jeikl/JeikChat
一个基于lanchain+rag+llm+vue+react的全栈JeikChat全能知识库Bot 支持多模型 语音识别 rag导入
What's novel
一个基于lanchain+rag+llm+vue+react的全栈JeikChat全能知识库Bot 支持多模型 语音识别 rag导入
Code Analysis
11 files read · 3 roundsA full-stack AI chat assistant that integrates LangGraph agent orchestration, MCP protocol tool calling, Qdrant-based RAG, and multi-model LLM support behind a FastAPI backend and React frontend.
Strengths
The MCP connection management with on-demand service loading, filtered STDIO sessions, and dynamic tool schema generation is well-designed and handles real-world complexity. The overall architecture cleanly separates concerns across agent, services, API routes, and frontend stores.
Weaknesses
No meaningful test suite exists (only demo scripts), sessions are stored in-memory despite claiming PostgreSQL persistence, a hardcoded API key is committed in RAG.py, and there's a bug where `selected_tools` is referenced before assignment in chatRouterStream.py. The LLM service uses ChatDeepSeek as a generic client for all providers, which is semantically misleading.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
85
Contributors
2
Files
107
Active weeks
3
Repository
Language
Python
Stars
1
Forks
0
License
—