: "细雨连绵" }, "设计.旁白": "旁白一句" }],
|
||||
reply: "写了两份",
|
||||
};
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, step === 1),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
},
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: {
|
||||
modules: [
|
||||
...catalog.modules,
|
||||
{ id: "narration", name: "旁白", artifact: "设计.旁白", declaration: "旁白" },
|
||||
],
|
||||
} as ModuleCatalog,
|
||||
moduleReadIndex: new Map([
|
||||
["mood", modulePayload],
|
||||
["narration", narration],
|
||||
]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen[1]).toContain("产物 tag：设计.氛围");
|
||||
expect(seen[1]).toContain("产物 tag：设计.旁白");
|
||||
expect(products.map((item) => item.tag).sort()).toEqual(["设计.旁白", "设计.氛围"]);
|
||||
expect(products.find((item) => item.tag === "设计.氛围")?.content).toBe("细雨连绵");
|
||||
expect(result.reply).not.toContain("未给出");
|
||||
});
|
||||
|
||||
it("asks for self_score dimensions before a write can land", async () => {
|
||||
const scored: DictateModuleReadPayload = {
|
||||
...modulePayload,
|
||||
score_dimensions: ["短句"],
|
||||
};
|
||||
const products: Array<{ tag: string; content: string }> = [];
|
||||
const tags = new Map<string, string>();
|
||||
const seen: string[] = [];
|
||||
const result = await runDictateTurn({
|
||||
llm: {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
const body =
|
||||
seen.length === 1
|
||||
? { nodes: [{ module_id: "mood" }] }
|
||||
: { products: [{ tag: "设计.氛围", content: "细雨" }], reply: "已落盘" };
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: toToolCalls(body, seen.length === 1),
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
},
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([["mood", scored]]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen[1]).toContain("self_score 维度（这次写入必填，name 必须一致）：短句");
|
||||
expect(seen[2]?.startsWith(seen[1] ?? "")).toBe(true);
|
||||
expect(seen[2]).toContain("还有未通过");
|
||||
expect(products).toHaveLength(0);
|
||||
expect(result.reply).toContain("self_score 必填");
|
||||
expect(result.reply).toContain("短句");
|
||||
expect(result.reply).not.toContain("未给出");
|
||||
});
|
||||
|
||||
it("keeps the node table and refuses to rewrite a tag that already passed", async () => {
|
||||
const narration: DictateModuleReadPayload = {
|
||||
...modulePayload,
|
||||
module_id: "narration",
|
||||
name: "旁白",
|
||||
artifact: "设计.旁白",
|
||||
task: "写旁白",
|
||||
score_dimensions: ["一句"],
|
||||
};
|
||||
const writes: string[] = [];
|
||||
const tags = new Map<string, string>();
|
||||
const seen: string[] = [];
|
||||
const rounds = [
|
||||
[{ id: "select", name: "select_nodes", arguments: JSON.stringify({ nodes: [{ module_id: "mood" }, { module_id: "narration" }] }) }],
|
||||
[
|
||||
{ id: "a", name: "insert", arguments: JSON.stringify({ position: "设计.氛围", content: "细雨" }) },
|
||||
{ id: "b", name: "insert", arguments: JSON.stringify({ position: "设计.旁白", content: "旁白" }) },
|
||||
{ id: "r1", name: "reply_user", arguments: JSON.stringify({ text: "都写了" }) },
|
||||
],
|
||||
[
|
||||
{ id: "a2", name: "insert", arguments: JSON.stringify({ position: "设计.氛围", content: "改掉细雨" }) },
|
||||
{
|
||||
id: "b2",
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({
|
||||
position: "设计.旁白",
|
||||
content: "旁白",
|
||||
self_score: { dims: [{ name: "一句", score: 4 }] },
|
||||
}),
|
||||
},
|
||||
],
|
||||
[{ id: "r2", name: "reply_user", arguments: JSON.stringify({ text: "旁白补上了" }) }],
|
||||
];
|
||||
let index = 0;
|
||||
await runDictateTurn({
|
||||
llm: {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: rounds[index++] ?? [],
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
},
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: {
|
||||
modules: [
|
||||
...catalog.modules,
|
||||
{ id: "narration", name: "旁白", artifact: "设计.旁白", declaration: "旁白" },
|
||||
],
|
||||
} as ModuleCatalog,
|
||||
moduleReadIndex: new Map([
|
||||
["mood", modulePayload],
|
||||
["narration", narration],
|
||||
]),
|
||||
handlers: {
|
||||
listProducts: () => [],
|
||||
writeProduct: (tag, content) => {
|
||||
writes.push(`${tag}:${content}`);
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen[2]?.startsWith(seen[1] ?? "")).toBe(true);
|
||||
expect(seen[2]).toContain("写旁白");
|
||||
expect(writes).toEqual(["设计.氛围:细雨", "设计.旁白:旁白"]);
|
||||
});
|
||||
|
||||
it("puts reply_user guidance early in the fill prompt and does not mention a fallback reply step", async () => {
|
||||
const seen: string[] = [];
|
||||
const products: Array<{ tag: string; content: string }> = [];
|
||||
const rounds = [
|
||||
[{ id: "select", name: "select_nodes", arguments: JSON.stringify({ nodes: [{ module_id: "mood" }] }) }],
|
||||
[
|
||||
{ id: "a", name: "insert", arguments: JSON.stringify({ position: "设计.氛围", content: "细雨" }) },
|
||||
{ id: "r", name: "reply_user", arguments: JSON.stringify({ text: "细雨留下了" }) },
|
||||
],
|
||||
];
|
||||
let index = 0;
|
||||
const tags = new Map<string, string>();
|
||||
await runDictateTurn({
|
||||
llm: {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: rounds[index++] ?? [],
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
},
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([["mood", modulePayload]]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content) => {
|
||||
products.push({ tag, content });
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(seen).toHaveLength(2);
|
||||
const fill = seen[1] ?? "";
|
||||
expect(fill.indexOf("填写收尾")).toBeLessThan(fill.indexOf("写入规则"));
|
||||
expect(fill.indexOf("本步收尾")).toBeLessThan(fill.indexOf("更早的对话"));
|
||||
expect(fill.indexOf("同一次请求必须调用 reply_user")).toBeLessThan(fill.indexOf("写氛围"));
|
||||
expect(fill).toContain("分数一般或还有更好写法的缺口");
|
||||
expect(fill).toContain("下一步可以提及哪些");
|
||||
expect(fill).not.toContain("本轮刚落盘");
|
||||
expect(fill).not.toContain("你只生成给用户看的综合回复");
|
||||
});
|
||||
|
||||
