What it is
A browser UI for log triage with a decision engine.
What it does: Chunks log events, asks Laya or Jev for a category and confidence, and lets operators filter uncertain rows and record corrections.
Why it belongs in the showcase: An ops-facing log classifier that can swap a local decision model for hosted Jev without changing the review workflow.
Caveats: No accuracy numbers published; Jev is optional.
Evidence
Public MIT repo (2 GitHub stars as of intake) with Docker + Laya GPU quick start, Korean README, architecture diagram and a YouTube demo. Jev is a selectable paid engine via .env (model jev-1.13.0); Laya is the default self-hosted path with no automatic paid fallback. Supports batch scan, follow mode, reviewer corrections and JSON export.
Caveats
- No published classification accuracy benchmark in the README at intake
- Default path is self-hosted Laya; Jev is optional and paid
- README states results are not an accuracy guarantee; in-memory jobs are lost on restart unless exported
Unofficial case page. Stars / list position observed for heat; do not treat heat as a quality SLA. Review the upstream README before you depend on it.