完成世界书、骰子、apiconfig页面处理
This commit is contained in:
231
backend/models/README.md
Normal file
231
backend/models/README.md
Normal file
@@ -0,0 +1,231 @@
|
||||
# Backend Models 数据模型说明
|
||||
|
||||
## 目录结构
|
||||
|
||||
```
|
||||
models/
|
||||
├── __init__.py # 包初始化,导出所有模型
|
||||
├── sillytavern.py # SillyTavern 兼容模型 (仅用于导入/导出)
|
||||
├── internal.py # 内部业务模型 (项目核心使用)
|
||||
└── README.md # 本文件
|
||||
```
|
||||
|
||||
## 模型分类
|
||||
|
||||
### 1. SillyTavern 兼容模型 (`sillytavern.py`)
|
||||
|
||||
**用途**: 仅用于与 SillyTavern 格式的数据进行导入/导出兼容
|
||||
|
||||
**特点**:
|
||||
- 严格遵循 SillyTavern 官方规范
|
||||
- 不参与内部业务逻辑
|
||||
- 所有字段名、结构与 SillyTavern 保持一致
|
||||
- 前缀 `ST` 表示 SillyTavern
|
||||
|
||||
**主要模型**:
|
||||
- `STWorldInfo` - SillyTavern 世界书
|
||||
- `STCharacterCard` - SillyTavern 角色卡
|
||||
- `STChatHeader` / `STChatMessage` - SillyTavern 聊天记录
|
||||
- `STGenerationPreset` - SillyTavern 采样预设
|
||||
- `STPromptPreset` - SillyTavern 提示词预设
|
||||
|
||||
**使用场景**:
|
||||
```python
|
||||
# 从 SillyTavern 导入时
|
||||
st_data = json.load(file)
|
||||
st_character = STCharacterCard(**st_data)
|
||||
|
||||
# 转换为内部模型
|
||||
internal_character = converter.st_to_internal(st_character)
|
||||
|
||||
# 导出到 SillyTavern 时
|
||||
st_data = converter.internal_to_st(internal_character)
|
||||
json.dump(st_data.dict(), file)
|
||||
```
|
||||
|
||||
### 2. 内部业务模型 (`internal.py`)
|
||||
|
||||
**用途**: 项目内部真正使用的数据结构,所有业务逻辑都基于这些模型
|
||||
|
||||
**特点**:
|
||||
- 继承并扩展了 SillyTavern 的功能
|
||||
- 添加了项目特色功能 (如 LOGIC 激活、RAG 配置、outputSchema 等)
|
||||
- 所有 API 响应、数据存储、工作流交换都使用这些模型
|
||||
- 无前缀,直接使用语义化名称
|
||||
|
||||
**主要模型**:
|
||||
|
||||
#### 世界书相关
|
||||
- `ActivationType` - 激活方式枚举 (PERMANENT/KEYWORD/RAG/LOGIC)
|
||||
- `LogicExpression` - 逻辑表达式
|
||||
- `RAGConfig` - RAG 检索配置
|
||||
- `WorldInfoEntry` - 世界书条目
|
||||
- `WorldInfo` - 世界书
|
||||
|
||||
#### 角色卡相关
|
||||
- `OutputSchemaField` - 结构化输出 schema
|
||||
- `CharacterCard` - 角色卡
|
||||
|
||||
#### 聊天记录相关
|
||||
- `ChatHeader` - 聊天头
|
||||
- `ChatMessage` - 聊天消息
|
||||
- `ChatLog` - 完整聊天记录
|
||||
|
||||
#### 预设相关
|
||||
- `GenerationPreset` - 采样参数预设
|
||||
- `PromptRole` - Prompt 角色枚举
|
||||
- `PromptEntry` - Prompt 条目
|
||||
- `PromptPresetView` - Prompt 预设视图
|
||||
|
||||
#### RAG 配置
|
||||
- `RAGSearchConfig` - RAG 搜索配置
|
||||
- `CharacterRAGConfig` - 角色卡 RAG 配置
|
||||
- `ChatRAGConfig` - 聊天 RAG 配置
|
||||
|
||||
**使用场景**:
|
||||
```python
|
||||
# 业务逻辑中直接使用
|
||||
from models import CharacterCard, WorldInfo
|
||||
|
||||
character = CharacterCard(
|
||||
id="uuid-123",
|
||||
name="Alice",
|
||||
description="...",
|
||||
...
