← all repositories
fossasia/susi_skill_data

A Git repo that pulls itself every 60 seconds

SUSI.AI's skill database is literally just Git plus a cron job, and that's somehow both the hack and the architecture.

1.5k stars Data Tooling
susi_skill_data
Velocity · 7d
+0.5
★ / day
Trend
steady
star history

What it does This is the skill store for SUSI.AI, an open-source voice assistant. Skills live as files in a Git repository. The production server clones this repo alongside susi_server and runs git pull origin master every minute to stay current. Contributors write skills through a web CMS or by hand, following a custom skill language documented elsewhere.

The interesting bit The “deployment pipeline” is just Git’s own mechanics. No packages, no containers, no API versioning — the server yanks fresh code from master like a kid checking the mailbox. The README calls this “a temporary solution for a wiki-like skill editing service,” which suggests the team knows it’s a hack but hasn’t replaced it yet.

Key highlights

  • All skills must be CC0-licensed, matching Wikidata’s public domain approach
  • Changes propagate to production within about a minute via automated pulls
  • Skill authoring happens through https://susi.ai or by learning the custom skill language
  • Requires co-installation with susi_server to function
  • Explicitly described as temporary infrastructure

Caveats

  • The “every minute” pull strategy means broken commits go live immediately; there’s no staging gate mentioned
  • The README’s “DO NOT PANIC!” reassurance about skill creation suggests the actual process may not be as frictionless as claimed

Verdict Worth a look if you’re building a low-friction content ecosystem and want to see how far you can stretch Git before building proper infrastructure. Skip it if you need rollback safety, review gates, or any deployment sophistication — this is deliberately primitive.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.