I’ve been thinking of an AI assistant in Everdo for certain tasks like automatically tagging, estimating, time, prioritizing items based on horizons of focus, etc.
A rough idea of what I’m thinking about would be to have a user’s personal info on Everdo vectorized (which keeps getting updated in the background), langchain, and then use that to interact with an AI agent so it can go through a user’s personal info without running into token limits and find (or modify) the precise data relevant to the task the user is asking for.
One project I’m currently using to “chat” with my personal data is called Khoj. It has a plugin to interact with my Obsidian vault.
The workflow to interact with Obsidian is: pip install khoj-assistant → server runs in the background (127.0.0.1:42110) → install community plugin in Obsidian → insert API key for GPT or Llama (for the privacy conscious who prefer a local AI) → Obsidian vault gets indexed → user can then chat with their data.
For Everdo, I was thinking of something like:
- Click on a button available on the Inbox section to auto-suggest → pop-up appears with auto-suggestions for me to approve/deny/edit. These suggestions are based on my Horizons of Focus, tag suggestions according to current tags or suggest new ones, time suggestions, energy suggestions.
- Similar button for the other Action sections.
- I can interact with my Horizons of Focus and improve through chatting with the AI agent, which can also suggest changes based on my list of past/present/future Everdo data.
- Overall Everdo “analysis”, where the AI agent goes through all my Everdo data and suggests ways to improve.
I understand this project would significantly increase the size of an Everdo installation, but it would be awesome.