it("silently asks for reply_user with only this round's products when fill omitted it", async () => {
|
||||
const products: Array<{ tag: string; content: string }> = [
|
||||
{ tag: "设计.舞台", content: "旧舞台不应进入回复请求" },
|
||||
];
|
||||
const seen: string[] = [];
|
||||
const choices: Array<unknown> = [];
|
||||
let index = 0;
|
||||
const tags = new Map<string, string>();
|
||||
const result = await runDictateTurn({
|
||||
llm: {
|
||||
async complete() {
|
||||
throw new Error("对话落盘走工具，不走纯文本完成");
|
||||
},
|
||||
async completeWithTools(messages, options) {
|
||||
seen.push(messages.map((message) => message.content ?? "").join("\n"));
|
||||
choices.push(options?.toolChoice);
|
||||
const step = index++;
|
||||
if (step === 0) {
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: [
|
||||
{
|
||||
id: "select",
|
||||
name: "select_nodes",
|
||||
arguments: JSON.stringify({ nodes: [{ module_id: "mood" }] }),
|
||||
},
|
||||
],
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
}
|
||||
if (step === 1) {
|
||||
return {
|
||||
content: "细雨留下了，接下来可以定主角怎么进城。",
|
||||
toolCalls: [
|
||||
{
|
||||
id: "a",
|
||||
name: "insert",
|
||||
arguments: JSON.stringify({
|
||||
position: "设计.氛围",
|
||||
content: "细雨",
|
||||
self_score: {
|
||||
dims: [{ name: "短句", score: 3, gap: "还缺进城方式", options: ["步行", "马车"] }],
|
||||
},
|
||||
}),
|
||||
},
|
||||
],
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
}
|
||||
return {
|
||||
content: null,
|
||||
toolCalls: [
|
||||
{
|
||||
id: "r",
|
||||
name: "reply_user",
|
||||
arguments: JSON.stringify({ text: "细雨已落盘，接下来可以定进城方式。" }),
|
||||
},
|
||||
],
|
||||
usage: { promptTokens: 10, completionTokens: 4, totalTokens: 14 },
|
||||
};
|
||||
},
|
||||
},
|
||||
dialogue: [{ role: "user", text: "写成细雨" }],
|
||||
moduleCatalog: catalog,
|
||||
moduleReadIndex: new Map([
|
||||
["mood", { ...modulePayload, score_dimensions: ["短句"] }],
|
||||
]),
|
||||
handlers: {
|
||||
listProducts: () => products,
|
||||
writeProduct: (tag, content, _order, metadata) => {
|
||||
products.push({
|
||||
tag,
|
||||
content,
|
||||
...(metadata?.selfScore ? { selfScore: metadata.selfScore } : {}),
|
||||
});
|
||||
},
|
||||
deleteProduct: () => false,
|
||||
readTag: (tag) => tags.get(tag),
|
||||
writeTag: (tag, content) => {
|
||||
tags.set(tag, content);
|
||||
},
|
||||
deleteTag: () => false,
|
||||
clearDialogue: () => undefined,
|
||||
lookupInsertFeedback: (tag) =>
|
||||
tag === "设计.氛围"
|
||||
? `# 本轮关注 · 氛围\n\n## 写入用户回复时做什么\n1. 自然回应\n3. 结合 self_score 说明最强处与关键缺口\n\n## 自评关注\n- 短句`
|
||||
: undefined,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result.reply).toBe("细雨已落盘，接下来可以定进城方式。");
|
||||
expect(result.iterations).toBe(3);
|
||||
expect(seen).toHaveLength(3);
|
||||
const replyReq = seen[2] ?? "";
|
||||
expect(replyReq).toContain("本轮刚落盘");
|
||||
expect(replyReq).toContain("细雨");
|
||||
expect(replyReq).toContain("self_score");
|
||||
expect(replyReq).toContain("还缺进城方式");
|
||||
expect(replyReq).toContain("本轮关注");
|
||||
expect(replyReq).toContain("结合 self_score 说明最强处与关键缺口");
|
||||
expect(replyReq).toContain("每个缺口写三样");
|
||||
expect(replyReq).toContain("2～3 个可以接着说的选项");
|
||||
expect(replyReq).not.toContain("旧舞台不应进入回复请求");
|
||||
expect(replyReq).not.toContain("#### task");
|
||||
expect(choices[2]).toEqual({ type: "function", function: { name: "reply_user" } });
|
||||
});
|
||||
});
|
||||
+53
-15
@@ -2,6 +2,7 @@ import { describe, expect, it } from "vitest";
|
||||
import {
|
||||
buildDictateMessages,
|
||||
buildDictateSystemPrompt,
|
||||
formatDictateRoundTail,
|
||||
buildDictateInsertFeedbackIndex,
|
||||
defaultDictateOrder,
|
||||
extractDictateDialogue,
|
||||
@@ -47,7 +48,7 @@ describe("dictate context", () => {
|
||||
});
|
||||
const body = messages[1]?.content ?? "";
|
||||
expect(body).toContain("冷色短句");
|
||||
expect(body).toContain("尚无对话");
|
||||
expect(body).toContain("尚无更早对话");
|
||||
});
|
||||
|
||||
it("extracts only user + dictate_reply turns", () => {
|
||||
@@ -73,8 +74,9 @@ describe("dictate context", () => {
|
||||
expect(isDictateModeValue("recipe")).toBe(false);
|
||||
});
|
||||
|
||||
it("system prompt starts from a plan and keeps generic write tools", () => {
|
||||
const prompt = buildDictateSystemPrompt({
|
||||
it("system prompt keeps stable rules; catalog and plan progress sit in the tail", () => {
|
||||
const prompt = buildDictateSystemPrompt();
|
||||
const tail = formatDictateRoundTail({
|
||||
recipeName: "数据化跑团体验",
|
||||
recipeBrief:
|
||||
"设定 + 真值 + 分档映射\n怎么做：\n- declare_variable\n- declare_map",
|
||||
@@ -90,17 +92,31 @@ describe("dictate context", () => {
|
||||
expect(prompt).toContain("declare_map");
|
||||
expect(prompt).toContain("remove_map");
|
||||
expect(prompt).toContain("同 position 再 insert");
|
||||
expect(prompt).toContain("数据化跑团体验");
|
||||
expect(prompt).not.toContain("数据化跑团体验");
|
||||
expect(prompt).toMatch(/order/);
|
||||
expect(prompt).toContain("【能力 · 何时落盘】");
|
||||
expect(prompt).toContain("设计.美学纲领与交互范式");
|
||||
expect(prompt).not.toContain("【能力 · 何时落盘】");
|
||||
expect(prompt).toContain("通常 2～3 个");
|
||||
expect(prompt).toContain("给 2～3 个选项来问");
|
||||
expect(prompt).toContain("不是 DAG");
|
||||
expect(prompt).toContain("不是用户已选路径");
|
||||
expect(prompt).not.toContain("不是用户已选路径");
|
||||
expect(prompt).toContain("自然回应");
|
||||
expect(prompt.indexOf("填写收尾")).toBeLessThan(prompt.indexOf("写入规则"));
|
||||
expect(prompt).not.toContain("体验锚定（程序打分）");
|
||||
expect(prompt).not.toContain("## 本轮落什么");
|
||||
expect(tail).toContain("数据化跑团体验");
|
||||
expect(tail).toContain("【能力 · 何时落盘】");
|
||||
expect(tail).toContain("不是用户已选路径");
|
||||
expect(tail).toContain("尚无本局创作方案");
|
||||
});
|
||||
|
||||
it("puts the fill reply_user duty before plan progress in the tail", () => {
|
||||
const fillTail = formatDictateRoundTail({
|
||||
phase: "fill",
|
||||
creationPlanState: "【当前决策】\n- mood：已落",
|
||||
