update _get_cache_key

README.md

@@ -59,80 +59,230 @@ docker-compose up -d
 ---
 
 ## 📖 使用指南
+# Catalog Agent API 文档
 
-公网：
+## 概述
-dev: https://catalog-agent-dev.gbase.ai
+本文档提供 Catalog Agent 服务的 API 接口说明，支持多语言对话、文件上传和解析功能。
-prod: https://catalog-agent.gbase.ai
 
-内网：
+## 服务地址
-prod http://catalog-agent.default.svc.cluster.local
-dev http://catalog-agent.gbase-dev.svc.cluster.local
 
-### 1. 聊天接口 (OpenAI 兼容)
+### 公网地址
+- **开发环境**: `https://catalog-agent-dev.gbase.ai`
+- **生产环境**: `https://catalog-agent.gbase.ai`
+
+### 内网地址
+- **生产环境**: `http://catalog-agent.default.svc.cluster.local`
+- **开发环境**: `http://catalog-agent.gbase-dev.svc.cluster.local`
+
+---
+
+## 接口列表
+
+### 1. 通用聊天接口 V1 (OpenAI 兼容)
 
 **端点**: `POST /api/v1/chat/completions`
 
+**认证方式**: Bearer Token (使用大语言模型的 API Key)
+
+**请求示例**:
 ```bash
 curl -X POST "{host}/api/v1/chat/completions" \
   -H "Content-Type: application/json" \
-  -H "Authorization: Bearer {api_key}"
+  -H "Authorization: Bearer {api_key}" \
   -d '{
-  "messages": [
+    "messages": [
-    {
+      {
-      "role": "user",
+        "role": "user",
-      "content": "1kg未満のノートPCを知りたいので表で出力してください"
+        "content": "1kg未満のノートPCを知りたいので表で出力してください"
-    }
+      }
-  ],
+    ],
-  "stream": true,
+    "stream": true,
-  "language": "ja",
+    "language": "ja",
-  "robot_type": "catalog_agent",
+    "robot_type": "catalog_agent",
-  "model": "gpt-4.1",
+    "model": "gpt-4.1",
-  "model_server": "https://one-dev.felo.me/v1",
+    "model_server": "https://one-dev.felo.me/v1",
-  "unique_id": "1624be71-5432-40bf-9758-f4aecffd4e9c",
+    "bot_id": "f4aecffd4e9c-624be71-5432-40bf-9758",
-  "tool_response": false
+    "dataset_ids": ["624be71-5432-40bf-9758-f4aecffd4e9c"],
-}'
+    "tool_response": false
+  }'
 ```
 
-### 2. 异步文件处理队列
+**请求参数说明**:
+| 参数 | 类型 | 必填 | 说明 |
+|------|------|------|------|
+| messages | array | 是 | 对话消息列表 |
+| stream | boolean | 是 | 是否启用流式输出 |
+| language | string | 是 | 语言代码: zh/en/ja |
+| robot_type | string | 是 | 固定值: catalog_agent |
+| model | string | 是 | AI 模型名称 |
+| model_server | string | 是 | AI 模型服务器地址 |
+| bot_id | string | 是 | 机器人唯一标识 |
+| dataset_ids | array | 是 | 知识库 ID 数组 |
+| tool_response | boolean | 是 | 是否返回工具响应 |
 
-#### 启动队列系统
+---
 
+### 2. GBase 聊天接口 V2
+
+**端点**: `POST /api/v2/chat/completions`
+
+**认证方式**: Bearer Token (使用 `md5(master:{bot_id})` 生成)
+
+> 注意：此接口的模型、服务器、API Key、数据集信息会自动从 GBase 读取，无需手动传入。
+
+**请求示例**:
 ```bash
-# 终端1：启动队列消费者
+curl -X POST "{host}/api/v2/chat/completions" \
-poetry run python task_queue/consumer.py --workers 2
+  -H "Content-Type: application/json" \
-
+  -H "Authorization: Bearer {md5(master:{bot_id})}" \
-# 终端2：启动API服务器
+  -d '{
-poetry run python fastapi_app.py
+    "messages": [
+      {
+        "role": "user",
+        "content": "1kg未満のノートPCを知りたいので表で出力してください"
+      }
+    ],
+    "stream": true,
+    "language": "ja",
+    "bot_id": "f4aecffd4e9c-624be71-5432-40bf-9758",
+    "tool_response": false
+  }'
 ```
 
