Skip to content

Changelog

Last updated: 2026-10-06

Notable changes to Kawa, most recent first.

For more info, (see the FAQ and ROADMAP.md for where things are headed).

For the Python SDK's own release history, see its CHANGELOG.md.

Unreleased

  • A failed job's error now includes next_step (what to do about it) and docs_url (its entry in the new Job Errors page).
  • GET /v1/transcriptions returns error on each failed row, so a list shows why a job failed without fetching it.
  • New error codes: invalid_audio, transcription_timeout and unsupported_language. Error messages are now written in plain words rather than passed on from the processing service.

2026-10-06

  • The Gradio playground has new Summary and Tagging tabs for pasted transcripts.
  • Your Jobs as a filterable list to see past jobs.
  • Its Status filter (formerly Outcome) adds Queued and Processing, and each row shows its status as an icon rather than in words.
  • Summary and tags rows name the transcription job_id they were made from, if relevant.

2026-09-16

  • New POST /v1/summarize and POST /v1/tag endpoints derive a summary or a fixed-taxonomy tag set from a succeeded transcription; see the API Reference.
  • GET /v1/transcriptions/{job_id} now returns type (transcription, summary, or tags) and, for a postprocessing job, source_job_id.
  • KawaClient (demo/kawa_client.py) gained create_summary() and create_tags(), plus the SummaryResult and TagsResult accessors on Job; see Using KawaClient.
  • The Gradio playground was updated to reflect the above API changes.

2026-07-01 — Launch

  • Post-conversation transcription API, hosted in Singapore
  • Open-source interactive playground and Python reference client (demo/)
  • Orchestration layer published under the MIT licence
  • Developer portal and API documentation at makimoto.ai