Resources & further reading
External material referenced by or complementary to this series.
Deep dives on individual topics
- Approximate Nearest Neighbor (ANN) search β how HNSW, IVF, trees, and hashing make vector search fast at scale (referenced in post 2b): https://outcomeschool.com/blog/how-does-approximate-nearest-neighbor-ann-search-work
- How does a Reranker work? β bi-encoder vs cross-encoder, ColBERT and late interaction (post 03 follows a similar structure): https://outcomeschool.com/blog/how-does-a-reranker-work
- How does an Embedding Cache work? β caching textβvector calls, the complement to post 06βs semantic cache: https://outcomeschool.com/blog/how-does-an-embedding-cache-work
- Agentic RAG β the concept overview post 07βs explanation follows: https://outcomeschool.com/blog/agentic-rag
- How does LangChain work? β chains, memory, agents in one framework: https://outcomeschool.com/blog/how-does-langchain-work
Primary sources
- Contextual Retrieval (Anthropic, 2024) β the technique behind post 05, including contextual BM25 and the ~67% retrieval-failure reduction figure: https://www.anthropic.com/news/contextual-retrieval
- Reciprocal Rank Fusion (Cormack, Clarke & Buettcher, 2009) β the RRF formula and the k=60 constant used in posts 2c and 04.
- OpenAI Agents SDK β the agent framework used in post 07: https://github.com/openai/openai-agents-python
- Okapi BM25 (Robertson & Walker) β the ranking function built from scratch in post 2a.
- HyDE: Precise Zero-Shot Dense Retrieval without Relevance Labels (Gao et al., 2022) β the hypothetical-document technique in post 04.
- Interactive tokenizer β watch text become tokens (post 2aβs aside): https://platform.openai.com/tokenizer
Libraries used throughout
- FastEmbed β local embeddings, sparse BM25 vectors, and cross-encoders: https://github.com/qdrant/fastembed
- Qdrant β vector database used for dense, sparse, and hybrid search: https://qdrant.tech
- FlashRank β local ONNX cross-encoder reranking: https://github.com/PrithivirajDamodaran/FlashRank
- LlamaIndex β the production framework in the course code: https://www.llamaindex.ai