ELECTRON · LOCAL AI · KNOWLEDGE
Repo Notebook
A visual knowledge environment that collects and analyses GitHub repositories and presents them as connected information — with Harvest, local AI and smart suggestions.
RESULT
127 repositories managed and categorised by local AI — no cloud API, no subscription.
01 THE PROBLEM
Anyone who works a lot with open source collects hundreds of repositories: in browser tabs, stars, notes and loose folders. Knowledge gets lost there. Why did you save something? Does it still work? What is it connected to?
02 WHAT IT DOES
- Collecting without friction
- Paste a repo URL and it's saved with README, stars, forks, language and licence. Search, filter by category and status, and an A–Z index for large collections.
- Harvest
- Paste a GitHub account, a list of tool names or a page URL (blog, Medium, linktree). Harvest fetches the page, pulls out all the GitHub links and lets you choose what goes into the notebook. Nothing is saved until you tick it.
- Local AI
- Categories, summaries and suggestions come from a locally running language model. No data to the cloud, no API costs, works offline too.
- Notes & status
- A note per repo (‘why did you save this?’) and a status: to try, installed, liked, archived. That keeps the collection alive instead of a graveyard.
- Map & Quick start
- View repositories as a connected map, and clone or open them on GitHub in one click.
- MCP link
- Through the Model Context Protocol, AI assistants can search and read the notebook directly — the basis for the Obsidian plugin.
03 WHAT I LEARNED
Local AI is fast enough for this kind of work if you keep the tasks small: categorising, summarising, suggesting. The biggest gain wasn't in the model, but in the workflow around it — Harvest and the status buttons made the difference between ‘handy’ and ‘used daily’.
INTERESTED?
Let's talk.
Tell me what you think — or what you'd want to do with it.