Status

Status

Live status is published at the public endpoint /api/status — a heartbeat the engine writes once per scheduled data run (freshness, active-listing count, last run, MCP health). This page fetches it and renders it below. If the fetch fails, or before the first probe lands, the manual check further down still works.

Loading live status… If this doesn't update in a moment, the manual check below always works.

Live engine health (/api/health)

The heartbeat above is a curated snapshot written once per scheduled run. This panel calls /api/health directly, live, on page load — a deeper (and slightly heavier — it reads D1 directly) probe of the same underlying state: engine version, last run, and active-listing count, straight from the database rather than from the once-per-run snapshot above. If it fails to load, the heartbeat above and the manual commands further down still work.

Loading live engine health… If this doesn't update in a moment, the heartbeat above or the manual check below still work.

What is probed

• Listings freshness: did the daily 13:00 UTC run land, and how many active listings resulted (listings_count, last_run).
• /mcp: the tool catalog is reachable (mcp_ok).
• Model version: the engine build stamped on the run (model_version).
• Per-feed last-refresh: each source in data sources is checked against its freshness target during a run.
• Live engine health panel above: the same version/last-run/active-listings signals, read directly from /api/health instead of the once-per-run heartbeat — useful right after a manual rescore, when the heartbeat above hasn’t caught up yet.

Checking manually

curl https://api.plantedmoney.ai/api/status   # public heartbeat (this page)
curl https://api.plantedmoney.ai/api/health   # deeper bindings + last run

What to read in /api/health:

Field	Meaning
ok, d1, kv	service and storage bindings up
last_run	most recent listings run: kind, finished timestamp, found, new_count. Older than ~24 h = the daily run failed
active_listings	current active inventory
stale_model_rows	active rows still scored by an older model_version — nonzero right after a release, draining via rescore

/api/metrics/scorecard also returns last_data_run alongside the published accuracy numbers. If something looks stuck for more than a day, open a GitHub issue (repository link to be published).