Confidant is a voice note transcription app, designed to handle multiple input & output directories, sorting by keyword, note aggregation, & conversion to PDF. It runs locally, with no data sent to third parties. Currently in an alpha state, and not yet ready for use by the general public.

Prerequisites

1. Python >=3.12
2. uv - a Python package installer (pip install uv).
3. An NVIDIA GPU with CUDA is recommended for hardware acceleration of the transcription process. The application will fall back to using the CPU if a compatible GPU is not found.

Setup

1. Clone this repo
2. Run uv sync && uv run pre-commit install from the root directory
3. Create a config.yaml in the root directory with the following structure (omit any sections not needed):

Note: If filters are set for the text-outputs in the config.yaml, the transcriber will check the first twelve words of a recording for filter keywords and route the resulting transcription files accordingly.

transcription-sets:
  - transcription-set:
      audio-inputs:
        - directory: Recordings/Unsorted
        - directory: Recordings/Unsorted 2021
        - directory: Recordings/Unsorted 2022
        - directory: Recordings/Unsorted 2023
        - directory: Recordings/Misc
      text-outputs:
        - directory: Transcription/Workouts
          filter:
            keyword: workouts
          retry:
            count: 20
            delay: 5
        - directory: Transcription/Workouts
          filter:
            keyword: work
          retry:
            count: 20
            delay: 5
        - directory: Transcription/Workouts
          filter:
            keyword: work-outs
          retry:
            count: 20
            delay: 5
        - directory: Transcription/Journal
          filter:
            keyword: journal
          retry:
            count: 20
            delay: 5
      deadletter:
        text-output:
          directory: Transcription/Unsorted
          retry:
            count: 20
            delay: 5
  - transcription-set:
      audio-inputs:
        - directory: Recordings/Notes
      text-outputs:
        - directory: Transcription/Notes
          retry:
            count: 20
            delay: 5
aggregations:
  - input:
      directory: Transcription/Notes
    output:
      filepath: Transcription/NoteAggregation.txt
      retry:
        count: 20
        delay: 5
pdf_conversions:
  - input:
      directory: "/mnt/documents/Text Files"
    output:
      directory: "/mnt/documents/PDFs"

Commmands

Execute: uv run python src/confidant/main.py Linting: uvx ruff check src Type Checking: uvx ty check src Tests: uv run pytest Precommit: uv run pre-commit run --all-files