|
||||
)
|
||||
|
||||
# API 响应
|
||||
@app.get("/characters/{id}")
|
||||
async def get_character(id: str):
|
||||
character = service.get_character(id)
|
||||
return character # 返回 internal 模型
|
||||
```
|
||||
|
||||
## 数据转换流程
|
||||
|
||||
```
|
||||
SillyTavern 文件
|
||||
↓ (导入)
|
||||
STCharacterCard (sillytavern.py)
|
||||
↓ (转换器)
|
||||
CharacterCard (internal.py)
|
||||
↓ (业务处理)
|
||||
CharacterCard (internal.py)
|
||||
↓ (转换器)
|
||||
STCharacterCard (sillytavern.py)
|
||||
↓ (导出)
|
||||
SillyTavern 文件
|
||||
```
|
||||
|
||||
## 开发规范
|
||||
|
||||
### ✅ 正确做法
|
||||
|
||||
1. **业务逻辑使用 internal 模型**
|
||||
```python
|
||||
from models import CharacterCard
|
||||
|
||||
def create_character(data: dict) -> CharacterCard:
|
||||
return CharacterCard(**data)
|
||||
```
|
||||
|
||||
2. **导入时使用转换器**
|
||||
```python
|
||||
from models import STCharacterCard, CharacterCard
|
||||
from models.converters import CharacterConverter
|
||||
|
||||
def import_character(file_path: str) -> CharacterCard:
|
||||
st_data = load_json(file_path)
|
||||
st_char = STCharacterCard(**st_data)
|
||||
return CharacterConverter.st_to_internal(st_char)
|
||||
```
|
||||
|
||||
3. **API 响应使用 internal 模型**
|
||||
```python
|
||||
@app.get("/characters")
|
||||
async def list_characters() -> List[CharacterCard]:
|
||||
return service.list_characters()
|
||||
```
|
||||
|
||||
### ❌ 错误做法
|
||||
|
||||
1. **不要在业务逻辑中直接使用 ST 模型**
|
||||
```python
|
||||
# 错误!
|
||||
from models import STCharacterCard
|
||||
|
||||
def process_character(char: STCharacterCard):
|
||||
...
|
||||
```
|
||||
|
||||
2. **不要混合使用两种模型**
|
||||
```python
|
||||
# 错误!
|
||||
character = CharacterCard(...)
|
||||
character.name = st_character.data.name # 不要混用
|
||||
```
|
||||
|
||||
3. **不要在 API 中暴露 ST 模型**
|
||||
```python
|
||||
# 错误!
|
||||
@app.get("/characters")
|
||||
async def list_characters() -> List[STCharacterCard]:
|
||||
...
|
||||
```
|
||||
|
||||
## 添加新模型
|
||||
|
||||
当需要添加新的数据类型时:
|
||||
|
||||
1. **判断用途**:
|
||||
- 如果是为了 SillyTavern 兼容 → 添加到 `sillytavern.py`
|
||||
- 如果是项目内部使用 → 添加到 `internal.py`
|
||||
|
||||
2. **遵循命名规范**:
|
||||
- SillyTavern 模型: 前缀 `ST`
|
||||
- 内部模型: 无前缀,使用清晰的语义化名称
|
||||
|
||||
3. **添加详细注释**:
|
||||
```python
|
||||
class MyModel(BaseModel):
|
||||
"""
|
||||
模型用途说明
|
||||
|
||||
详细描述该模型的作用、使用场景等
|
||||
"""
|
||||
field1: str = Field(..., description="字段说明")
|
||||
```
|
||||
|
||||
4. **在 `__init__.py` 中导出**:
|
||||
```python
|
||||
from .internal import MyModel
|
||||
|
||||
__all__ = [
|
||||
...,
|
||||
'MyModel',
|
||||
]
|
||||
```
|
||||
|
||||
## 转换器 (待实现)
|
||||
|
||||
`models/converters.py` 将提供双向转换功能:
|
||||
|
||||
```python
|
||||
class CharacterConverter:
|
||||
@staticmethod
|
||||
def st_to_internal(st_char: STCharacterCard) -> CharacterCard:
|
||||
"""SillyTavern → Internal"""
|
||||
...
|
||||
|
||||
@staticmethod
|
||||
def internal_to_st(int_char: CharacterCard) -> STCharacterCard:
|
||||
"""Internal → SillyTavern"""
|
||||
...
|
||||
```
|
||||
|
||||
## 总结
|
||||
|
||||
- **sillytavern.py** = 外部兼容层 (Import/Export Only)
|
||||
- **internal.py** = 内部业务层 (Core Business Logic)
|
||||
- **永远在业务逻辑中使用 internal 模型**
|
||||
- **通过转换器进行格式转换**
|
||||
61
backend/models/__init__.py
Normal file
61
backend/models/__init__.py
Normal file
@@ -0,0 +1,61 @@
|
||||
"""
|
||||
数据模型包
|
||||
|
||||
导出项目内部真正使用的数据结构 (Internal Models)。
|
||||
SillyTavern 兼容模型将在需要导入/导出时单独引用。
|
||||
"""
|
||||
|
||||
# 内部业务模型 (项目核心使用)
|
||||
from .internal import (
|
||||
# 世界书
|
||||
ActivationType,
|
||||
LogicOperator,
|
||||
LogicExpression,
|
||||
RAGConfig,
|
||||
WorldInfoEntry,
|
||||
WorldInfo,
|
||||
|
||||
# 角色卡
|
||||
OutputSchemaField,
|
||||
CharacterCard,
|
||||
|
||||
# 聊天记录
|
||||
ChatHeader,
|
||||
ChatMessage,
|
||||
ChatLog,
|
||||
|
||||
# 预设
|
||||
GenerationPreset,
|
||||
|
||||
# 提示词预设
|
||||
PromptRole,
|
||||
PromptEntry,
