Chunk ID

A chunk ID is the unique identifier for a single chunk (segment) of a document. Chunks are produced during chunking after parsing; each is embedded and stored in the vector index with metadata that includes the chunk ID and document ID. Vector search and RAG return chunks; citations reference the chunk (and thus document and page) so you can show the exact source passage.

Where It Appears

• Vector index — Each vector has metadata including chunk ID and document-id for filtering and display.
• Retrieval — Top-k results return chunk IDs; the API or chat response uses them for citation.
• Citation — Citations map answer excerpts to chunk ID, document, and page (and optionally bounding box).

Chunk IDs are stable for the lifetime of the document’s index. Reindex may create new chunk IDs if chunking or content changes.

Related Concepts

Chunk ID identifies a chunk within a document. Metadata and citation use it. Vector search and retrieval return chunks by ID for RAG answers.