checklist: "本轮写入：设计.氛围",
|
||||
});
|
||||
expect(fillTail.indexOf("必须调用 reply_user")).toBeLessThan(fillTail.indexOf("当前决策"));
|
||||
expect(fillTail.indexOf("必须调用 reply_user")).toBeLessThan(fillTail.indexOf("本轮写入"));
|
||||
});
|
||||
|
||||
it("formats module catalog for dictate with land tags and when", async () => {
|
||||
@@ -172,6 +188,12 @@ describe("dictate context", () => {
|
||||
expect(
|
||||
resolveModuleForInsertTag("设计.开场白", catalog!)?.id,
|
||||
).toBe("opening-setup");
|
||||
expect(
|
||||
resolveModuleForInsertTag("设计.开场白#opening-cycle2", catalog!)?.id,
|
||||
).toBe("opening-setup");
|
||||
expect(
|
||||
resolveModuleForInsertTag("设计.开场白与开场变量", catalog!)?.id,
|
||||
).toBe("opening-setup");
|
||||
expect(
|
||||
resolveModuleForInsertTag("设计.具体实例#lei-ying", catalog!)?.id,
|
||||
).toBe("concrete-instances");
|
||||
@@ -244,31 +266,47 @@ describe("dictate context", () => {
|
||||
expect("error" in overwriteBare).toBe(true);
|
||||
|
||||
const openingFamilies = new Set(["设计.开场白", "设计.主角设定"]);
|
||||
const firstOpen = resolveRepeatableInsertTag({
|
||||
const bareOpen = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白",
|
||||
content: "@玩家推开门。",
|
||||
existingTags: [],
|
||||
repeatableFamilies: openingFamilies,
|
||||
});
|
||||
expect("tag" in firstOpen && firstOpen.tag).toBe("设计.开场白");
|
||||
const overwriteFirst = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白",
|
||||
content: "@玩家改写第一幕。",
|
||||
expect("error" in bareOpen).toBe(true);
|
||||
const namedFromBundle = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白与开场变量",
|
||||
content: JSON.stringify({
|
||||
schema: "present.v1",
|
||||
id: "opening-cycle2",
|
||||
blocks: { body: "@玩家推开门。" },
|
||||
}),
|
||||
existingTags: ["设计.开场白"],
|
||||
repeatableFamilies: openingFamilies,
|
||||
});
|
||||
expect("tag" in overwriteFirst && overwriteFirst.tag).toBe("设计.开场白");
|
||||
expect("tag" in namedFromBundle && namedFromBundle.tag).toBe(
|
||||
"设计.开场白#opening-cycle2",
|
||||
);
|
||||
const keepNamed = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白#opening-cycle2",
|
||||
content: JSON.stringify({
|
||||
schema: "present.v1",
|
||||
blocks: { body: "@玩家改写这一幕。" },
|
||||
}),
|
||||
existingTags: ["设计.开场白", "设计.开场白与开场变量", "设计.开场白#opening-cycle2"],
|
||||
repeatableFamilies: openingFamilies,
|
||||
});
|
||||
expect("tag" in keepNamed && keepNamed.tag).toBe("设计.开场白#opening-cycle2");
|
||||
const secondOpen = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白#gate",
|
||||
content: "@玩家在闸机前。",
|
||||
existingTags: ["设计.开场白"],
|
||||
existingTags: ["设计.开场白#opening-cycle2"],
|
||||
repeatableFamilies: openingFamilies,
|
||||
});
|
||||
expect("tag" in secondOpen && secondOpen.tag).toBe("设计.开场白#gate");
|
||||
const bareOpenConflict = resolveRepeatableInsertTag({
|
||||
tag: "设计.开场白",
|
||||
content: "又写一条",
|
||||
existingTags: ["设计.开场白", "设计.开场白#gate"],
|
||||
existingTags: ["设计.开场白#opening-cycle2", "设计.开场白#gate"],
|
||||
repeatableFamilies: openingFamilies,
|
||||
});
|
||||
expect("error" in bareOpenConflict).toBe(true);
|
||||
|
||||
@@ -18,6 +18,7 @@ describe("display-labels", () => {
|
||||
it("maps creation intake modes and recipe display names", () => {
|
||||
expect(displayCreationModeLabel("recipe")).toBe("工序编排");
|
||||
expect(displayCreationModeLabel("dictate")).toBe("对话落盘");
|
||||
expect(displayCreationModeLabel("listen_serial")).toBe("听书连载");
|
||||
expect(displaySkillPackLabel("world-simulator")).toBe("回合推演");
|
||||
expect(displaySkillPackLabel("快穿短局")).toBe("快穿短局");
|
||||
});
|
||||
|
||||
@@ -0,0 +1,53 @@
|
||||
import { describe, expect, it, beforeEach, afterEach } from "vitest";
|
||||
import {
|
||||
applyPlayAuxLlmBubbles,
|
||||
isPlayAuxLlmBubblesEnabled,
|
||||
isPlayAuxLlmOutput,
|
||||
} from "../web/feed-prefs.js";
|
||||
|
||||
describe("play aux LLM bubbles", () => {
|
||||
beforeEach(() => {
|
||||
applyPlayAuxLlmBubbles(false);
|
||||
});
|
||||
afterEach(() => {
|
||||
applyPlayAuxLlmBubbles(false);
|
||||
});
|
||||
|
||||
it("stores the more-menu preference", () => {
|
||||
expect(isPlayAuxLlmBubblesEnabled()).toBe(false);
|
||||
applyPlayAuxLlmBubbles(true);
|
||||
expect(isPlayAuxLlmBubblesEnabled()).toBe(true);
|
||||
});
|
||||
|
||||
it("treats auditor worker_output as aux and world-simulator as main", () => {
|
||||
const play = { lifecycleStage: "play" };
|
||||
expect(
|
||||
isPlayAuxLlmOutput(
|
||||
{ role: "assistant", kind: "worker_output", actor: "auditor" },
|
||||
play,
|
||||
),
|
||||
).toBe(true);
|
||||
expect(
|
||||
isPlayAuxLlmOutput(
|
||||
{
|
||||
role: "assistant",
|
||||
kind: "worker_output",
|
||||
actor: "world-simulator",
|
||||
title: "主世界层 · 产出",
|
||||
},
|
||||
play,
|
||||
),
|
||||
).toBe(false);
|
||||
expect(
|
||||
isPlayAuxLlmOutput(
|
||||
{
|
||||
role: "assistant",
|
||||
kind: "worker_output",
|
||||
actor: "narrator",
|
||||
title: "叙事转述 · 产出",
|
||||
},
|
||||
play,
|
||||
),
|
||||
).toBe(false);
|
||||
});
|
||||
});
|
||||
@@ -13,7 +13,7 @@ import {
|
||||
sanitizeReasoningEffort,
|
||||
sanitizeVerbosity,
|
||||
} from "../src/llm/generation-compat.js";
|
||||
import { buildRequestBody } from "../src/llm/client.js";
|
||||
import { buildRequestBody, extractMessageParts, openRouterSessionId } from "../src/llm/client.js";
|
||||
import type { LlmConfig } from "../src/config/env.js";
|
||||
|
||||
const stGeneration = {
|
||||
@@ -410,6 +410,24 @@ describe("fetchWithGenerationCompat", () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe("extractMessageParts", () => {
|
||||
it("keeps reply_user when arguments are an object and id is missing", () => {
|
||||
const parts = extractMessageParts({
|
||||
content: [{ type: "text", text: "同一次请求里的说明" }],
|
||||
tool_calls: [
|
||||
{
|
||||
type: "function",
|
||||
function: { name: "reply_user", arguments: { text: "形成了气氛" } },
|
||||
},
|
||||
],
|
||||
});
|
||||
expect(parts.toolCalls).toEqual([
