Back to changelog

One Document ID Across Search, Jobs and Listings

A document now has one ID, and every endpoint returns it.

Before, List Documents returned one ID, and search results, indexing jobs and the chunk endpoints returned a different one for the same file. Now document_id (document.id in v3 search) is the same value everywhere. It never changes, even when you upload the file again.

New field: version_id. This is the ID of the current version of the document's contents. It changes each time the document is re-indexed. It is the value search results and jobs used to return as the document ID.

Do I need to change anything?

  • If you take IDs from List Documents: no.
  • If you saved IDs from search results or jobs: those are version IDs. Each one works in requests until its document is re-indexed or overwritten, and then it stops matching. To match saved values against new responses, compare with version_id. From now on, save document_id, which keeps working after a re-index.
  • Documents indexed before August 2026: nothing changes. Both fields hold the same value.

Chunk IDs are unchanged. Treat a chunk_id as an opaque value.