README.md

Dashboard

This directory contains a lightweight streaming dashboard for uni_agent run directories such as /tmp/swebench_qwen3_coder.

Features

• Per-run status inferred from the existing uni_agent log flow:
  • queued: run.log exists, but no actual run markers yet
  • running: environment startup or interaction steps have begun
  • verify: reward or shutdown phase has started, but the final result file is not written yet
  • completed: interaction_result.json exists
• Parallel run cards so multiple samples can be watched together
• Incremental log streaming from run.log
• Recent file create, modify, and delete events inside each run directory
• Search, status filtering, and overall counters

Start the Dashboard

From the repository root, run:

python -m dashboard.server --log-dir /tmp/swebench_qwen3_coder --port 8765

By default, the server listens on:

http://0.0.0.0:8765

Open the Page

If you are running locally, open:

http://127.0.0.1:8765

Remote Server Access

If Cursor is connected to a remote server, this default bind mode is closer to python -m http.server, so a single command is usually enough for port forwarding or remote preview tools to detect the service:

python -m dashboard.server --log-dir /tmp/swebench_qwen3_coder --port 8765

If you still need manual forwarding, you can use:

ssh -L 8765:127.0.0.1:8765 <user>@<remote-host>

Then open:

http://127.0.0.1:8765

If you want to bind only to localhost instead, run:

python -m dashboard.server --host 127.0.0.1 --log-dir /tmp/swebench_qwen3_coder --port 8765

Notes

• The server only uses the Python standard library.
• It polls the filesystem incrementally and only reads the appended part of run.log.
• The page animates newly appended log text word-by-word so active runs feel live.