|
||||
{ id: "call-0", name: "reply_user", arguments: JSON.stringify({ text: "形成了气氛" }) },
|
||||
]);
|
||||
expect(parts.content).toBe("同一次请求里的说明");
|
||||
});
|
||||
});
|
||||
|
||||
describe("buildRequestBody", () => {
|
||||
const messages = [{ role: "user" as const, content: "hi" }];
|
||||
|
||||
@@ -597,6 +615,24 @@ describe("buildRequestBody", () => {
|
||||
expect(body.messages).toBe(messages);
|
||||
});
|
||||
|
||||
it("writes session_id only for OpenRouter", () => {
|
||||
const openrouter: LlmConfig = {
|
||||
baseUrl: "https://openrouter.ai/api/v1",
|
||||
apiKey: "k",
|
||||
model: "google/gemini-3.8-flash",
|
||||
};
|
||||
const direct: LlmConfig = {
|
||||
baseUrl: "https://api.deepseek.com/v1",
|
||||
apiKey: "k",
|
||||
model: "deepseek-v4-pro",
|
||||
};
|
||||
const id = "9b76dcb6-0703-4b0d-ba75-faa985b532b4";
|
||||
expect(buildRequestBody(openrouter, messages, { sessionId: id }).session_id).toBe(id);
|
||||
expect(buildRequestBody(direct, messages, { sessionId: id }).session_id).toBeUndefined();
|
||||
expect(buildRequestBody(openrouter, messages, {}).session_id).toBeUndefined();
|
||||
expect(openRouterSessionId(openrouter.baseUrl, ` ${"a".repeat(300)} `)).toHaveLength(256);
|
||||
});
|
||||
|
||||
it("does not forward top_k from an imported ST preset", () => {
|
||||
const config: LlmConfig = {
|
||||
baseUrl: "https://example.test/v1",
|
||||
|
||||
@@ -0,0 +1,117 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { Blackboard } from "../src/blackboard/blackboard.js";
|
||||
import {
|
||||
applyRedirectCut,
|
||||
isConfirmText,
|
||||
isListenSerialModeValue,
|
||||
LISTEN_SERIAL_MODE_VALUE,
|
||||
makeOutlineItems,
|
||||
markListenConsumed,
|
||||
readListenState,
|
||||
toListenView,
|
||||
unreadChars,
|
||||
writeListenBrief,
|
||||
writeListenOutline,
|
||||
writeListenPhase,
|
||||
writeListenSections,
|
||||
} from "../src/listen-serial/index.js";
|
||||
import { ensureBufferTarget } from "../src/listen-serial/state.js";
|
||||
|
||||
describe("listen-serial state", () => {
|
||||
it("recognizes mode value", () => {
|
||||
expect(isListenSerialModeValue(LISTEN_SERIAL_MODE_VALUE)).toBe(true);
|
||||
expect(isListenSerialModeValue("dictate")).toBe(false);
|
||||
});
|
||||
|
||||
it("round-trips brief/outline/sections on blackboard", () => {
|
||||
const board = new Blackboard();
|
||||
ensureBufferTarget(board);
|
||||
writeListenPhase(board, "running");
|
||||
writeListenBrief(board, {
|
||||
thrill: "身份反差打脸",
|
||||
world: "现代都市；主角是隐退龙王",
|
||||
structure: "三段式",
|
||||
});
|
||||
const outline = makeOutlineItems(["被嘲", "亮身份", "全员后悔"], "全员后悔");
|
||||
writeListenOutline(board, outline);
|
||||
writeListenSections(board, [
|
||||
{
|
||||
id: "s1",
|
||||
outlineId: outline[0]!.id,
|
||||
title: "被嘲",
|
||||
text: "甲".repeat(100),
|
||||
summary: "被嘲讽",
|
||||
consumedChars: 0,
|
||||
},
|
||||
]);
|
||||
const state = readListenState(board);
|
||||
expect(state.phase).toBe("running");
|
||||
expect(state.thrill).toContain("身份反差");
|
||||
expect(state.outline).toHaveLength(3);
|
||||
expect(state.outline[2]?.payoff).toBe(true);
|
||||
expect(unreadChars(state)).toBe(100);
|
||||
const view = toListenView(state);
|
||||
expect(view.unreadChars).toBe(100);
|
||||
});
|
||||
|
||||
it("applyRedirectCut drops unconsumed sections and keeps consumed", () => {
|
||||
const outline = makeOutlineItems(["A", "B", "C"], "C");
|
||||
outline[0]!.status = "done";
|
||||
const state = {
|
||||
phase: "running" as const,
|
||||
thrill: "打脸",
|
||||
world: "都市",
|
||||
structure: "三段式",
|
||||
outline,
|
||||
sections: [
|
||||
{
|
||||
id: "s1",
|
||||
outlineId: outline[0]!.id,
|
||||
title: "A",
|
||||
text: "已读完的一节",
|
||||
summary: "A小结",
|
||||
consumedChars: 6,
|
||||
},
|
||||
{
|
||||
id: "s2",
|
||||
outlineId: "pending-x",
|
||||
title: "B",
|
||||
text: "整节未读应丢掉",
|
||||
summary: "B小结",
|
||||
consumedChars: 0,
|
||||
},
|
||||
],
|
||||
summaries: ["A小结", "B小结"],
|
||||
bufferTargetChars: 4000,
|
||||
consumedChars: 6,
|
||||
};
|
||||
const cut = applyRedirectCut(state, ["主角反杀", "对面后悔"]);
|
||||
expect(cut.sections).toHaveLength(1);
|
||||
expect(cut.sections[0]?.title).toBe("A");
|
||||
expect(cut.outline.filter((o) => o.status === "pending")).toHaveLength(2);
|
||||
expect(cut.outline.some((o) => o.title === "主角反杀")).toBe(true);
|
||||
});
|
||||
|
||||
it("markListenConsumed advances pointer", () => {
|
||||
const board = new Blackboard();
|
||||
writeListenSections(board, [
|
||||
{
|
||||
id: "s1",
|
||||
outlineId: "o1",
|
||||
title: "一",
|
||||
text: "一二三四五",
|
||||
summary: "一",
|
||||
consumedChars: 0,
|
||||
},
|
||||
]);
|
||||
const next = markListenConsumed(board, 3);
|
||||
expect(next.consumedChars).toBe(3);
|
||||
expect(next.sections[0]?.consumedChars).toBe(3);
|
||||
});
|
||||
|
||||
it("detects confirm phrases", () => {
|
||||
expect(isConfirmText("确认")).toBe(true);
|
||||
expect(isConfirmText("开始！")).toBe(true);
|
||||
expect(isConfirmText("主角反过来打脸")).toBe(false);
|
||||
});
|
||||
});
|
||||
@@ -0,0 +1,162 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { Blackboard } from "../src/blackboard/blackboard.js";
|
||||
import {
|
||||
applyDictatePlayBind,
|
||||
bindDictateProductsToPlaySpec,
|
||||
} from "../src/dictate/play-bind.js";
|
||||
import {
|
||||
applyMaintenanceStrategyToPlaySlots,
|
||||
normalizeMaintenanceAdvice,
|
||||
parseMaintenanceAdviceFromVariableDesign,
|
||||
resolveMaintenanceStrategyFromBoard,
|
||||
strategyFromAdvice,
|
||||
MAINTENANCE_STRATEGY_TAG,
|
||||
} from "../src/skills/maintenance-strategy.js";
|
||||
import { defaultPlaySlots } from "../src/skills/play-slots.js";
|
||||
import { parseWorkerSetYaml } from "../src/skills/worker-set-parse.js";
|
||||
import { VARIABLE_DESIGN_TAG } from "../src/skills/variable-design-compile.js";
|
||||
|
||||
function designWithAdvice(advice: string) {
|
||||
