Skip to content
HUIAHUB

Topic

RAG

Retrieval-augmented generation: grounding a model in your own documents with embeddings, chunking, vector search and evaluation, step by step.

Retrieval-augmented generation (RAG) is a technique that finds passages relevant to a question in your own documents and adds them to a language model's prompt, so the answer is grounded in that data. To start with RAG, follow the guide to building a RAG pipeline with Postgres and pgvector step by step, then read the guide on choosing between pgvector, Qdrant, Chroma and Milvus.

The tool pages cover products that answer from your documents, including Gemini Notebook for cited answers over your sources and website support agents that work from your help content. Graphify, an open-source project, takes a different route and builds a queryable knowledge graph from code and documents without embeddings.

The videos are talks and explainers for developers on agentic RAG, agent memory, production metrics for LLM apps and when to fix retrieval instead of fine-tuning a model. Their cards show length and level. Start with the guides if you want to build a pipeline yourself, and with the tool pages if you want a hosted product.