search

Retrieval-Augmented Generation — Deep Dive

From raw documents to grounded answers. Each chapter: visual journey overview + under-the-hood deep dive.
Co-Created by Kiran Shirol and Claude
Core Stack RAG Embeddings Vector DBs LlamaIndex LangChain
home Learning Portal play_arrow Start Learning summarize Key Insights dictionary Glossary 11 chapters · Each with High Level + Under the Hood
Ingestion

Documents to Vectors

Loading, chunking, embedding, and storing — the offline indexing pipeline.
1
lightbulb
What Is RAG & Why It Matters
The knowledge cutoff problem, hallucination, and the retrieve-then-generate pattern.
2
upload_file
Document Loading & Preprocessing
PDFs, web pages, databases, APIs — loaders, metadata extraction, and normalization.
3
content_cut
Chunking Strategies
Fixed-size, recursive, semantic chunking, overlap, and parent-child strategies.
4
scatter_plot
Embeddings: Text to Vectors
OpenAI, Cohere, BGE, E5, Matryoshka representations, and the MTEB benchmark.
5
database
Vector Stores & Indexing
Pinecone, Weaviate, Qdrant, Chroma, pgvector, HNSW, IVF, and hybrid indexes.
Retrieval

Query to Context

Finding the right chunks, transforming queries, and synthesizing answers.
6
manage_search
Retrieval Strategies
Dense, sparse (BM25), hybrid search, cross-encoder reranking, RRF, and MMR.
7
auto_fix_high
Query Transformation
HyDE, multi-query expansion, step-back prompting, decomposition, and self-query.
8
edit_note
Generation: Synthesizing Answers
Stuff/map-reduce/refine chains, citation, lost-in-the-middle, and faithfulness.
Production

Advanced Patterns & Production

GraphRAG, agentic RAG, the solutions landscape, evaluation, and scaling.
9
hub
Advanced RAG Patterns
GraphRAG, agentic RAG, Self-RAG, Corrective RAG, multi-modal, and adaptive loops.
10
landscape
RAG Solutions Landscape
LlamaIndex, LangChain, Haystack, Vectara, Cohere RAG, AWS Bedrock KB, and more.
11
monitoring
RAG in Production & Evaluation
RAGAS, DeepEval, semantic caching, incremental indexing, monitoring, and A/B testing.