return JSON.stringify({
|
||||
schema: "context-fragment.v1",
|
||||
技能: "变量设计与更新规则",
|
||||
正文: {
|
||||
真值: [
|
||||
{
|
||||
id: "affection",
|
||||
名称: "好感",
|
||||
类型: "number",
|
||||
初始化: { 默认值: 0 },
|
||||
公开性: "public",
|
||||
},
|
||||
],
|
||||
维护复杂度: {
|
||||
建议: advice,
|
||||
依据: ["测试依据"],
|
||||
},
|
||||
},
|
||||
});
|
||||
}
|
||||
|
||||
describe("maintenance-strategy", () => {
|
||||
it("normalizes advice aliases", () => {
|
||||
expect(normalizeMaintenanceAdvice("auditor_bundle")).toBe("auditor_bundle");
|
||||
expect(normalizeMaintenanceAdvice("旁观")).toBe("auditor_bundle");
|
||||
expect(normalizeMaintenanceAdvice("inline_tool")).toBe("inline_tool");
|
||||
expect(normalizeMaintenanceAdvice("内联")).toBe("inline_tool");
|
||||
expect(normalizeMaintenanceAdvice("inline_structured")).toBe(
|
||||
"inline_structured",
|
||||
);
|
||||
expect(normalizeMaintenanceAdvice("none")).toBe("none");
|
||||
});
|
||||
|
||||
it("parses variable-design maintenance box", () => {
|
||||
const parsed = parseMaintenanceAdviceFromVariableDesign(
|
||||
designWithAdvice("auditor_bundle"),
|
||||
);
|
||||
expect(parsed?.advice).toBe("auditor_bundle");
|
||||
expect(parsed?.依据).toContain("测试依据");
|
||||
});
|
||||
|
||||
it("defaults to auditor_bundle when truths exist but advice missing", () => {
|
||||
const raw = JSON.stringify({
|
||||
schema: "context-fragment.v1",
|
||||
正文: { 真值: [{ id: "a", 类型: "number", 初始化: { 默认值: 1 } }] },
|
||||
});
|
||||
expect(parseMaintenanceAdviceFromVariableDesign(raw)?.advice).toBe(
|
||||
"auditor_bundle",
|
||||
);
|
||||
});
|
||||
|
||||
it("maps advice to auditor/inline flags at strategyFromAdvice level", () => {
|
||||
expect(strategyFromAdvice("auditor_bundle")).toMatchObject({
|
||||
auditor: true,
|
||||
inline: false,
|
||||
});
|
||||
expect(strategyFromAdvice("inline_tool")).toMatchObject({
|
||||
auditor: false,
|
||||
inline: true,
|
||||
});
|
||||
expect(strategyFromAdvice("none")).toMatchObject({
|
||||
auditor: false,
|
||||
inline: false,
|
||||
});
|
||||
});
|
||||
|
||||
it("forces auditor_bundle when truths exist even if design said inline", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({
|
||||
tag: VARIABLE_DESIGN_TAG,
|
||||
content: designWithAdvice("inline_tool"),
|
||||
source: "test",
|
||||
});
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(board);
|
||||
expect(strategy.advice).toBe("auditor_bundle");
|
||||
expect(strategy.auditor).toBe(true);
|
||||
expect(strategy.inline).toBe(false);
|
||||
expect(strategy.依据?.some((x) => x.includes("inline_tool"))).toBe(true);
|
||||
});
|
||||
|
||||
it("keeps auditor off when there are no truths", () => {
|
||||
const board = new Blackboard();
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(board);
|
||||
expect(strategy.advice).toBe("none");
|
||||
expect(strategy.auditor).toBe(false);
|
||||
});
|
||||
|
||||
it("overrides play_slots.auditor from strategy", () => {
|
||||
const slots = applyMaintenanceStrategyToPlaySlots(defaultPlaySlots(), {
|
||||
schema: "maintenance-strategy.v1",
|
||||
advice: "auditor_bundle",
|
||||
auditor: true,
|
||||
inline: false,
|
||||
source: VARIABLE_DESIGN_TAG,
|
||||
});
|
||||
expect(slots.auditor).toBe(true);
|
||||
expect(slots.gm).toBe(true);
|
||||
});
|
||||
});
|
||||
|
||||
describe("play-bind applies maintenance strategy", () => {
|
||||
it("turns auditor on for auditor_bundle", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({
|
||||
tag: VARIABLE_DESIGN_TAG,
|
||||
content: designWithAdvice("auditor_bundle"),
|
||||
source: "test",
|
||||
});
|
||||
board.write({
|
||||
tag: "设计.世界蓝图",
|
||||
content: "舞台",
|
||||
source: "test",
|
||||
});
|
||||
const result = bindDictateProductsToPlaySpec(board);
|
||||
const parsed = parseWorkerSetYaml(result.workerSetJson);
|
||||
expect(parsed?.play_slots?.auditor).toBe(true);
|
||||
const strategy = resolveMaintenanceStrategyFromBoard(board);
|
||||
expect(strategy.advice).toBe("auditor_bundle");
|
||||
});
|
||||
|
||||
it("still opens auditor when design said inline_tool", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({
|
||||
tag: VARIABLE_DESIGN_TAG,
|
||||
content: designWithAdvice("inline_tool"),
|
||||
source: "test",
|
||||
});
|
||||
board.write({
|
||||
tag: "设计.世界蓝图",
|
||||
content: "舞台",
|
||||
source: "test",
|
||||
});
|
||||
applyDictatePlayBind(board);
|
||||
const worker = parseWorkerSetYaml(
|
||||
board.getContentByTag("设计.worker集") ?? "",
|
||||
);
|
||||
expect(worker?.play_slots?.auditor).toBe(true);
|
||||
const strategyRaw = board.getContentByTag(MAINTENANCE_STRATEGY_TAG);
|
||||
expect(strategyRaw).toContain("auditor_bundle");
|
||||
expect(strategyRaw).toContain('"auditor": true');
|
||||
expect(strategyRaw).toContain("inline_tool");
|
||||
});
|
||||
});
|
||||
@@ -1,4 +1,5 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { formatTableValue, parseTableDraft } from "../web/play-aux.js";
|
||||
import { Blackboard } from "../src/blackboard/blackboard.js";
|
||||
import { stringifyTableDoc, createTableFromValues } from "../src/blackboard/table-cells.js";
|
||||
import { buildPlayAuxView } from "../src/skills/play-aux.js";
|
||||
@@ -72,6 +73,8 @@ describe("play-aux view", () => {
|
||||
expect(view.openings).toHaveLength(1);
|
||||
expect(view.openings[0]?.text).toBe("@玩家站在债契前，堂屋很静。");
|
||||
expect(view.openings[0]?.text).not.toContain("present.v1");
|
||||
expect(view.openings[0]?.document).toContain("present.v1");
|
||||
expect(view.openings[0]?.document).toContain("@玩家站在债契前，堂屋很静。");
|
||||
});
|
||||
|
||||
it("lists dictate 设计.开场白#slot openings with same-slot 主角设定", () => {
|
||||
@@ -109,6 +112,45 @@ describe("play-aux view", () => {
|
||||
expect(gate?.persona?.name).toBe("阿泽");
|
||||
});
|
||||
|
||||
it("prefers slotted 设计.开场白#场景短码 over the bare card and the legacy bundle", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({
|
||||
tag: "设计.开场白",
