AudioSumma

Record your local audio and summarize it with whisper.cpp and llama.cpp! Open source, local on-prem transcription and summarization!

Installation

pip install -r requirements.txt

Configuration

Copy sample.env to .env and point your endpoint URLs for a working llama.cpp and whisper.cpp running in server/api mode.

llama.cpp and whisper.cpp

These need to be running in server mode somewhere on your local machine or on your network. Make sure the PROMPT_FORMAT in your .env file matches exactly to what the LLM model expects.

Running

Run either meetings.bat or meetings.sh to start app.

Usage

Hit record to record your global audio, hit stop to save the wav file. Hit transcribe to transcribe all wav files collected into a single summary markdown document (with date stamp). Hit the Clean button to remove old wav files and transcripts.