Wang-Shuo/A-Guide-to-Retrieval-Augmented-LLM

Combines retrieval-based techniques with LLMs to address hallucinations, outdated knowledge, and long-tail information gaps by dynamically fetching relevant external context during inference. The system integrates information retrieval (vector embeddings, semantic search) with prompt augmentation, allowing LLMs to ground responses in curated data sources while improving explainability and cost efficiency over simply expanding context windows. Covers architectural components including data ingestion/chunking pipelines, retrieval indexing strategies, and generation mechanisms—with analysis of frameworks like vector databases and embedding models commonly deployed in production RAG applications.

306 stars. No commits in the last 6 months.