|
||||
content: JSON.stringify({
|
||||
schema: "present.v1",
|
||||
shell: "prose",
|
||||
blocks: { body: "这张没有短码。" },
|
||||
}),
|
||||
source: "agent",
|
||||
});
|
||||
board.write({
|
||||
tag: "设计.开场白#opening-cycle2",
|
||||
content: JSON.stringify({
|
||||
schema: "present.v1",
|
||||
shell: "status_card",
|
||||
blocks: { status_card: "天数|阶段" },
|
||||
}),
|
||||
source: "agent",
|
||||
});
|
||||
board.write({
|
||||
tag: "设计.开场白与开场变量",
|
||||
content: JSON.stringify({
|
||||
schema: "context-fragment.v1",
|
||||
技能: "开场白与开场变量",
|
||||
正文: { 开场白全文: "@玩家站在债契前。" },
|
||||
}),
|
||||
source: "agent",
|
||||
});
|
||||
const view = buildPlayAuxView(board);
|
||||
expect(view.openings).toHaveLength(1);
|
||||
expect(view.openings[0]?.tag).toBe("设计.开场白#opening-cycle2");
|
||||
expect(view.openings[0]?.slot).toBe("opening-cycle2");
|
||||
expect(view.openings[0]?.document).toContain("status_card");
|
||||
expect(view.openings[0]?.document).toContain("天数|阶段");
|
||||
// 无 body 时预览回退整包；选用仍走 document，不会压丢 status_card
|
||||
expect(view.openings[0]?.document).toBe(view.openings[0]?.text);
|
||||
});
|
||||
|
||||
it("lists proliferated zero-layer openings for play swipe", () => {
|
||||
const board = new Blackboard();
|
||||
board.write({
|
||||
@@ -126,3 +168,38 @@ describe("play-aux view", () => {
|
||||
expect(view.selectedOpeningIndex).toBe(1);
|
||||
});
|
||||
});
|
||||
|
||||
describe("play table drafts", () => {
|
||||
it("keeps numbers, booleans, and JSON values in their original types", () => {
|
||||
expect(parseTableDraft(1, "12")).toEqual({ ok: true, value: 12 });
|
||||
expect(parseTableDraft(1, "nope").ok).toBe(false);
|
||||
expect(parseTableDraft(false, "是")).toEqual({ ok: true, value: true });
|
||||
expect(parseTableDraft(["基础侦察术"], '["基础侦察术","追踪"]')).toEqual({
|
||||
ok: true,
|
||||
value: ["基础侦察术", "追踪"],
|
||||
});
|
||||
expect(parseTableDraft(["基础侦察术"], "[")).toEqual({ ok: false, error: "JSON 不合法" });
|
||||
expect(parseTableDraft("空", " 还在 ")).toEqual({ ok: true, value: "还在" });
|
||||
});
|
||||
|
||||
it("pretty-prints long structured values and keeps short ones on one line", () => {
|
||||
expect(formatTableValue(["基础侦察术"])).toBe('["基础侦察术"]');
|
||||
const long = { a: "x".repeat(40), b: [1, 2, 3] };
|
||||
expect(formatTableValue(long)).toContain("\n");
|
||||
expect(formatTableValue(10)).toBe("10");
|
||||
});
|
||||
|
||||
it("parses JSON strings so nested matrices are editable as objects", () => {
|
||||
const matrix =
|
||||
'{"塞琳娜":{"认知度":"有所耳闻","亲密度":-10},"崔岚":{"认知度":"有所耳闻","亲密度":-10}}';
|
||||
expect(formatTableValue(matrix)).toContain("\n");
|
||||
expect(formatTableValue(matrix)).toContain("塞琳娜");
|
||||
expect(parseTableDraft(matrix, formatTableValue(matrix))).toEqual({
|
||||
ok: true,
|
||||
value: {
|
||||
塞琳娜: { 认知度: "有所耳闻", 亲密度: -10 },
|
||||
崔岚: { 认知度: "有所耳闻", 亲密度: -10 },
|
||||
},
|
||||
});
|
||||
});
|
||||
});
|
||||
|
||||
+148
-1
@@ -4,6 +4,7 @@ import {
|
||||
PRESENT_FRAME_SANDBOX,
|
||||
buildPresentSrcdoc,
|
||||
isFrontendBundle,
|
||||
presentFrameSlotHtml,
|
||||
presentSourceBytes,
|
||||
previewPacket,
|
||||
} from "../web/present-frame.js";
|
||||
@@ -24,7 +25,8 @@ describe("present frame srcdoc", () => {
|
||||
});
|
||||
expect(srcdoc).toContain(`<div id="app"></div>`);
|
||||
expect(srcdoc).toContain(`#app { color: #e1e7f0; }`);
|
||||
expect(srcdoc).toContain("present.onData");
|
||||
expect(srcdoc).toContain("present-host");
|
||||
expect(srcdoc).toContain("applyPacket");
|
||||
expect(srcdoc).toContain("示例正文");
|
||||
expect(srcdoc).toContain(`default-src 'none'`);
|
||||
expect(srcdoc.indexOf("window.present")).toBeLessThan(srcdoc.indexOf("data.blocks.body"));
|
||||
@@ -54,6 +56,112 @@ describe("present frame srcdoc", () => {
|
||||
expect(PRESENT_BYTE_LIMIT).toBe(60 * 1024);
|
||||
});
|
||||
|
||||
it("sizes the frame to content and fills contract slots", () => {
|
||||
const srcdoc = buildPresentSrcdoc({
|
||||
html: `<article data-block="body"></article>`,
|
||||
css: `html,body{height:100vh;overflow:auto}`,
|
||||
js: `present.onData(function () {});`,
|
||||
packet: { schema: "present.v1", blocks: { body: "正文", forum_feed: ["一帖"] } },
|
||||
slots: [
|
||||
{ id: "body", type: "markdown" },
|
||||
{ id: "forum_feed", type: "list" },
|
||||
],
|
||||
frameId: "pf3",
|
||||
});
|
||||
expect(srcdoc).not.toContain("100vh");
|
||||
expect(srcdoc).not.toContain("overflow:auto");
|
||||
expect(srcdoc).toContain("overflow:hidden!important");
|
||||
expect(srcdoc).toContain("fillSlots");
|
||||
expect(srcdoc).toContain("isProgramOwned");
|
||||
expect(srcdoc).toContain("forum_feed");
|
||||
expect(srcdoc).toContain('"type":"list"');
|
||||
});
|
||||
|
||||
it("turns markdown newlines into paragraphs even if the shell wrote textContent first", () => {
|
||||
const srcdoc = buildPresentSrcdoc({
|
||||
html: `<article id="body" data-block="body"></article>`,
|
||||
css: "",
|
||||
js: `present.onData(function (data) {
|
||||
document.getElementById("body").textContent = data.blocks.body;
|
||||
});`,
|
||||
packet: {
|
||||
schema: "present.v1",
|
||||
blocks: { body: "第一段\\n\\n第二段 **加粗**" },
|
||||
},
|
||||
slots: [{ id: "body", type: "markdown" }],
|
||||
frameId: "pf4",
|
||||
});
|
||||
// packet JSON keeps real newlines; template sample above used escaped for the test string —
|
||||
// rebuild with actual newlines:
|
||||
const withBreaks = buildPresentSrcdoc({
|
||||
html: `<article id="body" data-block="body"></article>`,
|
||||
css: "",
|
||||
js: `present.onData(function (data) {
|
||||
document.getElementById("body").textContent = data.blocks.body;
|
||||
});`,
|
||||
packet: {
|
||||
schema: "present.v1",
|
||||
blocks: { body: "第一段\n\n第二段 **加粗**" },
|
||||
},
|
||||
slots: [{ id: "body", type: "markdown" }],
|
||||
frameId: "pf4b",
|
||||
});
|
||||
const match = withBreaks.match(/function markdown\(raw\) \{[\s\S]*?\n \}/);
|
