How mustafa-shoukat1/ragcraft-by-mustafa-shoukat works
Project type and purpose
• This is a monolithic “monorepo” hosting a curated collection of Retrieval-Augmented Generation (RAG) tutorial apps and AI-agent demos.
• Its purpose is to showcase and provide runnable examples of LLM-based RAG applications using various models (OpenAI, Anthropic, Google Gemini, DeepSeek, Qwen, Llama, Cohere, etc.), vector stores, and retrieval strategies.
File structure
• Top-level:
• A root Readme.md describing the overall “Awesome LLM Apps” catalog.
• Multiple peer directories, each a self-contained RAG or agent project (agentic_rag, agentic_rag_math_agent, ai_blog_search, autonomous_rag, …, vision_rag).
• Within each project folder:
• A README.md explaining that specific app’s purpose, usage, and dependencies.
• A requirements.txt listing Python dependencies.
• One or more Python modules (e.g., main.py, app.py, <agent>.py) implementing the RAG pipeline or agent logic.
• Some have subdirectories for config (.env), data ingestion, benchmarking, logs, or specialized modules (vector.py, query_router.py, guardrails.py).
• There are no shared libraries at root—each app is effectively a standalone microservice.
README for additional insights
• The root README classifies projects into categories (AI Agents, Multi-agent Teams, Voice, MCP, RAG, etc.).
• The RAG section lists these tutorial apps and hints at common patterns: retrieving from vector store, routing queries, LLM calls, corrective loops, hybrid search.