初始化仅前端版本

backend/app/core/Start.py

@@ -0,0 +1,46 @@
+import base64
+from typing import Any, Dict
+from IPython.core.magic_arguments import defaults
+from  .. import nodes
+
+class StartNode():
+    name = "开始节点"
+    inputs = {
+        "user_input": "string",  # 用户输入文本
+        "stream": "boolean",  # 是否流式输出
+        "img_switch": "boolean",  # 是否处理图片
+        "table_switch": "boolean",  # 是否处理表格
+        "role_name": "string",  # 角色名称
+        "chat_name": "string"  # 会话名称
+    }
+
+    async def run(self, text: str = None, image: bytes = None, **kwargs) -> Dict[str, Any]:
+        # 查空：文本不能为空字符串
+        if not text or text.strip() == "":
+            raise ValueError("文本输入不能为空")
+
+        # 查空：图片数据不能为空
+        if image is None or len(image) == 0:
+            raise ValueError("图片输入不能为空")
+
+        # 将图片字节转换为 Base64 字符串，便于在节点间传递
+        image_base64 = base64.b64encode(image).decode('utf-8')
+
+        return {
+            "text": text,
+            "image": image_base64
+        }
+
+async def run(is_user,floor_number,mes: str = None, stream: bool = False, img_switch: bool = False,name = "default",
+              table_switch: bool = False, role_name: str = None, chat_name: str = None,preset: str = None):
+    # 将输入内容持久化存储到本地json方便前端读
+    nodes.save_input_to_json(mes=mes, role_name=role_name, chat_name=chat_name, name=name, is_user=is_user, floor_number=floor_number)
+    # 对上一条输入内容（已确定不变的内容）调用向量化，根据role和chat嵌入到对应本地数据库
+    embed_input(user_input, role_name, chat_name)
+    # 根据role和chat去读取绑定的世界书
+    # 读取预设，进行拼接
+    # 调用模型，返回结果
+    # 将结果持久化存储到本地json方便前端读（用JSONL）
+    # 如果img_switch是开的，那么异步调用生图，并存储到目标文件夹里
+    # 如果table_switch是开的，那么异步调用表格生成，并存储到目标文件夹里
+    # 将结果返回给前端

backend/app/nodes/Start.py

@@ -1,29 +0,0 @@
-import base64
-from core.node_base import BaseNode
-from typing import Any, Dict
-
-
-class StartNode(BaseNode):
-    name = "开始节点"
-    inputs = {}  # 没有输入参数
-    outputs = {
-        "text": "string",  # 文本内容
-        "image": "string"  # 图片以 Base64 编码的字符串传递
-    }
-
-    async def run(self, text: str = None, image: bytes = None, **kwargs) -> Dict[str, Any]:
-        # 查空：文本不能为空字符串
-        if not text or text.strip() == "":
-            raise ValueError("文本输入不能为空")
-
-        # 查空：图片数据不能为空
-        if image is None or len(image) == 0:
-            raise ValueError("图片输入不能为空")
-
-        # 将图片字节转换为 Base64 字符串，便于在节点间传递
-        image_base64 = base64.b64encode(image).decode('utf-8')
-
-        return {
-            "text": text,
-            "image": image_base64
-        }

