优化提供商的导入和系统提示词的导入

.env.example

@@ -0,0 +1,77 @@
+# ===== 通用 API 配置 =====
+# 选择一个服务商，取消对应注释并填入 API_KEY
+# 注意：一次只使用一个服务商
+
+# --- SiliconFlow (硅基流动) - 推荐国内使用 ---
+API_KEY=your_siliconflow_api_key
+BASE_URL=https://api.siliconflow.cn/v1
+
+# --- OpenRouter ---
+# API_KEY=your_openrouter_api_key
+# BASE_URL=https://openrouter.ai/api/v1
+
+# --- DeepSeek ---
+# API_KEY=your_deepseek_api_key
+# BASE_URL=https://api.deepseek.com/v1
+
+# --- OpenAI ---
+# API_KEY=your_openai_api_key
+# BASE_URL=https://api.openai.com/v1
+
+# --- Anthropic ---
+# API_KEY=your_anthropic_api_key
+# BASE_URL=https://api.anthropic.com/v1
+
+# --- 火山引擎 (Volcengine) ---
+# API_KEY=your_volcengine_api_key
+# BASE_URL=https://ark.cn-beijing.volces.com/api/v3
+
+# ===== 模型配置 =====
+# 根据选择的服务商使用对应模型名称
+
+# --- SiliconFlow 模型 ---
+MODEL=Qwen/Qwen3.5-122B-A10B
+EMBEDDER=Qwen/Qwen3-Embedding-4B
+RERANK_MODEL=Qwen/Qwen3-Reranker-4B
+RERANK_BINDING_HOST=https://api.siliconflow.cn/v1/rerank
+
+# --- OpenRouter 示例 ---
+# MODEL=meta-llama/llama-3.1-70b-instruct
+# EMBEDDER=text-embedding-3-small
+
+# --- DeepSeek 示例 ---
+# MODEL=deepseek-chat
+# EMBEDDER=deepseek-embedding
+
+# --- OpenAI 示例 ---
+# MODEL=gpt-4o
+# EMBEDDER=text-embedding-3-small
+
+# --- Anthropic 示例 ---
+# MODEL=claude-3-5-sonnet-20241022
+# EMBEDDER=（需配合其他嵌入服务使用）
+
+# --- 火山引擎 示例 ---
+# MODEL=ep-20240910063496-xxxxxx
+# EMBEDDER=doubao-embedding-text-xxxxx
+
+# ===== RAG Pipeline =====
+GRADE_MODEL=Qwen/Qwen3.5-122B-A10B
+
+# ===== Milvus =====
+MILVUS_HOST=127.0.0.1
+MILVUS_PORT=19530
+MILVUS_COLLECTION=embeddings_collection
+
+# ===== Auto-merging =====
+AUTO_MERGE_ENABLED=true
+AUTO_MERGE_THRESHOLD=2
+LEAF_RETRIEVE_LEVEL=3
+
+# ===== Tools (可选) =====
+AMAP_WEATHER_API=https://restapi.amap.com/v3/weather/weatherInfo
+AMAP_API_KEY=your_amap_api_key
+
+# ===== Server =====
+HOST=127.0.0.1
+PORT=8000

.gitignore

@@ -16,3 +16,6 @@ volumes/
 # OS files
 .DS_Store
 data/
+
+# cc
+.claude/

CLAUDE.md

@@ -0,0 +1,84 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## 常用命令
+
+```bash
+# 安装依赖
+uv sync
+
+# 启动 Milvus（向量库）
+docker compose up -d
+
+# 启动后端服务
+uv run uvicorn backend.app:app --host 127.0.0.1 --port 8000 --reload
+
+# 访问
+# 前端: http://127.0.0.1:8000/
+# API 文档: http://127.0.0.1:8000/docs
+```
+
+## 环境配置
+
+创建 `backend/.env` 文件：
+
+```bash
+API_KEY=your_api_key_here
+MODEL=Qwen/Qwen3.5-122B-A10B          # 主模型
+EMBEDDER=Qwen/Qwen3-Embedding-4B      # 嵌入模型
+RERANK_MODEL=Qwen/Qwen3-Reranker-4B   # 重排序模型
+BASE_URL=https://api.siliconflow.cn/v1 # 兼容 OpenAI API 的服务商
+MILVUS_HOST=127.0.0.1
+MILVUS_PORT=19530
+```
+
+## 项目架构
+
+```
+SuperMew/
+├── backend/                 # FastAPI 后端
+│   ├── app.py              # 入口、CORS、静态资源
+│   ├── api.py              # 聊天/会话/文档 API
+│   ├── agent.py            # LangChain Agent + 会话存储
+│   ├── config.py           # 环境变量配置
+│   ├── schemas.py          # Pydantic 数据模型
+│   ├── soul/
+│   │   └── soul.md         # Agent 系统提示词
+│   ├── tools.py            # 工具：天气查询、知识库检索
+│   ├── rag_pipeline.py     # RAG 工作流（LangGraph）
+│   ├── rag_utils.py        # 检索、查询重写、HyDE
+│   ├── embedding.py        # 稠密向量 + BM25 稀疏向量
+│   ├── milvus_client.py    # Milvus 混合检索
+│   ├── milvus_writer.py    # 向量写入
+│   ├── document_loader.py  # PDF/Word 分块
+│   └── parent_chunk_store.py  # 父级分块 DocStore
+├── frontend/               # Vue 3 单页应用
+│   ├── index.html
+│   ├── script.js
+│   └── style.css
+├── data/                   # 本地存储
+│   ├── customer_service_history.json  # 会话历史
+│   ├── parent_chunks.json             # 父级分块
+│   └── documents/                     # 上传文档
+└── docker-compose.yml      # Milvus 向量库
+```
+
+## 核心模块
+
+- **RAG 管道** (`rag_pipeline.py`)：基于 LangGraph 构建，包含 retrieve → grade → rewrite → retrieve\_expanded 节点
+- **检索** (`rag_utils.py`)：Hybrid Search (Dense + Sparse + RRF) + Jina Rerank
+- **分块** (`document_loader.py`)：三级滑动窗口（L1/L2/L3），仅 L3 叶子块写入 Milvus
+- **记忆** (`agent.py`)：
+  - `SummarizationMiddleware` 在 token 超过 80000 时触发
+  - 保留最近 12 条消息（6 轮交互）
+  - 使用 `backend/soul/soul.md` 作为系统提示词
+
+## 优化方向
+
+优先优化**上下文工程**和**记忆系统**：
+
+- 改进会话摘要策略
+- 考虑集成 Mem0/LangMem 等外部记忆方案
+- 优化上下文压缩和注入方式 -- Context Engineering
+

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Chasen Liao
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.