RAG over Kubernetes Docs

Hybrid semantic RAG pipeline with safety guard over Kubernetes documentation.

Answers questions over Kubernetes docs locally with no API keys required.

Built a local RAG pipeline over Kubernetes documentation. It indexes markdown into ParadeDB/Postgres, retrieves with BM25 and pgvector, fuses results with RRF, reranks with a cross-encoder, filters weak chunks before generation, answers through Ollama, and evaluates runs with RAGAS.

Tech Stack

Python

LlamaIndex

PostgreSQL

ParadeDB

show more