Dashboard (eq dashboard)¶
Launches the FastHTML dashboard, the primary UI for browsing saved runs. It is the only report surface receiving new features; eq redteam ui and eq sim ui are the retired Streamlit viewers, still registered but no longer documented.
For what the dashboard shows once it is open — the run index, the red-team and simulation walkthroughs, and applying recommendations to an agent — see Dashboard. This page covers the command only.
| Flag / Argument | Type / Default | Description |
|---|---|---|
PATHS | Path (repeatable) | Directories to scan for reports, or a single report file. Omit to scan both default stores. |
--host | str / 127.0.0.1 | Host to bind the server to. |
--port | int / 8080 | Port for the server. |
What gets scanned¶
The argument is what decides which runs appear, and the three forms behave differently:
- No path — scans both default stores together:
.evaluatorq/runs/(red team) and.evaluatorq/sim-runs/(simulation). This is the usual invocation. - One or more directories — scans exactly those, together. Useful for putting simulation runs from one repo next to red-team runs from another.
- A file — scans that file's parent directory, so sibling reports still resolve, and prints the direct URL for that one report. You land on the report rather than the index.
# every saved run, both stores
eq dashboard
# simulation runs only
eq dashboard .evaluatorq/sim-runs
# two stores from different checkouts, side by side
eq dashboard ~/work/api-agent/.evaluatorq/runs ~/work/support-bot/.evaluatorq/sim-runs
# open one report directly
eq dashboard .evaluatorq/runs/red-team-2026-08-18T09-14-02.json
Binding¶
--host defaults to 127.0.0.1, so the dashboard is reachable only from the machine running it. Binding elsewhere exposes every saved run — transcripts, system prompts, and judge verdicts — to anyone who can reach the port. There is no authentication layer. Put it behind one before binding to 0.0.0.0.