Back to Context Engineering
Library
Retrieval-augmented generation
Ground answers in your data.
Retrieval
Overview
Retrieval-Augmented Generation grounds an LLM in external knowledge by retrieving relevant chunks and injecting them into the prompt. It reduces hallucination and lets models answer over private, up-to-date data.
How it works
RetrievalClientServiceEdgeData
Step by step, with examples
- 1
Embed docs
- Chunk documents into a vector store.
- 2
Top-k
- ANN search for the query.
- 3
Inject
- Add retrieved snippets to the prompt.
- 4
Grounded
- Answer with citations.
- Example: cuts hallucination
Overview
Chunk + embed a corpus, retrieve top-k by similarity, and inject the snippets so the model answers from grounded facts.
Common pitfalls
- Bad chunking
- No reranking
- Stale index
- No citation/grounding check
Where this content comes from
For full transparency, this content is curated and verified from these sources:
Frontier-lab prompting & agent guidesRetrieval-augmented generation literatureOppZen-authored context-engineering playbooks