Skip to content

README.md

Latest commit

 

History

History
47 lines (32 loc) · 1.28 KB

README.md

File metadata and controls

47 lines (32 loc) · 1.28 KB

OpenHands Eval Monitor

A lightweight dashboard for monitoring OpenHands evaluation runs. Deployed on Vercel as a thin JavaScript client that reads public metadata from results.eval.all-hands.dev.

Features

  • Date-based navigation — browse evaluation runs by date (UTC)
  • Run list — see all runs for a given day, grouped by benchmark
  • Run detail — view metadata, parameters, and pipeline progress for each run
  • Status tracking — see whether a run is pending, running inference, running evaluation, completed, or errored
  • Auto-refresh — quickly refresh run status

Data Source

The dashboard reads from publicly accessible metadata files:

  • https://results.eval.all-hands.dev/metadata/YYYY-MM-DD.txt — list of runs for a date
  • https://results.eval.all-hands.dev/{run}/metadata/{file}.json — individual run metadata

Development

cd frontend
npm install
npm run dev

The dev server runs on port 12001 with a proxy to results.eval.all-hands.dev to avoid CORS issues.

Build

cd frontend
npm run build

Deployment

Deployed automatically via Vercel. The vercel.json configures:

  • Build command and output directory
  • API rewrites to proxy requests to the data bucket

Tech Stack

  • React + TypeScript
  • Vite
  • Tailwind CSS