summscriber 0.4.0

Transcribe audio with Whisper (faster-whisper), with summarization (pysummarization, sumy, OpenAI) and short reply with OpenAI.

Summscriber

Transcribe audio with faster-whisper (Whisper), with summarization options (pysummarization, sumy, OpenAI) and short reply generation via OpenAI API.

Repository: github.com/pablogventura/summscriber

Installation

With pipx (recommended: isolated env, global command)

pipx install git+https://github.com/pablogventura/summscriber.git

To upgrade:

pipx upgrade summscriber

With pip (from project directory)

pip install .

Or in editable mode (development):

pip install -e .

Or from the repository:

pip install git+https://github.com/pablogventura/summscriber.git

Or from PyPI (once published):

pip install summscriber

Usage

After installing, the summscriber command is available:

summscriber FILE [options]

Examples:

summscriber recording.mp3
summscriber interview.ogg --reply
summscriber audio.wav --reply --json
summscriber audio.wav --no-summary

Main options

• FILE: audio file to transcribe (required).
• Summarization is on by default: OpenAI if configured, otherwise the shortest of pysummarization and sumy.
• --no-summary: do not generate a summary (transcription only, or with --reply/--json if you pass those).
• --summary-pysummarization / --summary-sumy / --summary-openai: also print a summary from a specific backend.
• --summary-sentences N: number of sentences in the summary (default 3).
• --reply: generate a short reply to the message with OpenAI.
• --json: output as JSON.

For summarization and reply with OpenAI, use a config file or environment variables. Configuration is read from the current directory, then from the user config dir (~/.config/summscriber/config.ini on Linux/macOS, %APPDATA%\summscriber\config.ini on Windows). See config.ini.example. Save your token and URL once (stored in the user config dir so it works from any directory):

summscriber --save-config --api-key YOUR_TOKEN --base-url https://...

Development

From the repo root without installing:

python -m summscriber FILE [options]

Publishing to PyPI

1. Install build tools: pip install build twine (or pip install ".[publish]").

2. Create a PyPI account and an API token at pypi.org/manage/account/token/.

3. From the project root run:

   ./publish.sh
   

   Or manually:

   rm -rf build dist *.egg-info
   python -m build
   twine upload dist/*
   

   When prompted, use username __token__ and password your PyPI token. Or set TWINE_USERNAME=__token__ and TWINE_PASSWORD=pypi-your-token.