Track 2 · RAG · Teal
RAG & knowledge systems
Retrieval-augmented generation for production: turn private documents into searchable chunks, fetch the right passages at query time, and force the LLM to answer from evidence—not parametric memory. This track covers chunking, vector stores, hybrid retrieval, advanced patterns, and ingestion pipelines. Assumes you can call an LLM API and understand embeddings from Track 1.
Guides in this track
Six deep-dive chapters. All guides are live—read in order.
Reading order: RAG explained → Chunking strategies → Vector databases → Retrieval & query rewriting → Advanced RAG patterns → Knowledge base ingestion
-
01
RAG explained
Retrieval-augmented generation, RAG vs fine-tuning, basic pipeline, failure modes, minimal implementation, eval basics, production checklist.
-
02
Chunking strategies
Fixed, recursive, semantic, and document-aware chunking; overlap; metadata; tables and code blocks.
-
03
Vector databases
Pinecone, pgvector, Chroma, Weaviate; ANN indexes, metadata filters, multi-tenant isolation.
-
04
Retrieval & query rewriting
Hybrid search, BM25 + dense, reranking, HyDE, multi-query, parent-child retrieval.
-
05
Advanced RAG patterns
Self-RAG, corrective RAG, GraphRAG, agentic retrieval, multi-hop reasoning.
-
06
Knowledge base ingestion
Incremental sync, webhooks, ACL propagation, deduplication, observability for ingest jobs.