How to build a pilot dashboard

Convert Harbourmaster telemetry logfmt into a read-only pilot report.

Capture the log

The Action writes Effect logfmt to the GitHub job log. Copy the timestamp=... level=... message=... lines into a local telemetry.log file.

Do not use the harbourmaster-opencode-events.jsonl artifact as dashboard input. That file is the raw OpenCode event stream.

For a local CLI run, redirect standard output to telemetry.log.

See the telemetry reference for the accepted Effect logfmt fields. See How to diagnose a Harbourmaster review failure when a run is missing from the report.

Build a Markdown report

Run the dashboard command from the EOS repository root:

pnpm --filter @eos/harbourmaster dashboard:build -- \
  --input telemetry.log \
  --output dashboard.md

Open dashboard.md. The report groups events by review run and summarizes run status, findings, latency, retries, model usage, estimated cost, and links back to GitHub when the required identity fields are present.

Select another format

Use an .html or .json output extension:

pnpm --filter @eos/harbourmaster dashboard:build -- \
  --input telemetry.log \
  --output dashboard.html

Alternatively, pass --format markdown, --format html, or --format json to override extension inference.

Confirm the result

The command writes the output path and the run count to standard error. Open the output file and confirm it lists the expected runs.

If the command fails to parse a line, confirm the file contains Effect logfmt and not OpenCode JSONL.