|
||||
PromptPresetView,
|
||||
|
||||
# RAG 配置
|
||||
RAGSearchConfig,
|
||||
CharacterRAGConfig,
|
||||
ChatRAGConfig,
|
||||
)
|
||||
|
||||
__all__ = [
|
||||
# 内部模型
|
||||
'ActivationType',
|
||||
'LogicOperator',
|
||||
'LogicExpression',
|
||||
'RAGConfig',
|
||||
'WorldInfoEntry',
|
||||
'WorldInfo',
|
||||
'OutputSchemaField',
|
||||
'CharacterCard',
|
||||
'ChatHeader',
|
||||
'ChatMessage',
|
||||
'ChatLog',
|
||||
'GenerationPreset',
|
||||
'PromptRole',
|
||||
'PromptEntry',
|
||||
'PromptPresetView',
|
||||
'RAGSearchConfig',
|
||||
'CharacterRAGConfig',
|
||||
'ChatRAGConfig',
|
||||
]
|
||||
379
backend/models/converters.py
Normal file
379
backend/models/converters.py
Normal file
@@ -0,0 +1,379 @@
|
||||
"""
|
||||
数据模型转换器
|
||||
|
||||
提供 SillyTavern 格式与内部格式之间的双向转换功能。
|
||||
所有导入/导出操作都应该通过转换器进行,确保数据格式的一致性。
|
||||
"""
|
||||
import uuid
|
||||
from typing import Dict, Any, List, Optional
|
||||
from datetime import datetime
|
||||
|
||||
from models.internal import (
|
||||
WorldInfo,
|
||||
WorldInfoEntry,
|
||||
ActivationType,
|
||||
)
|
||||
|
||||
|
||||
class WorldBookConverter:
|
||||
"""世界书数据转换器
|
||||
|
||||
负责 SillyTavern 格式和项目内部格式之间的转换。
|
||||
|
||||
SillyTavern 格式特点:
|
||||
- entries 是 dict (key 为 uid)
|
||||
- 使用 constant 字段表示常驻激活
|
||||
- position 是字符串 (如 "after_char")
|
||||
|
||||
项目内部格式特点:
|
||||
- entries 是 list
|
||||
- 使用 activationType 枚举
|
||||
- position 是数字 (0-5)
|
||||
- 包含 trigger_config 结构(前端需要)
|
||||
"""
|
||||
|
||||
@staticmethod
|
||||
def detect_format(data: Dict[str, Any]) -> str:
|
||||
"""
|
||||
智能检测世界书数据格式
|
||||
|
||||
Args:
|
||||
data: 世界书数据
|
||||
|
||||
Returns:
|
||||
'sillytavern' | 'internal' | 'unknown'
|
||||
"""
|
||||
# 检查 entries 类型
|
||||
entries = data.get("entries")
|
||||
if not entries:
|
||||
return "unknown"
|
||||
|
||||
# SillyTavern 特征: entries 是 dict
|
||||
if isinstance(entries, dict):
|
||||
return "sillytavern"
|
||||
|
||||
# 内部格式特征: entries 是 list
|
||||
if isinstance(entries, list):
|
||||
# 进一步检查是否有 trigger_config
|
||||
if len(entries) > 0 and isinstance(entries[0], dict):
|
||||
first_entry = entries[0]
|
||||
if "trigger_config" in first_entry:
|
||||
return "internal"
|
||||
# 也可能是简化的内部格式
|
||||
if "activationType" in first_entry or "position" in first_entry:
|
||||
return "internal"
|
||||
|
||||
return "unknown"
|
||||
|
||||
# 位置映射: SillyTavern 字符串 -> 内部数字
|
||||
POSITION_MAP_ST_TO_INTERNAL = {
|
||||
"after_char": 0,
|
||||
"before_char": 1,
|
||||
"before_example": 2,
|
||||
"after_example": 3,
|
||||
"author_note": 4,
|
||||
"system_prompt": 5,
|
||||
}
|
||||
|
||||
# 位置映射: 内部数字 -> SillyTavern 字符串
|
||||
POSITION_MAP_INTERNAL_TO_ST = {
|
||||
0: "after_char",
|
||||
1: "before_char",
|
||||
2: "before_example",
|
||||
3: "after_example",
|
||||
4: "author_note",
|
||||
5: "system_prompt",
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def st_to_internal(st_data: Dict[str, Any], name: str = None) -> Dict[str, Any]:
|
||||
"""
|
||||
将 SillyTavern 格式的世界书转换为内部格式
|
||||
|
||||
Args:
|
||||
st_data: SillyTavern 格式的世界书数据
|
||||
name: 世界书名称(可选,优先使用 st_data 中的 name)
|
||||
|
||||
Returns:
|
||||
内部格式的世界书字典(包含 trigger_config)
|
||||
"""
|
||||
now = int(datetime.now().timestamp())
|
||||
|
||||
# 转换条目
|
||||
entries = []
|
||||
st_entries = st_data.get("entries", {})
|
||||
|
||||
# SillyTavern 的 entries 可能是 dict 或 list
|
||||
if isinstance(st_entries, dict):
|
||||
entries_list = list(st_entries.values())
|
||||
elif isinstance(st_entries, list):
|
||||
entries_list = st_entries
|
||||
else:
|
||||
entries_list = []
|
||||
|
||||
for st_entry in entries_list:
|
||||
if not isinstance(st_entry, dict):
|
||||
continue
|
||||
|
||||
# 判断激活类型
|
||||
is_constant = st_entry.get("constant", False)
|
||||
activation_type = ActivationType.PERMANENT if is_constant else ActivationType.KEYWORD
|
||||
|
||||
# 转换位置
|
||||
