Skip to main content

Audio Transcribe Tool

audio_transcribe is a native tool that can transcribe audio media.

When To Use It​

Use audio_transcribe when a bot or workflow needs this capability inside the current workspace. Prefer this tool over ad hoc text handling when the task depends on workspace artifacts, blobs, links, events, channels, forms, media, or operation state.

Inputs And Outputs​

  • Tool name: audio_transcribe.
  • Input contract: AudioTranscribeArgs with fields artifact_id, slot, output_language.
  • Output contract: AudioTranscribeOutput with fields state, provider, model, content_type, task, run_id, request_id, provider_status, result_url, source_language, output_language, duration_seconds, transcript_artifact_id, text, note.

The caller should provide exact workspace, artifact, blob, channel, or operation references from the current context whenever the contract requires them.

Example Task​

A user can ask the selected workspace bot to run this kind of tool-backed work when the needed source context is visible or named. The bot should return a concise result and point to any artifact, blob, event, message, or operation output that was created or inspected.

Limits And Review​

  • Confirm the bot has access to the workspace and relevant artifacts.
  • Check required references before retrying a failed call.
  • Keep results bounded enough for channel delivery and operation review.
  • Review any durable workspace changes this tool creates before treating them as final.

Media To Add​

  • Table: example audio_transcribe input fields, output fields, and review action. It helps developers and admins understand the contract without reading raw schemas. Source: a sample operation that uses this tool.