Fetching metadata from the HF Docker repository...
-
Week4: Gradio 前端 + PDF 入库 + HF Spaces 部署件
Maggei +1: Claude Opus 4.8 - frontend/app.py:Gradio 三 Tab(研报分析/问答/上传),纯 HTTP 客户端接 FastAPI
- rag/ingest.py + read_pdf:研报 PDF 入库(CLI/API /upload/pdf/前端三路径)
- agents/llm.py + rag/embed.py:env 驱动 OpenAI 兼容后端,模型名 env 化(部署用)
- app.py + Dockerfile + .dockerignore + README front-matter:HF Spaces 单进程部署件
- README.md:架构/快速启动/接口/评估结论
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
-
Week3: FastAPI layer + wire RAG into graph qa node
Maggei +1: Claude Opus 4.8 - graph/workflow.py: qa 节点未给 context 时自动走 HybridRetriever 混合检索注入,返回 sources;检索器单例 + refresh hook
- api/main.py: /health /analyze /chat /chat/stream(流式) /upload(入库即刷新);全请求经 LangGraph,同步图调用走线程池并发
- agents/llm.py: 加 chat_stream 流式接口
- 5 接口端到端实测通过(含上传后即时可检索);requirements 加 fastapi+uvicorn
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>