||||
expect(match).toBeTruthy();
|
||||
const markdown = new Function(`${match![0]}; return markdown;`)();
|
||||
const html = markdown("第一段\n\n第二段 **加粗**");
|
||||
expect(html).toContain("<p>第一段</p>");
|
||||
expect(html).toContain("<p>第二段 <strong>加粗</strong></p>");
|
||||
expect(withBreaks).toContain("isProgramOwned");
|
||||
expect(srcdoc).toContain("isProgramOwned");
|
||||
});
|
||||
|
||||
it("applies program fill after shell onData so textContent cannot keep a wall of text", () => {
|
||||
const srcdoc = buildPresentSrcdoc({
|
||||
html: `<article id="body" data-block="body"></article>`,
|
||||
css: "",
|
||||
js: `present.onData(function (data) {
|
||||
document.getElementById("body").textContent = data.blocks.body;
|
||||
});`,
|
||||
packet: { schema: "present.v1", blocks: { body: "a\\nb" } },
|
||||
slots: [{ id: "body", type: "markdown" }],
|
||||
frameId: "pf5",
|
||||
});
|
||||
const sealIdx = srcdoc.indexOf("present.__seal");
|
||||
const ownedIdx = srcdoc.indexOf("isProgramOwned");
|
||||
expect(ownedIdx).toBeGreaterThan(-1);
|
||||
expect(sealIdx).toBeGreaterThan(-1);
|
||||
// applyPacket: renderCb then fillSlots
|
||||
const applyIdx = srcdoc.indexOf("function applyPacket");
|
||||
const applyChunk = srcdoc.slice(applyIdx, applyIdx + 450);
|
||||
expect(applyChunk.indexOf("renderCb")).toBeLessThan(applyChunk.lastIndexOf("fillSlots"));
|
||||
expect(srcdoc).toContain('t === "kv" || t === "table" || t === "topology"');
|
||||
});
|
||||
|
||||
it("renders list objects by text/label instead of JSON.stringify", () => {
|
||||
const srcdoc = buildPresentSrcdoc({
|
||||
html: `<ol data-block="options"></ol>`,
|
||||
css: "",
|
||||
js: `present.onData(function () {});`,
|
||||
packet: { schema: "present.v1", blocks: { options: [] } },
|
||||
slots: [{ id: "options", type: "list" }],
|
||||
frameId: "pf6",
|
||||
});
|
||||
const match = srcdoc.match(/function listItemText\(item\) \{[\s\S]*?\n \}/);
|
||||
expect(match).toBeTruthy();
|
||||
const listItemText = new Function(`${match![0]}; return listItemText;`)();
|
||||
expect(listItemText({ type: "继续推进1", text: "顺着旧话试探" })).toBe("顺着旧话试探");
|
||||
expect(listItemText({ label: "换方向" })).toBe("换方向");
|
||||
expect(listItemText("纯字符串")).toBe("纯字符串");
|
||||
expect(listItemText('{"type":"x","text":"从字符串JSON取出"}')).toBe("从字符串JSON取出");
|
||||
});
|
||||
|
||||
it("keeps the previous frame height so a remount does not collapse the page", () => {
|
||||
expect(presentFrameSlotHtml("pf9", 1280)).toContain('style="min-height:1280px"');
|
||||
expect(presentFrameSlotHtml("pf9")).not.toContain("min-height");
|
||||
expect(presentFrameSlotHtml("pf9", 4)).not.toContain("min-height");
|
||||
});
|
||||
|
||||
it("normalizes sample data into present.v1 blocks", () => {
|
||||
expect(previewPacket({ schema: "present.v1", blocks: { body: "x" } })).toEqual({
|
||||
schema: "present.v1",
|
||||
@@ -61,4 +169,43 @@ describe("present frame srcdoc", () => {
|
||||
});
|
||||
expect(previewPacket(null)).toEqual({ schema: "present.v1", blocks: {} });
|
||||
});
|
||||
|
||||
it("accepts Chinese block ids and expands nested matrix JSON into a table", () => {
|
||||
const srcdoc = buildPresentSrcdoc({
|
||||
html: `<section data-block="卷宗表"></section><article data-block="正文"></article>`,
|
||||
css: "",
|
||||
js: "present.onData(function () {});",
|
||||
packet: {
|
||||
schema: "present.v1",
|
||||
blocks: {
|
||||
卷宗表: {
|
||||
source: "heroines_matrix",
|
||||
rows: [
|
||||
{ name: "塞琳娜", 亲密度: -10, 服从度: 10 },
|
||||
{ name: "崔岚", 亲密度: -10, 服从度: 10 },
|
||||
],
|
||||
},
|
||||
正文: "第一段\n\n第二段",
|
||||
},
|
||||
},
|
||||
slots: [
|
||||
{ id: "卷宗表", type: "table" },
|
||||
{ id: "正文", type: "markdown" },
|
||||
],
|
||||
frameId: "pf-zh",
|
||||
});
|
||||
expect(srcdoc).toContain("function isBlockId");
|
||||
expect(srcdoc).toContain("present-data-table");
|
||||
expect(srcdoc).toContain("matrixMapToRows");
|
||||
expect(srcdoc).not.toContain("/^[A-Za-z_]");
|
||||
|
||||
const writeMatch = srcdoc.match(/function writeDataTable\(el, rows, columns\) \{[\s\S]*?\n \}/);
|
||||
expect(writeMatch).toBeTruthy();
|
||||
const idMatch = srcdoc.match(/function isBlockId\(id\) \{[\s\S]*?\n \}/);
|
||||
expect(idMatch).toBeTruthy();
|
||||
const isBlockId = new Function(`${idMatch![0]}; return isBlockId;`)();
|
||||
expect(isBlockId("正文")).toBe(true);
|
||||
expect(isBlockId("body")).toBe(true);
|
||||
expect(isBlockId('bad"id')).toBe(false);
|
||||
});
|
||||
});
|
||||
|
||||
@@ -203,6 +203,39 @@ describe("protagonist overlay", () => {
|
||||
expect(protagonistTagForOpeningSlot("设计.开场白")).toBe(PROTAGONIST_TAG);
|
||||
});
|
||||
|
||||
it("binds present.v1 meta.用户角色 to the slotted card and follows opening rewrites", async () => {
|
||||
const {
|
||||
ensureProtagonistFromPersona,
|
||||
personaBoundToOpening,
|
||||
protagonistTagForOpeningSlot,
|
||||
} = await import("../src/persona/protagonist-bind.js");
|
||||
const { splitOpeningDocument } = await import("../src/skills/opening-seal.js");
|
||||
|
||||
const openingTag = "设计.开场白#debt-hall";
|
||||
const card = (简介: string) =>
|
||||
JSON.stringify({
|
||||
schema: "present.v1",
|
||||
blocks: { body: "清晨，@玩家推开堂屋门。" },
|
||||
meta: { 用户角色: { 名字: "沈聿", 简介 } },
|
||||
});
|
||||
|
||||
const board = new Blackboard();
|
||||
const personaTag = protagonistTagForOpeningSlot(openingTag);
|
||||
for (const 简介 of ["本地放贷人。", "持有这家债契的放贷人。"]) {
|
||||
const content = card(简介);
|
||||
const split = splitOpeningDocument(content);
|
||||
expect(split.persona?.name).toBe("沈聿");
|
||||
expect(split.text).toBe(content);
|
||||
ensureProtagonistFromPersona(board, split.persona, personaTag, {
|
||||
overwrite: true,
|
||||
});
|
||||
board.write({ tag: openingTag, content, source: "agent" });
|
||||
expect(personaBoundToOpening(board, openingTag, content)?.description).toContain(
|
||||
简介,
|
||||
);
|
||||
}
|
||||
});
|
||||
|
||||
