API reference

The runtime serves exactly three public endpoints, and never reads your database on the request path — every query is answered from your published list at the edge. Every SDK binding in the integration guides calls these underneath; this reference is for calling them directly.

Prefer a machine-readable description? The same API is published as an OpenAPI 3.1 document.

• GET /v1/query — the autocomplete endpoint

  Call the runtime directly for pre-ranked autocomplete suggestions — the endpoint every SDK binding calls underneath.

  Copy GET /v1/query — the autocomplete endpoint instructions for AI / LLM

• GET /v1/bootstrap — session policy and version resolution

  Resolve a publishable key to its list identity, active version hash and effective query policy in one call, before the first query.

  Copy GET /v1/bootstrap — session policy and version resolution instructions for AI / LLM

• POST /v1/selection — report which suggestion was chosen

  Tell AutoSugges which suggestion a user picked, so future ranking for that list can improve. Entirely optional and fire-and-forget.

  Copy POST /v1/selection — report which suggestion was chosen instructions for AI / LLM

• Mint an access token

  The one control-plane call in this reference — exchange a secret key for a one-hour access token on your server, for querying a private list.

• Error codes

  The complete typed error vocabulary — every code, its HTTP status, whether it is safe to retry, and what to do about it.

• Control-plane API

  Manage lists, items, keys, subscriptions and publishing from your own server with a secret key — everything the dashboard itself calls, callable directly.