Enrichment
LLM Enrichment
Enrich documents with LLM-generated fields like summaries, extractions, and classifications
Why do anything?
Documents lack derived fields that users need. Manual enrichment doesn't scale.
Why now?
LLMs can generate useful derived content (summaries, entities, classifications) automatically.
Why this feature?
LLM enriches documents with generated fields based on configurable prompts and output schemas.
How It Works
LLM enrichment adds AI-generated fields to documents.
1
Content Extraction
Extract content for LLM processing
2
Prompt Formation
Build prompt with document content
3
LLM Generation
Generate structured output
4
Field Addition
Add generated fields to document
Why This Approach
LLM generates human-quality enrichments. Schema enforcement ensures consistent output.
Where This Is Used
Integration
retriever = client.retrievers.create(stages=[{"type": "llm_enrich", "prompt": "..."}])