🍫 Raspberry Pi
π

pi-a-day

AI-assisted community documentation for Raspberry Pi bootloaders & boot processes — because the Foundation's docs don't always keep up.

GitHub
What it does

Scrapes community sources

Pulls from Raspberry Pi forums, GitHub discussions, DeepWiki, and other community docs — each source capped at 6,000 characters, with Wayback Machine fallback if a URL goes down.

AI-generated report

Feeds the scraped text into MiniMax M2.5 via the HuggingFace inference router (max 4,096 tokens, temperature 0.3) to produce a structured Markdown report on Pi bootloader development.

Card-based live site

Builds a dark-themed card grid at pi-a-day.pages.dev — one card per Pi model, each linking to its full report. Currently covers Pi 5.

Report a missing Pi

Users can flag Pi models that haven't been documented yet. The pipeline picks them up and generates a report on the next run.

Auto-deploys

Commits the generated HTML to GitHub and Cloudflare Pages rebuilds automatically. No manual deploy steps.

State tracking

Keeps a state.json so each Pi model is only processed once unless you pass --force.

What the report covers
Usage
CLI python main.py [huggingface_token]
  --pi=NAME    Target a specific Pi model (default: pi5)
  --force      Regenerate even if already done
  --no-push    Build locally, don't push to GitHub
Stack
Python HTML HuggingFace API MiniMax M2.5 Cloudflare Pages GitHub Actions