st_position = st_entry.get("position", "after_char")
|
||||
internal_position = WorldBookConverter.POSITION_MAP_ST_TO_INTERNAL.get(st_position, 0)
|
||||
|
||||
# 构建 trigger_config (前端期望的格式)
|
||||
trigger_config = WorldBookConverter._build_trigger_config(
|
||||
is_constant=is_constant,
|
||||
key=st_entry.get("key", []),
|
||||
keysecondary=st_entry.get("keysecondary", []),
|
||||
selective=st_entry.get("selective", True)
|
||||
)
|
||||
|
||||
# 创建内部格式的条目
|
||||
entry_dict = {
|
||||
"uid": st_entry.get("uid", str(uuid.uuid4())),
|
||||
"key": st_entry.get("key", []),
|
||||
"keysecondary": st_entry.get("keysecondary", []),
|
||||
"content": st_entry.get("content", ""),
|
||||
"comment": st_entry.get("comment", ""),
|
||||
"activationType": activation_type.value,
|
||||
"trigger_config": trigger_config,
|
||||
"order": st_entry.get("order", 100),
|
||||
"position": internal_position,
|
||||
"depth": st_entry.get("depth", 4),
|
||||
"role": st_entry.get("role", 0),
|
||||
"probability": st_entry.get("probability", 100),
|
||||
"group": st_entry.get("group", []),
|
||||
"disable": st_entry.get("disable", False),
|
||||
"createdAt": now,
|
||||
"updatedAt": now
|
||||
}
|
||||
|
||||
entries.append(entry_dict)
|
||||
|
||||
# 创建内部格式的世界书
|
||||
worldbook_data = {
|
||||
"id": str(uuid.uuid4()),
|
||||
"name": name or st_data.get("name", "Unnamed"),
|
||||
"description": st_data.get("description", ""),
|
||||
"entries": entries,
|
||||
"createdAt": now,
|
||||
"updatedAt": now,
|
||||
"version": 1
|
||||
}
|
||||
|
||||
return worldbook_data
|
||||
|
||||
@staticmethod
|
||||
def internal_to_st(worldbook_data: Dict[str, Any]) -> Dict[str, Any]:
|
||||
"""
|
||||
将内部格式的世界书转换为 SillyTavern 格式
|
||||
|
||||
Args:
|
||||
worldbook_data: 内部格式的世界书字典
|
||||
|
||||
Returns:
|
||||
SillyTavern 格式的世界书数据
|
||||
"""
|
||||
# 转换条目
|
||||
st_entries = {}
|
||||
|
||||
for entry_data in worldbook_data.get("entries", []):
|
||||
if not isinstance(entry_data, dict):
|
||||
continue
|
||||
|
||||
uid = entry_data.get("uid", str(uuid.uuid4()))
|
||||
|
||||
# 从 trigger_config 或 activationType 判断是否常驻
|
||||
is_constant = WorldBookConverter._is_constant_entry(entry_data)
|
||||
|
||||
# 提取关键词
|
||||
key, keysecondary = WorldBookConverter._extract_keywords(entry_data)
|
||||
|
||||
# 转换位置
|
||||
internal_position = entry_data.get("position", 0)
|
||||
st_position = WorldBookConverter.POSITION_MAP_INTERNAL_TO_ST.get(internal_position, "after_char")
|
||||
|
||||
# 创建 SillyTavern 格式的条目
|
||||
st_entry = {
|
||||
"uid": uid,
|
||||
"key": key,
|
||||
"keysecondary": keysecondary,
|
||||
"content": entry_data.get("content", ""),
|
||||
"comment": entry_data.get("comment", ""),
|
||||
"constant": is_constant,
|
||||
"selective": not is_constant,
|
||||
"order": entry_data.get("order", 100),
|
||||
"position": st_position,
|
||||
"depth": entry_data.get("depth", 4),
|
||||
"probability": entry_data.get("probability", 100),
|
||||
"group": entry_data.get("group", []),
|
||||
"disable": entry_data.get("disable", False)
|
||||
}
|
||||
|
||||
st_entries[uid] = st_entry
|
||||
|
||||
# 创建 SillyTavern 格式的世界书
|
||||
st_data = {
|
||||
"name": worldbook_data.get("name", ""),
|
||||
"description": worldbook_data.get("description", ""),
|
||||
"entries": st_entries
|
||||
}
|
||||
|
||||
return st_data
|
||||
|
||||
@staticmethod
|
||||
def normalize_entry(entry_data: Dict[str, Any]) -> Dict[str, Any]:
|
||||
"""
|
||||
规范化条目数据,确保包含所有必需字段和 trigger_config
|
||||
|
||||
Args:
|
||||
entry_data: 条目数据(可能来自不同来源)
|
||||
|
||||
Returns:
|
||||
规范化后的条目数据
|
||||
"""
|
||||
now = int(datetime.now().timestamp())
|
||||
|
||||
# 如果已经有 trigger_config,直接返回
|
||||
if "trigger_config" in entry_data and entry_data["trigger_config"]:
|
||||
return entry_data
|
||||
|
||||
# 否则从其他字段构建 trigger_config
|
||||
is_constant = WorldBookConverter._is_constant_entry(entry_data)
