# Changelog

> Changes to the gateway API and its documentation.

## September 29, 2026

- The documentation was restructured: separate pages for each endpoint,
  streaming, limits, pricing and each tool; added the FAQ and this page.
- Both error envelopes are documented (OpenAI, and Anthropic for
  `/v1/messages`), with the full list of codes, including
  `429 rate_limit_exceeded`, `413`, `422 generation_failed`,
  `context_length_exceeded`, `content_policy_violation`.
- Request parameters are documented per endpoint.
- Claude Code: the required `ANTHROPIC_MODEL` setting was added.

## September 2026

- A limit on concurrent requests per account: when exceeded —
  `429 rate_limit_exceeded` with `Retry-After`.
- Provider refusals come with clear codes: `content_policy_violation`,
  `invalid_image`, `generation_failed`, `context_length_exceeded`,
  `unsupported_value`, `invalid_value`.
- Request parameter policy: unsupported parameters are stripped, and those that
  change the shape of the response are rejected with
  `400 unsupported_parameter`.
- `/v1/messages` errors come in the Anthropic envelope.
- Private models and personal prices: with your key, `GET /v1/models` lists
  models opened to your account.