Wanderer：漫遊美國國會圖書館分類圖譜的 AI
Wanderer: An AI That Roams the Library of Congress Map
Wanderer 是一個新推出的開源 AI 計劃，會透過 GitHub Actions 每四小時醒來一次，從美國國會圖書館分類中挑選主題。它會優先探索較少人注意的分類，選讀一篇 Wikipedia 文章，再以特定風格撰寫田野筆記；筆記以 Markdown 儲存。
Meet Wanderer
I’m proud to drop a new open‑source project: Wanderer. It’s an AI that lives on the internet, waking every four hours via GitHub Actions. It picks a subject from the Library of Congress classification—dark spots first—snags a strange real Wikipedia article, reads it, then writes a field note in a voice that mixes Carlin’s bite, Douglas Adams’ absurdity, and Tim Urban’s clarity.
How it works
-
wander.pyruns a single trip using only the Python standard library. - A cron workflow (
.github/workflows/wander.yml) is the heartbeat. - The brains are free cloud AIs (Groq, Gemini, OpenRouter) with keys stored as GitHub secrets.
- All notes land in
notes/as Markdown, tracked injournal.json. - A public page lives at https://2db2du-byte.github.io/wanderer/.
Running your own
Fork the repo, add GROQ_API_KEY (and optionally the others) as secrets, enable GitHub Pages on the /docs folder, and let it loose. The repo even includes an easy off switch: drop a file named PAUSED or disable the workflow.
The project is MIT‑licensed, so feel free to hack, remix, or just watch it grow.
Find more about me and my other AI experiments on my homepage: https://2db2du-byte.github.io
來源：DEV Community · dev.to