-#### 提交异步任务
+**请求参数说明**:
+| 参数 | 类型 | 必填 | 说明 |
+|------|------|------|------|
+| messages | array | 是 | 对话消息列表 |
+| stream | boolean | 是 | 是否启用流式输出 |
+| language | string | 是 | 语言代码: zh/en/ja |
+| bot_id | string | 是 | 机器人唯一标识 |
+| tool_response | boolean | 是 | 是否返回工具响应 |
 
+---
+
+### 3. 文件上传
+
+**端点**: `POST /api/v1/upload`
+
+**请求方式**: 表单数据上传
+
+**参数名**: `file`
+
+**响应示例**:
+```json
+{
+  "success": true,
+  "message": "文件上传成功",
+  "filename": "12345678-1234-5678-9abc-123456789def.pdf",
+  "original_filename": "document.pdf",
+  "file_path": "projects/uploads/12345678-1234-5678-9abc-123456789def.pdf"
+}
+```
+
+---
+
+### 4. 文件解析
+
+#### 4.1 全量文件解析
+
+**端点**: `POST /api/v1/files/process/async`
+
+**请求示例**:
 ```bash
 curl -X POST "{host}/api/v1/files/process/async" \
   -H "Content-Type: application/json" \
   -d '{
-    "unique_id": "1624be71-5432-40bf-9758-f4aecffd4e9c",
+    "dataset_id": "624be71-5432-40bf-9758-f4aecffd4e9c",
     "files": {
-      "group_name":{
+      "group_name1": [
-          "public/document.txt",
+        "public/document.txt",
-          "public/data.zip",
+        "public/data.zip",
-          "public/goods.xlsx"
+        "public/goods.xlsx"
-      }
+      ],
+      "group_name2": [
+        "public/document.txt",
+        "public/data.zip",
+        "public/goods.xlsx"
+      ]
     }
   }'
 ```
 
-**响应**:
+#### 4.2 增量文件解析
+
+**端点**: `POST /api/v1/files/process/incremental`
+
+**请求示例**:
+```bash
+curl -X POST "{host}/api/v1/files/process/incremental" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "dataset_id": "624be71-5432-40bf-9758-f4aecffd4e9c",
+    "files_to_add": {
+      "group_name1": [
+        "projects/uploads/report_list/1090571652550791207.md"
+      ],
+      "group_name2": [
+        "projects/uploads/report_list/1090570966043889684.md"
+      ]
+    },
+    "files_to_remove": {
+      "group_name1": [
+        "projects/uploads/report_list/1090570712888283212.md"
+      ],
+      "group_name2": [
+        "projects/uploads/report_list/1090570712888283214.md"
+      ]
+    }
+  }'
+```
+
+**解析响应**:
 ```json
 {
   "success": true,
   "task_id": "abc-123-def",
-  "unique_id": "my_project_123",
+  "dataset_id": "624be71-5432-40bf-9758-f4aecffd4e9c",
   "task_status": "pending",
   "estimated_processing_time": 30
 }
 ```
 
+---
+
+### 5. 查询任务状态
+
+**端点**: `GET /api/v1/task/{task_id}/status`
+
+**请求示例**:
+```bash
+curl "{host}/api/v1/task/{task_id}/status"
+```
+
+**响应示例**:
+```json
+{
+  "success": true,
+  "task_id": "abc-123-def",
+  "status": "completed",
+  "dataset_id": "my_project_123",
+  "result": {
+    "status": "success",
+    "message": "成功处理了 2 个文档文件",
+    "processed_files": ["projects/my_project_123/dataset/docs/document.txt"]
+  }
+}
+```
+
+---
+
+## 注意事项
+
+1. **认证方式**: V1 和 V2 接口使用不同的认证机制，请根据接口版本选择合适的认证方式
+2. **必填参数**: 所有标记为"必填"的参数必须提供
+3. **文件路径**: 文件解析接口中使用的文件路径应为文件上传接口返回的 `file_path`
+4. **任务状态**: 文件解析为异步操作，需要通过任务状态接口查询处理进度
+
 #### 项目目录结构
 
