# KolWrite API

Base URL: https://api.kolwrite.com

Use personal access tokens with either `Authorization: Bearer kol_...` or
`X-API-Key: kol_...`. Public API operations document their required scope in
the OpenAPI extension `x-required-scope`.

Core docs:

- Quickstart: /quickstart.md
- Authentication and scopes: /authentication.md
- Errors and rate limits: /errors.md
- OpenAPI: /openapi.json

Localized raw guides:

- Hebrew: /locales/he/quickstart.md, /locales/he/authentication.md, /locales/he/errors.md
- Modern Standard Arabic: /locales/ar/quickstart.md, /locales/ar/authentication.md, /locales/ar/errors.md
- Spain Spanish: /locales/es/quickstart.md, /locales/es/authentication.md, /locales/es/errors.md
- France French: /locales/fr/quickstart.md, /locales/fr/authentication.md, /locales/fr/errors.md
- Russian: /locales/ru/quickstart.md, /locales/ru/authentication.md, /locales/ru/errors.md
- Mainland Simplified Chinese: /locales/zh/quickstart.md, /locales/zh/authentication.md, /locales/zh/errors.md

Primary workflow:

1. Create an upload session.
2. Upload bytes to the returned storage URL.
3. Finalize the media.
4. Create a transcription or caption burn job with an `Idempotency-Key`.
5. Poll the job until it succeeds.
6. Export transcript formats or inspect durable artifacts.

Also:

- Speakers only: keep a transcript you already have (timed lines, for example
  a saved live session) word for word and add speakers with
  `"profile": "speakers-only"` and `options.reuse_transcript_segments`; only
  the diarization add-on is charged (`regular` 0.25, `pro` 1.00 credit per
  minute). See /openapi.json.
- Live transcription: `GET /api/v1/models` gives every transcription model a
  `live` block (where it runs, `request_model`, estimated credits per minute
  of speech).