Skip to content

Draft: feat: ASR integration for transcription

Bhaskar Battula requested to merge ASR-integration into develop

Overview

The feature enables the system to process user-uploaded or server-recorded audio files, generate transcripts using the ASR pipeline, and patch the generated transcript into the existing record data automatically.

What Does This MR Do?

Features Added

  • Added ASR integration flow into the CLI project
  • Supports:
    • User-uploaded audio files
    • Server-side recorded audio
  • Automatically transcribes audio content
  • Patches transcript data into existing records
  • Added logging and error handling for ASR workflows
  • Added automated tests for transcription and patching flows

Workflow

  1. Audio file is uploaded or existing audio record is fetched
  2. CLI triggers ASR processing
  3. Audio is sent to ASR service
  4. Transcript is generated
  5. Record is patched with transcript data
  6. Updated record is persisted

Technical Changes

Added

  • ASR service integration
  • Audio processing pipeline
  • Record patch/update handler
  • Transcription orchestration logic
  • Error handling and logging

Commands for Execution

Run ASR Processing

corpus-client asr <audio file>

Environment Variables

ASR_BASE_URL= <asr-service-url>

Test Coverage

Added Test Cases

  • Successful audio transcription
  • Record patch/update flow
  • Uploaded audio processing
  • Server-recorded audio processing
  • Invalid audio handling
  • ASR service failure handling
  • Invalid/missing record handling

Expected Outcome

After execution:

  • Audio files are transcribed successfully
  • Existing records are updated with transcript data
  • Errors are handled gracefully
  • Logs are generated for debugging and monitoring
  • Manual transcription effort is reduced

Checklist

  • ASR integration added
  • Record patch flow implemented
  • Error handling added
  • Logging added
  • Tests added
  • Coverage verified

Closes

Closes #54

Merge request reports