AI product Open source · MIT
Production Agentic RAG Course is a GitHub-hosted course on building production Retrieval-Augmented Generation systems through a seven-week, hands-on arXiv Paper Curator project. Its learning path covers Docker-based infrastructure, FastAPI, PostgreSQL, OpenSearch, Airflow, academic-paper ingestion with the arXiv API and Docling, BM25 keyword search, section-aware chunking, Jina embeddings, hybrid retrieval with reciprocal-rank fusion, local LLM serving with Ollama, streaming responses, Gradio, Langfuse tracing, Redis caching, and LangGraph-based agentic RAG.
The agentic RAG stage uses guardrail, retrieval, document-grading, query-rewriting, and generation nodes coordinated in a state-based LangGraph workflow. It includes adaptive retrieval, semantic relevance evaluation, domain-boundary detection, reasoning-step tracking, and an asynchronous Telegram bot interface.
The course is organized through weekly notebooks, blog posts, and tagged code releases in the repository. The project is distributed under the MIT License and is designed to run locally with Docker Compose, Python 3.12 or later, and the UV package manager; optional external services include Jina embeddings, Langfuse, and Telegram.