Script you can run on your computer in the background, to export your Safari history every hour and upload it to a D1 database. The goal with this is to be able to have browsing history as additional context for your AI.

Other related repos I made:

• https://github.com/janwilmake/ip-camera-to-s3-macos
• https://github.com/janwilmake/yaptabber
• https://github.com/janwilmake/efficient-recorder

How to use:

• Ensure you have node and bun

• install pm2 with npm install -g pm2 To look at the data locally:

• Go to privacy & security -> Full Disk Access, and allow full disk access to Terminal

• Run sudo bun run export-history-formatted.ts (or other scripts); (need administrator access to export browsing history)

To run this in the background and upload it every hour:

• run npm run create to create your db and copy the result in wrangler.toml
• run npx wrangler deploy and put the domain in .env as DOMAIN=Your Domain Here
• run pm2 start job.js --name "safari-history-tracker"
• run pm2 save to ensure it reruns at restart
• run pm2 logs safari-history-tracker to view if it worked

Ideas for making this useful

• Try scraping every page stored and augment the with its markdown and raw text (e.g. with https://github.com/janwilmake/llmtext.reader)
• Analyse each url+text into a summary of a sentence (with your desired LLM API)
• Create an endpoint that shows domains visited most, average page duration, total duration, duration percentage, etc.
• Create an endpoint to get context between two times (?from=...&until=...) that allows us to gather the context for cros-processing this with another context (e.g. audio or video)
• Analyse URL structure of domains visited (uncover patterns) using LLMs

Wishlist

• Run workerd locally (https://github.com/cloudflare/workerd) so this doesn't need to be hosted on cloudflare (Full privacy)