About BuildRAG

I built BuildRAG because too many explanations make RAG synonymous with embeddings and a vector database. That shortcut hides the harder problem: choosing, evaluating, and operating the right retrieval path for an information need.

BuildRAG teaches one system from transparent primitives. The Course starts with glob, grep, and selective reading. It then adds structured lookup, ranked lexical search, embeddings, PostgreSQL, hybrid ranking, grounded generation, evaluation, agents, caching, monitoring, and deployment.

Editorial standards

• Teach the baseline before the abstraction.
• Prefer focused libraries over broad RAG frameworks.
• Separate retrieval quality from Answer quality.
• Preserve Source locations and verify Citations.
• Date provider, model, and pricing claims that can change.
• Cite primary sources where practical.
• Correct errors in public.

BuildRAG does not favor a model provider, vector database vendor, or agent framework. Provider examples exist to make the Course runnable.

Author

I'm Goh Young Jiet. I have spent the last several years working with RAG and language-model applications. BuildRAG is my attempt to make the full pipeline understandable without hiding it inside a framework.

When the repository becomes public, every page will link to its source. You can report an error or propose a correction through GitHub.