it("overlays opening meta 用户角色 when the user switches to that preset", async () => {
|
||||
const { readSessionProtagonist, overlayPlayPersona } = await import(
|
||||
"../src/persona/protagonist-bind.js"
|
||||
|
||||
@@ -0,0 +1,88 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { parsePresentPacket } from "../src/skills/present-packet.js";
|
||||
import {
|
||||
openingContractError,
|
||||
parseReplyDataContract,
|
||||
presentSchemaFromContract,
|
||||
} from "../src/skills/reply-contract.js";
|
||||
|
||||
const replyFormat = JSON.stringify({
|
||||
schema: "context-fragment.v1",
|
||||
正文: {
|
||||
数据契约: {
|
||||
schema: "present.v1",
|
||||
blocks: {
|
||||
status_card: { 类型: "kv", 来源: "program_projection", 职责: "属性" },
|
||||
body: { 类型: "markdown", 来源: "model_generated", 职责: "主读正文" },
|
||||
forum_feed: { 类型: "list", 来源: "model_generated", 职责: "论坛" },
|
||||
danmu_stream: { 类型: "list", 来源: "model_generated", 职责: "弹幕" },
|
||||
},
|
||||
},
|
||||
frontend: { html: `<div id="app"></div>`, css: "body{}", js: "present.onData(function(){})" },
|
||||
},
|
||||
});
|
||||
|
||||
describe("reply data contract", () => {
|
||||
it("keeps model blocks and drops program projections from the fill schema", () => {
|
||||
const contract = parseReplyDataContract(replyFormat);
|
||||
expect(contract?.modelBlocks.map((block) => block.id)).toEqual([
|
||||
"body",
|
||||
"forum_feed",
|
||||
"danmu_stream",
|
||||
]);
|
||||
expect(contract?.programBlocks.map((block) => block.id)).toEqual(["status_card"]);
|
||||
expect(contract?.frontend?.html).toContain("app");
|
||||
const built = presentSchemaFromContract(contract!);
|
||||
const blocks = (
|
||||
built.schema.properties as { blocks: { required: string[] } }
|
||||
).blocks;
|
||||
expect(blocks.required).toEqual(["body", "forum_feed", "danmu_stream"]);
|
||||
expect(built.loose).toBe(false);
|
||||
});
|
||||
|
||||
it("rejects an old shell packet that only fills body", () => {
|
||||
const contract = parseReplyDataContract(replyFormat)!;
|
||||
const error = openingContractError(
|
||||
JSON.stringify({
|
||||
schema: "present.v1",
|
||||
shell: "prose",
|
||||
blocks: { body: "清晨，@玩家推开门。" },
|
||||
}),
|
||||
contract,
|
||||
);
|
||||
expect(error).toContain("forum_feed");
|
||||
expect(error).toContain("danmu_stream");
|
||||
});
|
||||
|
||||
it("accepts present.v1 that fills every model block", () => {
|
||||
const contract = parseReplyDataContract(replyFormat)!;
|
||||
expect(
|
||||
openingContractError(
|
||||
JSON.stringify({
|
||||
schema: "present.v1",
|
||||
blocks: {
|
||||
body: "清晨，@玩家推开门。",
|
||||
forum_feed: ["有人愿意修外墙吗？"],
|
||||
danmu_stream: [],
|
||||
},
|
||||
}),
|
||||
contract,
|
||||
),
|
||||
).toBeUndefined();
|
||||
});
|
||||
|
||||
it("preserves contract blocks when normalizing a present packet", () => {
|
||||
const view = parsePresentPacket(
|
||||
JSON.stringify({
|
||||
schema: "present.v1",
|
||||
blocks: {
|
||||
body: "正文",
|
||||
forum_feed: ["一帖"],
|
||||
danmu_stream: ["一条"],
|
||||
},
|
||||
}),
|
||||
);
|
||||
expect(view.packet.blocks.forum_feed).toEqual(["一帖"]);
|
||||
expect(view.packet.blocks.danmu_stream).toEqual(["一条"]);
|
||||
});
|
||||
});
|
||||
@@ -44,6 +44,16 @@ describe("parseUsage", () => {
|
||||
expect(usage?.cacheMissTokens).toBe(100);
|
||||
});
|
||||
|
||||
it("parses cache write tokens when the provider returns them", () => {
|
||||
const usage = parseUsage({
|
||||
prompt_tokens: 100,
|
||||
completion_tokens: 10,
|
||||
total_tokens: 110,
|
||||
cache_creation_input_tokens: 40,
|
||||
});
|
||||
expect(usage?.cacheWriteTokens).toBe(40);
|
||||
});
|
||||
|
||||
it("returns undefined for empty usage", () => {
|
||||
expect(parseUsage(null)).toBeUndefined();
|
||||
expect(parseUsage({})).toBeUndefined();
|
||||
@@ -52,8 +62,10 @@ describe("parseUsage", () => {
|
||||
|
||||
describe("TokenTrackingProvider", () => {
|
||||
it("stores pending usage with cache fields on context", async () => {
|
||||
let seenSession: string | undefined;
|
||||
const inner: LlmProvider = {
|
||||
async complete(): Promise<CompleteResult> {
|
||||
async complete(_messages, options): Promise<CompleteResult> {
|
||||
seenSession = options?.sessionId;
|
||||
return {
|
||||
content: "ok",
|
||||
model: "test-model",
|
||||
@@ -71,6 +83,7 @@ describe("TokenTrackingProvider", () => {
|
||||
const ctx: LlmTrackingContext = { sessionId: "sess-1" };
|
||||
const provider = new TokenTrackingProvider(inner, () => ctx);
|
||||
await provider.complete([], { caller: "worker:write-rules" });
|
||||
expect(seenSession).toBe("sess-1");
|
||||
|
||||
expect(ctx.pendingUsage).toMatchObject({
|
||||
totalTokens: 15,
|
||||
|
||||
@@ -0,0 +1,37 @@
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { ttsCacheKey } from "../src/tts/cache.js";
|
||||
import { DEFAULT_TTS_PROVIDER, resolveTtsProvider } from "../src/tts/index.js";
|
||||
|
||||
describe("tts cache / provider", () => {
|
||||
it("hashes stable keys", () => {
|
||||
const a = ttsCacheKey({
|
||||
provider: "edge",
|
||||
voice: "zh-CN-XiaoxiaoNeural",
|
||||
text: "你好",
|
||||
extra: "s1",
|
||||
});
|
||||
const b = ttsCacheKey({
|
||||
provider: "edge",
|
||||
voice: "zh-CN-XiaoxiaoNeural",
|
||||
text: "你好",
|
||||
extra: "s1",
|
||||
});
|
||||
const c = ttsCacheKey({
|
||||
provider: "edge",
|
||||
voice: "zh-CN-XiaoxiaoNeural",
|
||||
text: "你好啊",
|
||||
extra: "s1",
|
||||
});
|
||||
expect(a).toBe(b);
|
||||
expect(a).not.toBe(c);
|
||||
expect(a.length).toBe(40);
|
||||
});
|
||||
|
||||
it("defaults provider to edge", () => {
|
||||
expect(DEFAULT_TTS_PROVIDER).toBe("edge");
|
||||
const prev = process.env.WRITING_AGENT_TTS_PROVIDER;
|
||||
delete process.env.WRITING_AGENT_TTS_PROVIDER;
|
||||
expect(resolveTtsProvider()).toBe("edge");
|
||||
process.env.WRITING_AGENT_TTS_PROVIDER = prev;
|
||||
});
|
||||
});
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user