All work
Self-initiated technical buildAI / RAG
RAG Knowledge Platform
2025|Next.js, TypeScript, PostgreSQL, pgvector, OpenAI, LangChain, Vercel
The Problem
Internal knowledge bases become stale and unsearchable as teams grow. Standard search returns documents, not answers.
The Solution
Built a retrieval-augmented generation system that indexes structured and unstructured documents, maintains source traceability, and answers natural-language queries with grounded responses.
Key Decisions
- 1Chose pgvector over a dedicated vector store to keep the stack footprint minimal
- 2Implemented chunking strategy with contextual overlap to preserve semantic coherence across section boundaries
- 3Used streaming responses to reduce perceived latency on longer answers
Stack
Next.jsTypeScriptPostgreSQLpgvectorOpenAILangChainVercel
Start a similar project
Tell us about your constraints and goals.