Skip to main content
POST /v1/chunk splits documents into stable, embedding-ready chunks, the same as the SDK chunk(). Feed the result to /v1/embed.

Request

markdown accepts a plain string, a list of strings, or a list of document dicts with a text field (such as the outputs from an ingestion endpoint).

Choosing chunk size

chunk_size is the main lever. Too small loses surrounding context; too large retrieves noise around the answer. Start at 2000 with 200 overlap (~10%), lower the size for dense factual content.

Request fields

Response fields