How zhenqiangshi/about_ai_system works
This repository is an architecture and learning guide, not an executable application. Its documented principal workflow is user question → RAG retrieval and context assembly → LLM generation; a separate ingestion path turns image/PDF content into structured text for the knowledge base. The graph also preserves documented inference, agent, gateway, and observability capabilities. Relationships are limited to README statements; unsampled implementation details are not inferred.
Main entry points: RAG流程及技术.md RAG retrieval · OCR.md OCR extraction · Agent Agent system · LLM.md Language model
Components 12
Knowledge pipeline
- OCR extraction
OCR/OCR.md - RAG retrieval
RAG/RAG流程及技术.md
Model serving
- Language model
LLM/LLM.md - vLLM serving
LLM/vLLM.md - Xinference serving
LLM/Xinference.md
Agent execution
- Agent system
Agent
Operations
- AI gateway
LLM/AI网关相关.md - LiteLLM API layer
Observability/LiteLLM.md - Langfuse observability
Observability/langfuse.md
Other components
- User
- Image/PDF input
- External tools
Connections 5
- User → RAG retrieval: asks question
README.md - Image/PDF input → OCR extraction: submits files
README.md - OCR extraction → RAG retrieval: injects text
README.md - RAG retrieval → Language model: generates answer
README.md - Agent system → External tools: calls tools
README.md