|
||||
key, keysecondary = WorldBookConverter._extract_keywords(entry_data)
|
||||
|
||||
trigger_config = WorldBookConverter._build_trigger_config(
|
||||
is_constant=is_constant,
|
||||
key=key,
|
||||
keysecondary=keysecondary,
|
||||
selective=entry_data.get("selective", True)
|
||||
)
|
||||
|
||||
# 添加缺失的字段
|
||||
normalized = {
|
||||
"uid": entry_data.get("uid", str(uuid.uuid4())),
|
||||
"key": key,
|
||||
"keysecondary": keysecondary,
|
||||
"content": entry_data.get("content", ""),
|
||||
"comment": entry_data.get("comment", ""),
|
||||
"activationType": entry_data.get("activationType",
|
||||
ActivationType.PERMANENT.value if is_constant
|
||||
else ActivationType.KEYWORD.value),
|
||||
"trigger_config": trigger_config,
|
||||
"order": entry_data.get("order", 100),
|
||||
"position": entry_data.get("position", 0),
|
||||
"depth": entry_data.get("depth", 4),
|
||||
"role": entry_data.get("role", 0),
|
||||
"probability": entry_data.get("probability", 100),
|
||||
"group": entry_data.get("group", []),
|
||||
"disable": entry_data.get("disable", False),
|
||||
"createdAt": entry_data.get("createdAt", now),
|
||||
"updatedAt": entry_data.get("updatedAt", now)
|
||||
}
|
||||
|
||||
return normalized
|
||||
|
||||
@staticmethod
|
||||
def _build_trigger_config(
|
||||
is_constant: bool,
|
||||
key: List[str],
|
||||
keysecondary: List[str],
|
||||
selective: bool = True
|
||||
) -> Dict[str, Any]:
|
||||
"""
|
||||
构建 trigger_config 结构
|
||||
|
||||
Args:
|
||||
is_constant: 是否常驻激活
|
||||
key: 主关键词列表
|
||||
keysecondary: 次要关键词列表
|
||||
selective: 是否选择性匹配
|
||||
|
||||
Returns:
|
||||
trigger_config 字典
|
||||
"""
|
||||
return {
|
||||
"triggers": {
|
||||
"constant": [is_constant, None],
|
||||
"keyword": [
|
||||
not is_constant,
|
||||
{
|
||||
"key": key,
|
||||
"keysecondary": keysecondary,
|
||||
"selective": selective,
|
||||
"selectiveLogic": 0,
|
||||
"matchWholeWords": False,
|
||||
"caseSensitive": False
|
||||
}
|
||||
],
|
||||
"rag": [False, {
|
||||
"threshold": 0.75,
|
||||
"top_k": 5,
|
||||
"query_template": None
|
||||
}],
|
||||
"condition": [False, {
|
||||
"variable_a": "",
|
||||
"operator": "=",
|
||||
"variable_b": ""
|
||||
}]
|
||||
}
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def _is_constant_entry(entry_data: Dict[str, Any]) -> bool:
|
||||
"""
|
||||
判断条目是否为常驻激活
|
||||
|
||||
Args:
|
||||
entry_data: 条目数据
|
||||
|
||||
Returns:
|
||||
是否常驻激活
|
||||
"""
|
||||
# 优先从 trigger_config 判断
|
||||
if "trigger_config" in entry_data and entry_data["trigger_config"]:
|
||||
try:
|
||||
return entry_data["trigger_config"]["triggers"]["constant"][0]
|
||||
except (KeyError, IndexError, TypeError):
|
||||
pass
|
||||
|
||||
# 其次从 activationType 判断
|
||||
if "activationType" in entry_data:
|
||||
return entry_data["activationType"] == ActivationType.PERMANENT.value
|
||||
|
||||
# 最后从 constant 字段判断
|
||||
if "constant" in entry_data:
|
||||
return entry_data["constant"]
|
||||
|
||||
return False
|
||||
|
||||
@staticmethod
|
||||
def _extract_keywords(entry_data: Dict[str, Any]) -> tuple:
|
||||
"""
|
||||
从条目数据中提取关键词
|
||||
|
||||
Args:
|
||||
entry_data: 条目数据
|
||||
|
||||
Returns:
|
||||
(key, keysecondary) 元组
|
||||
"""
|
||||
# 优先从 trigger_config 提取
|
||||
if "trigger_config" in entry_data and entry_data["trigger_config"]:
|
||||
try:
|
||||
keyword_config = entry_data["trigger_config"]["triggers"]["keyword"][1]
|
||||
if keyword_config:
|
||||
key = keyword_config.get("key", [])
|
||||
keysecondary = keyword_config.get("keysecondary", [])
|
||||
return key, keysecondary
|
||||
except (KeyError, IndexError, TypeError):
|
||||
pass
|
||||
|
||||
# 否则从顶层字段提取
|
||||
key = entry_data.get("key", [])
|
||||
keysecondary = entry_data.get("keysecondary", [])
|
||||
|
||||
return key, keysecondary
|
||||
301
backend/models/internal.py
Normal file
301
backend/models/internal.py
