forked from fangtianchen/algonotes_rag
137 lines
2.9 KiB
Markdown
137 lines
2.9 KiB
Markdown
# 🔧 AlgoNotes RAG 配置说明
|
||
|
||
## 配置文件
|
||
|
||
项目使用两层配置：
|
||
|
||
```
|
||
.env → 敏感信息（API Key，不提交 Git）
|
||
config.toml → 业务配置（提交 Git，${VAR} 引用环境变量）
|
||
```
|
||
|
||
## 环境变量
|
||
|
||
复制 `.env.example` 为 `.env` 并填入你的 Key：
|
||
|
||
```bash
|
||
# .env
|
||
OPENAI_API_KEY = "your-api-key-here"
|
||
```
|
||
|
||
## config.toml 说明
|
||
|
||
```toml
|
||
# ── Chat Model ──
|
||
[llm]
|
||
model = "Qwen3-Next-80B-A3B-Instruct"
|
||
base_url = "https://ai.gitee.com/v1"
|
||
api_key = "${OPENAI_API_KEY}"
|
||
temperature = 0.7
|
||
max_tokens = 4096
|
||
timeout = 60.0
|
||
|
||
# ── Embedding Model ──
|
||
[embedding]
|
||
model = "Qwen/Qwen3-Embedding-4B"
|
||
base_url = "https://ai.gitee.com/v1"
|
||
api_key = "${OPENAI_API_KEY}"
|
||
|
||
# ── Reranker Model ──
|
||
[reranker]
|
||
model = "Qwen3-Reranker-4B"
|
||
base_url = "https://ai.gitee.com/v1"
|
||
api_key = "${OPENAI_API_KEY}"
|
||
top_n = 3
|
||
|
||
# ── Logging ──
|
||
[logging]
|
||
level = "INFO"
|
||
file = "logs/app.log"
|
||
max_bytes = 10485760
|
||
backup_count = 5
|
||
|
||
# ── Storage Paths ──
|
||
[store]
|
||
chroma_dir = "data/chroma_db"
|
||
files_dir = "data/files"
|
||
sqlite_path = "data/sql_db/notes.db"
|
||
```
|
||
|
||
## 切换模型提供商
|
||
|
||
只需修改 `config.toml` 中的 `base_url` 和 `model` 即可切换：
|
||
|
||
### 生产环境（如比赛要求使用的 Gitee.AI）
|
||
|
||
```toml
|
||
[llm]
|
||
model = "DeepSeek-R1"
|
||
base_url = "https://ai.gitee.com/v1"
|
||
|
||
[embedding]
|
||
model = "Qwen/Qwen3-Embedding-4B"
|
||
base_url = "https://ai.gitee.com/v1"
|
||
```
|
||
|
||
### 开发环境（如硅基流动）
|
||
|
||
```toml
|
||
[llm]
|
||
model = "deepseek-ai/DeepSeek-V4-Flash"
|
||
base_url = "https://api.siliconflow.cn/v1"
|
||
|
||
[embedding]
|
||
model = "Qwen/Qwen3-Embedding-4B"
|
||
base_url = "https://api.siliconflow.cn/v1"
|
||
```
|
||
|
||
## 代码中使用
|
||
|
||
```python
|
||
from src.config import config
|
||
|
||
# LLM 配置
|
||
model = init_chat_model(
|
||
model=config.llm.model,
|
||
model_provider="openai",
|
||
openai_api_key=config.llm.api_key,
|
||
base_url=config.llm.base_url,
|
||
temperature=config.llm.temperature,
|
||
)
|
||
|
||
# Embedding 配置
|
||
embeddings = init_embeddings(
|
||
model=config.embedding.model,
|
||
provider="openai",
|
||
openai_api_key=config.embedding.api_key,
|
||
base_url=config.embedding.base_url,
|
||
)
|
||
|
||
# 日志配置
|
||
logger = setup_logger()
|
||
# 自动读取 config.logging.level、config.logging.file 等
|
||
```
|
||
|
||
## 配置优先级
|
||
|
||
```
|
||
硬编码默认值 < config.toml < 环境变量 ${VAR}
|
||
```
|
||
|
||
1. 如果 `config.toml` 不存在，使用 Pydantic 模型中的硬编码默认值
|
||
2. 如果 `config.toml` 存在，以其为准
|
||
3. `${OPENAI_API_KEY}` 等占位符在加载时被 `.env` 中的实际值替换
|
||
|
||
## 公共题库（CPGraph）
|
||
|
||
`[cp_graph]` 配置段控制是否查询公共题库：
|
||
|
||
```toml
|
||
[cp_graph]
|
||
enabled = false # 是否启用公共题库查询
|
||
url = "https://mcp.cpgraph.top/mcp" # CPGraph MCP 服务地址
|
||
```
|
||
|
||
- `enabled = false`：仅查询个人笔记（默认）
|
||
- `enabled = true`：同时查询公共题库，需要 CPGraph MCP 服务可用
|