Notehold is a little app that helps you take better notes using your voice! It's primarily a proof of concept, and was created in a few hours for a hackathon.
It will transcribe your voice, generate a summary/title, and categorize it with it's urgency and into a category based on the ones you provide it with (or it will just put it in a general category if you don't provide any).
- Clone the repo
- Install requirements.txt
- You will need both a groq key and a openrouter key, which you can get from the following links:
We use groq for the transcription and openrouter for the summarization. But under the hood we use litellm, so changing openrouter to another provider should be trivial.
- Create a .env file in the root directory with the following variables:
GROQ_API_KEY=your_groq_key
OPENROUTER_KEY=your_openrouter_key
- Run main.py!
This should probably be a native app or a mobile app! Integration with pre-existing pre-existing siri/alexa/google assistant would be cool too!
- More it clearer when app is recording
- Verifiy and deal with weird output from LLMs
// this could be tackled by making the prompts more specific/longer too!
- Check if the output of title summary is surrounded by quotes
- add markdown note output/integrations with other apps
- add keyboard shortcuts to record notes quickly
- add a way to manually add a note to a specific file
- make it more model agnostic (.env variable for model selection)