Normal file
@@ -0,0 +1,301 @@
|
||||
"""
|
||||
项目内部数据结构定义
|
||||
|
||||
这是本项目真正使用的核心数据模型,所有业务逻辑都基于这些类型。
|
||||
与 sillytavern.py 不同,这里的模型不参与导入导出兼容,而是专注于:
|
||||
- 内部业务逻辑处理
|
||||
- API 响应数据结构
|
||||
- 数据存储格式
|
||||
- 工作流引擎数据交换
|
||||
|
||||
所有从 SillyTavern 导入的数据都会转换为这些内部模型进行处理,
|
||||
导出时再从内部模型转换回 SillyTavern 格式。
|
||||
"""
|
||||
from enum import Enum
|
||||
from typing import List, Optional, Dict, Any
|
||||
from pydantic import BaseModel, Field
|
||||
from datetime import datetime
|
||||
|
||||
|
||||
# ==================== 世界书 (World Info) ====================
|
||||
|
||||
class ActivationType(str, Enum):
|
||||
"""
|
||||
自定义激活方式类型（4种枚举）
|
||||
|
||||
这是项目的核心创新点之一,相比 SillyTavern 的简单 constant/selective 标志,
|
||||
我们提供了更灵活的激活机制。
|
||||
"""
|
||||
PERMANENT = 'permanent' # 永久激活 - 始终包含在上下文中
|
||||
KEYWORD = 'keyword' # 关键词触发 - 匹配关键词时激活
|
||||
RAG = 'rag' # RAG 检索激活 - 基于向量相似度检索
|
||||
LOGIC = 'logic' # 逻辑表达式激活 - 基于变量条件判断
|
||||
|
||||
|
||||
class LogicOperator(str, Enum):
|
||||
"""逻辑运算符（用于 LOGIC 激活类型）"""
|
||||
EQUALS = 'equals' # 等于
|
||||
NOT_EQUALS = 'not_equals' # 不等于
|
||||
CONTAINS = 'contains' # 包含
|
||||
NOT_CONTAINS = 'not_contains' # 不包含
|
||||
GREATER = 'greater' # 大于
|
||||
LESS = 'less' # 小于
|
||||
|
||||
|
||||
class LogicExpression(BaseModel):
|
||||
"""
|
||||
逻辑表达式结构（用于 LOGIC 激活类型）
|
||||
|
||||
示例: variable1="mood", operator="equals", variable2="happy"
|
||||
表示当 mood 变量等于 happy 时激活该条目
|
||||
"""
|
||||
variable1: str = Field(..., description="第一个变量名")
|
||||
operator: LogicOperator = Field(..., description="比较运算符")
|
||||
variable2: str = Field(..., description="第二个变量名或值")
|
||||
|
||||
|
||||
class RAGConfig(BaseModel):
|
||||
"""
|
||||
RAG 配置（用于 RAG 激活类型）
|
||||
|
||||
控制如何从向量数据库中检索相关内容
|
||||
"""
|
||||
libraryId: str = Field(..., description="绑定的 RAG 库 ID")
|
||||
threshold: Optional[float] = Field(0.7, ge=0, le=1, description="相似度阈值 (0-1)")
|
||||
maxEntries: Optional[int] = Field(5, gt=0, description="最大返回条目数")
|
||||
|
||||
|
||||
class WorldInfoEntry(BaseModel):
|
||||
"""
|
||||
项目内部世界书条目结构
|
||||
|
||||
这是世界书的核心单元,每个条目代表一段可以被动态注入到对话上下文中的知识。
|
||||
相比 SillyTavern,我们添加了 activationType、logicExpression、ragConfig 等高级功能。
|
||||
"""
|
||||
uid: str = Field(..., description="条目唯一标识符 (UUID)")
|
||||
key: Optional[List[str]] = Field(None, description="主关键词列表 (用于 KEYWORD 激活)")
|
||||
keysecondary: Optional[List[str]] = Field(None, description="次要关键词列表 (可选过滤)")
|
||||
content: str = Field(..., description="条目内容 - 激活时注入的文本")
|
||||
activationType: ActivationType = Field(..., description="激活方式")
|
||||
logicExpression: Optional[LogicExpression] = Field(None, description="逻辑表达式 (LOGIC 类型使用)")
|
||||
ragConfig: Optional[RAGConfig] = Field(None, description="RAG 配置 (RAG 类型使用)")
|
||||
order: int = Field(0, description="插入顺序 - 数值越大越靠近末尾")
|
||||
position: Optional[str] = Field('after_char', description="插入位置")
|
||||
depth: Optional[int] = Field(None, description="插入深度 (当 position='at_depth' 时使用)")
|
||||
probability: Optional[float] = Field(100, ge=0, le=100, description="激活概率 (0-100)")
|
||||
group: Optional[List[str]] = Field(None, description="所属组标签")
|
||||
disable: bool = Field(False, description="是否禁用")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
|
||||
|
||||
class WorldInfo(BaseModel):
|
||||
"""
|
||||
项目内部世界书结构
|
||||
|
||||
世界书是角色知识的集合,可以绑定到角色卡上,在对话中动态提供背景信息。
|
||||
"""
|
||||
id: str = Field(..., description="世界书唯一标识符 (UUID)")
|
||||
name: str = Field(..., description="世界书名称")
|
||||
description: Optional[str] = Field(None, description="世界书描述")
|
||||
entries: List[WorldInfoEntry] = Field(default_factory=list, description="条目数组")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
version: int = Field(1, description="版本号 (用于数据迁移)")
|
||||
|
||||
|
||||
# ==================== 角色卡 (Character Card) ====================
|
||||
|
||||
class OutputSchemaField(BaseModel):
|
||||
