Templates
Production RAG Starter Kit
A retrieval pipeline with the evaluation harness already wired in.
A complete retrieval-augmented generation service built the way I build them for clients: structure-aware chunking, hybrid dense and keyword search, cross-encoder reranking, and an evaluation harness so you can prove a change helped instead of hoping it did.
What is inside
- Structure-aware chunking that splits on headings and carries the heading trail into each chunk
- Hybrid retrieval combining dense vectors with BM25, so exact identifiers are still findable
- Cross-encoder reranking over a 30-candidate shortlist
- Evaluation harness with retrieval-level and answer-level scoring, plus 50 seed questions
- FastAPI service with streaming responses and per-request token accounting
- Docker Compose for local development against Postgres and pgvector
Built with
- Python
- FastAPI
- LangChain
- pgvector
- Sentence Transformers
- Docker
Next step
Have an AI problem worth solving?
Tell me what you are trying to build. I will tell you honestly whether AI is the right tool, what it would take, and how I would approach it.
Typical reply within 24 hours
Working from Dhaka, Bangladesh · remote worldwide