backend/app/nodes/save_input_to_json.py

@@ -0,0 +1,145 @@
+import json
+from typing import Dict, Any
+from datetime import datetime
+import config as cfg
+from pathlib import Path
+
+
+def save_input_to_json(
+        mes: str,
+        role_name: str,
+        chat_name: str,
+        name: str,
+        is_user: bool,
+        floor_number: int = 0
+) -> Dict[str, Any]:
+    """
+    保存消息到JSONL文件或处理重roll请求
+
+    参数:
+        mes: 消息内容
+        role_name: 角色名称
+        chat_name: 对话名称
+        name: 发送者名称
+        is_user: 是否为用户消息
+        floor_number: 楼层号(对话中的第几次回复)，用于判断是否为重roll请求
+
+    返回:
+        更新后的消息对象
+    """
+    config = cfg.settings
+    file_path = config.BASE_PATH / "data" / "chat" / role_name / f"{chat_name}.jsonl"
+
+    # 确保目录存在
+    Path(file_path).parent.mkdir(parents=True, exist_ok=True)
+
+    # 读取文件内容
+    try:
+        with open(file_path, 'r', encoding='utf-8') as f:
+            lines = f.readlines()
+    except FileNotFoundError:
+        lines = []
+
+    # 判断是否为重roll请求
+    is_regenerate = False
+    target_index = -1
+
+    if lines and floor_number > 0:
+        # 计算当前楼层号
+        current_floor = len(lines)
+
+        # 如果floor_number与当前楼层号相同，则为重roll请求
+        if floor_number == current_floor:
+            # 找到最后一条非用户消息
+            for i in range(len(lines) - 1, -1, -1):
+                try:
+                    line_data = json.loads(lines[i])
+                    if not line_data.get('is_user', False):
+                        is_regenerate = True
+                        target_index = i
+                        break
+                except json.JSONDecodeError:
+                    continue
+
+    # 处理重roll逻辑
+    if is_regenerate:
+        # 解析目标消息
+        try:
+            target_message = json.loads(lines[target_index])
+        except json.JSONDecodeError:
+            raise ValueError(f"无法解析楼层 {floor_number} 的JSON数据")
+
+        # 初始化swipes数组
+        if target_message.get('swipes') is None:
+            target_message['swipes'] = []
+
+        # 将新回复添加到swipes数组
+        target_message['swipes'].append(mes)
+
+        # 更新swipe_id和content
+        target_message['swipes_id'] = len(target_message['swipes']) - 1
+        target_message['content'] = mes
+
+        # 更新文件内容
+        lines[target_index] = json.dumps(target_message, ensure_ascii=False) + '\n'
+
+        # 写回文件
+        with open(file_path, 'w', encoding='utf-8') as f:
+            f.writelines(lines)
+
+        return target_message
+
+    # 处理普通消息保存逻辑
+    else:
+        # 获取当前时间
+        current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+        # 构建消息对象
+        message = {
+            "role": role_name,
+            "chat": chat_name,
+            "content": mes,
+            "name": name,
+            "is_user": is_user,
+            "send_date": current_time,
+            "floor_number": len(lines) + 1,  # 记录楼层号
+            "swipes": [],
+            "swipes_id": 0
+        }
+
+        # 追加到文件
+        with open(file_path, 'a', encoding='utf-8') as f:
+            f.write(json.dumps(message, ensure_ascii=False) + '\n')
+
+        return message
+
+
+if __name__ == '__main__':
+    # 测试普通消息保存
+    # save_input_to_json(
+    #     mes="你好",
+    #     role_name="test",
+    #     chat_name="111",
+    #     name="用户",
+    #     is_user=True,
+    #     floor_number=0
+    # )
+    #
+    # save_input_to_json(
+    #     mes="你好，我是AI助手",
+    #     role_name="test",
+    #     chat_name="111",
+    #     name="AI",
+    #     is_user=False,
+    #     floor_number=1
+    # )
+
+    # 测试重roll最后一条AI消息
+    save_input_to_json(
+        mes="这是重roll后的新回复2",
+        role_name="test",
+        chat_name="111",
+        name="AI",
+        is_user=False,
+        floor_number=2  # 与当前楼层号相同，表示重roll
+    )

config.py

@@ -0,0 +1,45 @@
+import os
+from pathlib import Path
+from dotenv import load_dotenv
+
+# 1. 动态计算项目根目录
+# 假设 config.py 位于 backend/ 目录下
+# __file__ 指向本文件的绝对路径
+# .parent 指向 backend/ 目录
+# .parent.parent 指向项目根目录 (即包含 backend/ 和 frontend/ 的目录)
+PROJECT_ROOT = Path(__file__).resolve().parent
+
+# 2. 加载 .env 文件
+# 假设 .env 文件位于项目根目录下
+load_dotenv(PROJECT_ROOT / ".env")
+
+
+class Settings:
+    # --- 主模型配置 ---
+    MAIN_LLM_API_KEY = os.getenv("MAIN_LLM_API_KEY")
+    MAIN_LLM_MODEL = os.getenv("MAIN_LLM_MODEL", "gpt-3.5-turbo")
+    MAIN_LLM_BASE_URL = os.getenv("MAIN_LLM_BASE_URL", "https://api.openai.com/v1")
+    MAIN_LLM_MAX_TOKENS = int(os.getenv("MAIN_LLM_MAX_TOKENS", "4096"))
+    MAIN_LLM_STREAM = os.getenv("MAIN_LLM_STREAM", "true").lower() == "true"
+
+    # --- 路径配置 (核心修改) ---
+
+    # 强制使用计算出的项目根目录，不再依赖 .env 中的 BASE_PATH
+    BASE_PATH = PROJECT_ROOT
+
+    # 数据目录：固定为根目录下的 data 文件夹
+    # 即使 .env 里写了 DATA_PATH=/data，这里也会强制指向项目根目录下的 data
+    DATA_PATH = BASE_PATH / "data"
+
+    # 其他文件路径：基于 DATA_PATH 拼接
+    STATE_FILE = DATA_PATH / "state.json"
+    SCHEMA_FILE = DATA_PATH / "schema.json"
+    PRESETS_FILE = DATA_PATH / "presets.json"
+    REGEX_FILE = DATA_PATH / "regex_rules.json"
+    VECTORSTORE_PATH = DATA_PATH / "vectorstore"
+
+    # ... 其他配置 ...
+
+
+# 实例化配置对象
+settings = Settings()

data/chat/test/111.jsonl

@@ -0,0 +1,2 @@
+{"role": "test", "chat": "111", "content": "你好", "name": "用户", "is_user": true, "send_date": "2026-03-12 18:26:50", "floor_number": 1, "swipes": [], "swipes_id": 0}
+{"role": "test", "chat": "111", "content": "这是重roll后的新回复2", "name": "AI", "is_user": false, "send_date": "2026-03-12 18:26:50", "floor_number": 2, "swipes": ["这是重roll后的新回复", "这是重roll后的新回复2"], "swipes_id": 1}