05 · Retrieval & Memory

Grounding models in your data. RAG and its building blocks — how to connect a model to external knowledge and give agents persistent memory.

Step	Topic	One-liner	Status
1	RAG	The core pattern: retrieve relevant context, inject into prompt, generate	🔴
2	Embedding models	Models that encode text into dense vectors for semantic comparison	🔴
3	Vector databases	Specialized stores for high-dimensional vectors with ANN search	🔴
4	Chunking strategies	How to split documents so retrieval finds the right piece	🔴
5	Semantic search	Finding documents by meaning, not keywords	🔴
6	Hybrid search	Combining BM25 keyword and dense vector search for better recall	🔴
7	Re-ranking	A second-pass model that reorders retrieval results by relevance	🔴
8	Knowledge graph	Structured entity-relationship graphs as an alternative to vector retrieval	🔴
9	GraphRAG	RAG over a knowledge graph — better for multi-hop and relational queries	🔴
10	Agentic RAG	Agents that decide when and how to retrieve, rather than retrieval being fixed	🔴
11	Long-context retrieval	Using extended context windows as an alternative to chunked retrieval	🔴
12	Working & episodic memory	Short-term (within session) and long-term (across sessions) agent memory	🔴

---

← Previous section: 04 · Model Optimization & Formats | Next section → 06 · Agents & Orchestration