"""
|
||||
Vercel AI SDK Output.object() 的表头定义
|
||||
|
||||
用于结构化输出,让 LLM 按照指定格式返回数据。
|
||||
这是项目的特色功能,支持动态表格生成。
|
||||
"""
|
||||
name: str = Field(..., description="字段名称")
|
||||
type: str = Field(..., description="字段类型 (string/number/boolean/array/object)")
|
||||
description: str = Field(..., description="字段描述")
|
||||
required: Optional[bool] = Field(None, description="是否必需")
|
||||
enum: Optional[List[str]] = Field(None, description="枚举值 (字符串固定选项)")
|
||||
fields: Optional[List['OutputSchemaField']] = Field(None, description="嵌套字段 (object 类型)")
|
||||
|
||||
|
||||
class CharacterCard(BaseModel):
|
||||
"""
|
||||
项目内部角色卡结构
|
||||
|
||||
角色卡是对话 AI 的核心定义,包含人设、场景、开场白等。
|
||||
相比 SillyTavern,我们添加了 categories、outputSchema、worldInfoId 等功能。
|
||||
"""
|
||||
id: str = Field(..., description="角色唯一标识符 (UUID)")
|
||||
name: str = Field(..., description="角色名称")
|
||||
description: str = Field(..., description="角色详细描述")
|
||||
personality: str = Field(..., description="角色性格特征")
|
||||
scenario: str = Field(..., description="场景设定")
|
||||
first_mes: str = Field(..., description="首条开场消息")
|
||||
mes_example: str = Field(..., description="对话示例")
|
||||
categories: List[str] = Field(default_factory=list, description="分类标签 (用于前端筛选)")
|
||||
worldInfoId: Optional[str] = Field(None, description="绑定的世界书 ID")
|
||||
outputSchema: Optional[List[OutputSchemaField]] = Field(None, description="输出 schema 定义 (结构化输出)")
|
||||
avatarPath: Optional[str] = Field(None, description="角色头像路径")
|
||||
alternate_greetings: Optional[List[str]] = Field(None, description="替代问候语数组")
|
||||
tags: Optional[List[str]] = Field(None, description="标签数组")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
lastChatAt: Optional[int] = Field(None, description="最后聊天时间戳")
|
||||
isFavorite: bool = Field(False, description="收藏状态")
|
||||
version: int = Field(1, description="版本号")
|
||||
|
||||
|
||||
# ==================== 聊天记录 (Chat Log) ====================
|
||||
|
||||
class ChatHeader(BaseModel):
|
||||
"""
|
||||
项目内部聊天记录头
|
||||
|
||||
包含聊天的元数据,如参与角色、创建时间等。
|
||||
"""
|
||||
id: str = Field(..., description="聊天唯一标识符 (UUID)")
|
||||
displayName: str = Field(..., description="显示名称 (聊天标题)")
|
||||
characterId: str = Field(..., description="关联的角色卡 ID")
|
||||
userName: str = Field("User", description="用户角色名")
|
||||
characterName: str = Field(..., description="AI 角色名称")
|
||||
tableData: Optional[Dict[str, Any]] = Field(None, description="表格数据 (对应 outputSchema)")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
messageCount: int = Field(0, description="消息数量")
|
||||
ragLibraryId: Optional[str] = Field(None, description="关联的 RAG 历史消息库 ID")
|
||||
|
||||
|
||||
class ChatMessage(BaseModel):
|
||||
"""
|
||||
项目内部聊天消息
|
||||
|
||||
单条对话消息,支持多版本 (swipes)、token 统计等功能。
|
||||
"""
|
||||
id: str = Field(..., description="消息唯一标识符 (UUID)")
|
||||
name: str = Field(..., description="发送者名称")
|
||||
is_user: bool = Field(..., description="是否为用户消息")
|
||||
is_system: Optional[bool] = Field(None, description="是否为系统消息")
|
||||
sendDate: str = Field(..., description="发送日期 ISO 字符串")
|
||||
mes: str = Field(..., description="消息内容文本")
|
||||
chatId: str = Field(..., description="关联的聊天 ID")
|
||||
swipes: Optional[List[str]] = Field(None, description="替换回答数组 (多版本)")
|
||||
swipe_id: Optional[int] = Field(0, description="当前选择的版本索引")
|
||||
tokenCount: Optional[int] = Field(None, description="Token 数量 (用于统计)")
|
||||
isTemporary: Optional[bool] = Field(None, description="是否为临时消息 (未保存)")
|
||||
|
||||
|
||||
class ChatLog(BaseModel):
|
||||
"""
|
||||
项目内部完整聊天记录
|
||||
|
||||
包含聊天头和所有消息,是完整的对话历史。
|
||||
"""
|
||||
header: ChatHeader = Field(..., description="聊天头")
|
||||
messages: List[ChatMessage] = Field(default_factory=list, description="消息列表")
|
||||
|
||||
|
||||
# ==================== 预设 (Preset) ====================
|
||||
|
||||
class GenerationPreset(BaseModel):
