Add jev-canvas - #22
gaborishka wants to merge 2 commits into
Conversation
|
Warning Review limit reachedNext included review available in 48 minutes. View limit detailsLimit details: You’ve used the included review currently available. You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Review configuration: ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Advanced Run ID: 📒 Files selected for processing (1)
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Advanced Run ID: 📒 Files selected for processing (1)
Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review. 📝 WalkthroughWalkthroughThe project catalog adds ChangesProject catalog
Priority: ⬇️ Low Estimated code review effort: 1 (Trivial) | ~5 minutes Change: Other Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches 💡 1⚔️ Resolve merge conflicts 💡
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Project: https://github.com/gaborishka/jev-canvas
Where Jev fits: one Decisions API request per partial transcript answers the typed questions (is this a command, is it complete, action, shape, colour, target, place, size) and deterministic code applies the thresholds and edits the canvas.
data/projects.jsonand rannpm run validate— the object is added, and I checked the file still parses as valid JSON, but I did not runnpm run validatelocally.README.mdorsite/index.html(CI regenerates them)Adds jev-canvas (I built and maintain it).
What it is: a browser demo where you draw on a tldraw canvas by speaking while pointing a finger at the screen (MediaPipe hand tracking + Web Speech API).
How it uses Jev: one Decisions API request per partial transcript with typed questions (noul: is_command, complete; choice: action, shape, colour, target, place, text span; score: size). Thresholds, pointing-word capture and execution are plain code. Questions: shared/questions.js, policy: shared/policy.js.
Evidence: scripts/probe.js runs 18 spoken commands (10 English, 8 Ukrainian) against the live API: 18/18 correct, 300–550 ms, about 1,900 input tokens per request. The README notes two findings: short Ukrainian imperatives scored is_command at 12% until the question said the speaker may use Ukrainian (87% after), and a visible finger alone pulls where=point up, so code only honours it when a pointing word was spoken.
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.Summary by CodeRabbit