Changelog

Changes to the API, the SDKs and this documentation. Breaking changes are called out explicitly. For SDK releases, the SDK pages always describe the current version.

  • Added A machine-readable OpenAPI 3.1 description of every public endpoint, for client generators and coding agents.
  • Added Reference sections for audio formats and errors — every status code, what failed_stage values mean, and which requests are safe to retry.
  • Fixed Documentation showed diarized speakers as SPEAKER_0. The API numbers them from one — SPEAKER_1, SPEAKER_2 — dense and in order of first appearance. The API did not change; the examples were wrong.

  • Added These docs, at docs.speechrevolutions.com: API reference, guides, tutorials, framework integrations and migration guides from five other providers.
  • Added Webhook delivery headers documented — X-SR-Signature, X-SR-Event and X-SR-Delivery — with the retry contract.

· SDKs 0.3.0

  • Changed Breaking. The base error type is now SpeechRevolutionsError (SpeechRevolutionsException in C#), replacing STTError / SttException. The STTClient and AsyncSTTClient aliases are gone.
  • Removed Breaking. The SDKs no longer read STT_API_KEY or STT_BASE_URL. Use SPEECHREVOLUTIONS_API_KEY and SPEECHREVOLUTIONS_BASE_URL.
  • Added Published per-endpoint rate limits.

· SDKs 0.2.1–0.2.3

  • Fixed Each SDK announced a version it was not. The User-Agent is now derived from the package's own metadata, so it always matches what is installed.
  • Fixed Job-creating requests are no longer blindly retried. A lost response to /api/v1/upload could create a second job for the same audio — transcribed twice, billed twice. Those two calls now retry only when the request provably never landed.

· SDKs 0.2.0