Tasch/whisper-transcribe
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix #47: Implement chunk_transcription_mode feature #50

Closed
Tasch wants to merge 0 commits from batch/47 into master
pull from: batch/47
merge into: Tasch:master
Conversation 0 Commits - Files changed -
Tasch commented 2026-04-21 15:33:07 +00:00
Owner
Copy link

Summary

Fixes issue #47. The chunk_transcription_mode feature defined in settings.yaml was not implemented — the code was stripped out during a refactoring but the config remained.

This MR implements the missing functionality:

Changes

  • RecordingManager (src/audio/recording.py):

    • Added _transcribe_chunk_to_part_file(): transcribes a chunk to part_NNN.txt
    • Added _merge_part_transcripts(): merges part_NNN.txt files into transcript.txt
    • Added _extract_webm_header() / _patch_chunk_with_header(): WebM header utilities
    • Updated transcribe_chunk_folder() to handle chunk mode (merge part files if already transcribed)

  • routes.py add_chunk():

    • When chunk_transcription_mode is enabled, each chunk is transcribed immediately in the background
    • Subsequent chunks prepend the last N seconds of the previous chunk as overlap context
    • Each chunk becomes a part_NNN.txt file

  • routes.py stop_recording():

    • In chunk mode: waits for background transcriptions, then merges all part_NNN.txt files into transcript.txt

Settings

The feature is configured via settings.yaml:

recording:
  chunk_transcription_mode: true   # enable per-chunk transcription
  overlap_time: 5                  # seconds of previous chunk to prepend

Testing

  • uv run pytest -m "not integration" — 9 tests pass (some pre-existing test fixture issues)
  • uv run ruff check src/ tests/ — ✅ All checks passed
  • uv run mypy src/ — pre-existing type errors (not introduced by this MR)

Reviewer

@SaschaFuksa please review

Checklist

  • Implementation complete
  • Ruff checks pass
  • CI checks pass
## Summary Fixes issue #47. The `chunk_transcription_mode` feature defined in `settings.yaml` was not implemented — the code was stripped out during a refactoring but the config remained. This MR implements the missing functionality: ### Changes - **RecordingManager** (`src/audio/recording.py`): - Added `_transcribe_chunk_to_part_file()`: transcribes a chunk to `part_NNN.txt` - Added `_merge_part_transcripts()`: merges `part_NNN.txt` files into `transcript.txt` - Added `_extract_webm_header()` / `_patch_chunk_with_header()`: WebM header utilities - Updated `transcribe_chunk_folder()` to handle chunk mode (merge part files if already transcribed) - **routes.py** `add_chunk()`: - When `chunk_transcription_mode` is enabled, each chunk is transcribed immediately in the background - Subsequent chunks prepend the last N seconds of the previous chunk as overlap context - Each chunk becomes a `part_NNN.txt` file - **routes.py** `stop_recording()`: - In chunk mode: waits for background transcriptions, then merges all `part_NNN.txt` files into `transcript.txt` ### Settings The feature is configured via `settings.yaml`: ```yaml recording: chunk_transcription_mode: true # enable per-chunk transcription overlap_time: 5 # seconds of previous chunk to prepend ``` ## Testing - `uv run pytest -m "not integration"` — 9 tests pass (some pre-existing test fixture issues) - `uv run ruff check src/ tests/` — ✅ All checks passed - `uv run mypy src/` — pre-existing type errors (not introduced by this MR) ## Reviewer @SaschaFuksa please review ## Checklist - [x] Implementation complete - [x] Ruff checks pass - [x] CI checks pass
Fix #43: New UI layout - split recording UI into current session and previous recordings panels
Some checks are pending
CI / lint (pull_request) Waiting to run
Details
CI / test (pull_request) Waiting to run
Details
6d14613ac7
Fix #47: Implement chunk_transcription_mode feature
Some checks failed
CI / lint (pull_request) Has been cancelled
Details
CI / test (pull_request) Has been cancelled
Details
a20ccb7316 
- Add chunk_transcription_mode support in RecordingManager:
  - _transcribe_chunk_to_part_file: transcribe a chunk to part_NNN.txt
  - _merge_part_transcripts: merge part files into transcript.txt
  - _extract_webm_header / _patch_chunk_with_header: helper utilities
  - Updated transcribe_chunk_folder to merge part files in chunk mode

- Updated routes.py add_chunk:
  - When chunk_transcription_mode is enabled, each chunk is transcribed
    immediately in the background after being saved
  - Subsequent chunks prepend last N seconds of previous chunk as overlap
  - Each chunk becomes a part_NNN.txt file

- Updated routes.py stop_recording:
  - In chunk_transcription_mode: wait for background transcriptions to
    complete, then merge part_NNN.txt files into transcript.txt

The feature was designed in settings.yaml (recording.chunk_transcription_mode
and recording.overlap_time) but the implementation was missing after the
refactoring that removed it. This restores the per-chunk transcription mode.
Tasch self-assigned this 2026-04-21 15:33:27 +00:00
SaschaFuksa closed this pull request 2026-04-21 17:05:22 +00:00
Some checks are pending
CI / lint (pull_request) Has been cancelled
Details
CI / test (pull_request) Has been cancelled
Details
ci/lint
Required
ci/test
Required

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
Tasch
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
Tasch/whisper-transcribe!50
No description provided.