|
||||
"""
|
||||
项目内部采样参数预设
|
||||
|
||||
控制 LLM 生成的参数配置,如温度、top_p 等。
|
||||
"""
|
||||
id: str = Field(..., description="预设唯一标识符 (UUID)")
|
||||
name: str = Field(..., description="预设名称")
|
||||
temperature: float = Field(1.0, ge=0, le=2, description="温度 (控制随机性)")
|
||||
topP: float = Field(1.0, ge=0, le=1, description="Top P (核采样)")
|
||||
topK: int = Field(0, ge=0, description="Top K")
|
||||
repetitionPenalty: float = Field(1.0, ge=0, description="重复惩罚")
|
||||
frequencyPenalty: Optional[float] = Field(None, description="频率惩罚")
|
||||
presencePenalty: Optional[float] = Field(None, description="存在惩罚")
|
||||
maxLength: Optional[int] = Field(None, gt=0, description="最大生成长度")
|
||||
isDefault: bool = Field(False, description="是否为默认预设")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
|
||||
|
||||
# ==================== 提示词预设 (Prompt Preset) ====================
|
||||
|
||||
class PromptRole(str, Enum):
|
||||
"""
|
||||
Prompt 角色类型
|
||||
|
||||
内部业务层只保留三种角色,简化了 SillyTavern 的复杂角色系统。
|
||||
"""
|
||||
SYSTEM = 'system' # 系统指令
|
||||
AI = 'ai' # AI 助手
|
||||
USER = 'user' # 用户
|
||||
|
||||
|
||||
class PromptEntry(BaseModel):
|
||||
"""
|
||||
内部业务层 - Prompt 条目
|
||||
|
||||
提示词模板的基本单元,可以组合成完整的提示词预设。
|
||||
这是基于某个 character_id 生成的"当前视图"。
|
||||
"""
|
||||
identifier: str = Field(..., description="稳定关联键 (用于回写)")
|
||||
name: str = Field(..., description="条目名 (前端显示)")
|
||||
enabled: bool = Field(True, description="是否启用 (当前作用域下的业务状态)")
|
||||
content: str = Field(..., description="条目内容 (静态内容视图)")
|
||||
order: int = Field(..., description="条目顺序 (前端展示和拖拽排序)")
|
||||
role: PromptRole = Field(..., description="角色类型")
|
||||
tokenCount: int = Field(0, description="总 token 数 (派生显示字段)")
|
||||
isSystemNode: bool = Field(False, description="是否固有节点 (不可删除)")
|
||||
|
||||
|
||||
class PromptPresetView(BaseModel):
|
||||
"""
|
||||
内部业务层 - Prompt 预设视图
|
||||
|
||||
基于某个 character_id 的"当前视图",包含已排序、已过滤的条目列表。
|
||||
"""
|
||||
characterId: str = Field(..., description="关联的角色 ID")
|
||||
entries: List[PromptEntry] = Field(default_factory=list, description="当前视图的条目列表")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
version: int = Field(1, description="版本号")
|
||||
|
||||
|
||||
# ==================== RAG 配置 ====================
|
||||
|
||||
class RAGSearchConfig(BaseModel):
|
||||
"""RAG 搜索配置"""
|
||||
topK: int = Field(5, gt=0, description="每次检索返回的结果数")
|
||||
threshold: float = Field(0.7, ge=0, le=1, description="相似度阈值 (0-1)")
|
||||
maxContextLength: int = Field(2000, gt=0, description="最大上下文长度 (字符数)")
|
||||
|
||||
|
||||
class CharacterRAGConfig(BaseModel):
|
||||
"""
|
||||
角色卡 RAG 世界书库配置
|
||||
|
||||
记录角色卡关联的 RAG 知识库,用于动态检索相关知识。
|
||||
"""
|
||||
characterId: str = Field(..., description="角色卡ID")
|
||||
ragLibraryIds: List[str] = Field(default_factory=list, description="关联的RAG库ID列表")
|
||||
enabled: bool = Field(True, description="是否启用")
|
||||
searchConfig: Optional[RAGSearchConfig] = Field(None, description="搜索配置")
|
||||
position: str = Field('after_char', description="RAG内容插入位置")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
|
||||
|
||||
class ChatRAGConfig(BaseModel):
|
||||
"""
|
||||
聊天会话 RAG 历史消息配置
|
||||
|
||||
记录聊天会话关联的 RAG 历史消息库,用于智能检索历史对话。
|
||||
"""
|
||||
chatId: str = Field(..., description="聊天会话ID")
|
||||
ragLibraryId: Optional[str] = Field(None, description="关联的RAG历史消息库ID")
|
||||
enabled: bool = Field(True, description="是否启用")
|
||||
searchConfig: Optional[Dict[str, Any]] = Field(None, description="搜索配置")
|
||||
autoIndex: bool = Field(True, description="是否自动索引新消息")
|
||||
indexConfig: Optional[Dict[str, Any]] = Field(None, description="索引配置")
|
||||
createdAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="创建时间戳")
|
||||
updatedAt: int = Field(default_factory=lambda: int(datetime.now().timestamp()), description="最后更新时间戳")
|
||||
Reference in New Issue
Block a user