 文件处理后，会在 `projects/{unique_id}/` 目录下生成以下结构：
@@ -156,66 +306,6 @@ projects/{unique_id}/
 - **分页数据层 (pagination.txt)**: 按页分割的数据，每页5000字符，便于检索
 - **向量嵌入层 (embedding.pkl)**: 文档的语义向量，支持语义搜索
 
-#### 查询任务状态
-
-```bash
-# 🎯 主要接口 - 只需要记住这一个
-curl "{host}/api/v1/task/{task_id}/status"
-```
-
-**状态响应**:
-```json
-{
-  "success": true,
-  "task_id": "abc-123-def",
-  "status": "completed",
-  "unique_id": "my_project_123",
-  "result": {
-    "status": "success",
-    "message": "成功处理了 2 个文档文件",
-    "processed_files": ["projects/my_project_123/dataset/docs/document.txt"]
-  }
-}
-```
-
-### 3. Python 客户端示例
-
-```python
-import requests
-import time
-
-def submit_and_monitor_task():
-    # 1. 提交任务
-    response = requests.post(
-        "http://localhost:8001/api/v1/files/process/async",
-        json={
-            "unique_id": "my_project",
-            "files": {"docs": ["public/file.txt"]}
-        }
-    )
-    
-    task_id = response.json()["task_id"]
-    print(f"任务已提交: {task_id}")
-    
-    # 2. 监控任务状态
-    while True:
-        response = requests.get(f"http://localhost:8001/api/v1/task/{task_id}/status")
-        data = response.json()
-        
-        status = data["status"]
-        print(f"任务状态: {status}")
-        
-        if status == "completed":
-            print("🎉 任务完成！")
-            break
-        elif status == "failed":
-            print("❌ 任务失败！")
-            break
-        
-        time.sleep(2)
-
-submit_and_monitor_task()
-```
 
 ### 4. 项目目录树接口
 

utils/file_loaded_agent_manager.py

@@ -39,11 +39,17 @@ class FileLoadedAgentManager:
         self.creation_times: Dict[str, float] = {}  # 创建时间记录
         self.max_cached_agents = max_cached_agents
     
-    def _get_cache_key(self, bot_id: str, system_prompt: str = None, mcp_settings: List[Dict] = None) -> str:
+    def _get_cache_key(self, bot_id: str, model_name: str = None, api_key: str = None,
-        """获取包含 bot_id、system_prompt 和 mcp_settings 的哈希值作为缓存键
+                       model_server: str = None, generate_cfg: Dict = None,
+                       system_prompt: str = None, mcp_settings: List[Dict] = None) -> str:
+        """获取包含所有相关参数的哈希值作为缓存键
 
         Args:
             bot_id: 机器人项目ID
+            model_name: 模型名称
+            api_key: API密钥
+            model_server: 模型服务器地址
+            generate_cfg: 生成配置
             system_prompt: 系统提示词
             mcp_settings: MCP设置列表
 
@@ -53,6 +59,10 @@ class FileLoadedAgentManager:
         # 构建包含所有相关参数的字符串
         cache_data = {
             'bot_id': bot_id,
+            'model_name': model_name or '',
+            'api_key': api_key or '',
+            'model_server': model_server or '',
+            'generate_cfg': json.dumps(generate_cfg or {}, sort_keys=True),
             'system_prompt': system_prompt or '',
             'mcp_settings': json.dumps(mcp_settings or [], sort_keys=True)
         }
@@ -124,7 +134,8 @@ class FileLoadedAgentManager:
         final_system_prompt = load_system_prompt(project_dir, language, system_prompt, robot_type, bot_id)
         final_mcp_settings = load_mcp_settings(project_dir, mcp_settings, bot_id, robot_type)
 
-        cache_key = self._get_cache_key(bot_id, final_system_prompt, final_mcp_settings)
+        cache_key = self._get_cache_key(bot_id, model_name, api_key, model_server,
+                                        generate_cfg, final_system_prompt, final_mcp_settings)
         
         # 检查是否已存在该助手实例
         if cache_key in self.agents: