RAG Assistant — augmented search — AI POC
The idea. An LLM doesn't know my documents. RAG gives it access to an up-to-date document base, without retraining, with sourced answers.
What I prototype
- Ingestion & chunking of documents, embedding generation.
- Vector search with pgvector (PostgreSQL), reranking.
- LLM + context orchestration, citations and anti-hallucination guardrails.
- NestJS API and integration with the self-hosted stack (see LLM MLOps).
Stack
NestJS · pgvector · PostgreSQL · Embeddings · LLM · Docker
POC — AI prototype, not deployed to production.