Gateway
Content passedLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
358 msScheduled history is unavailable. Showing bundled observations with their original dates.
Archive checked . Page cache: 5 minutes.
Source records ↗ · Download displayed JSONL
pnpm --filter @flop-lab/cli exec flop status probe --jsonContent checks passed
16 / 16 probesHTTP 200 plus the expected content.
Recorded rounds
2Each round attempts every listed service once.
Observed failure spans
0Sample boundaries, not continuous downtime.
01 / Public responses
Sample availability = passed probes / measured probes. Each probe has equal weight. No sample means no data.
Selected window: to .
Observers: local-build-c18. Sparklines use actual capture times and a separate vertical scale per service.
Latest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
358 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
329 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
699 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
800 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
536 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
518 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
502 msLatest HTTP 200
Sample availability
100.0%2 / 2 measured probes
Latest elapsed time
420 ms02 / Failure observations
A span joins consecutive failed samples from the same observer. Gaps over 4.5 hours break the span. A later passing sample marks observed recovery only.
No failed samples in this window. No incident duration can be inferred.
03 / Read the boundary
A probe makes one fixed GET or MCP initialize request. It has a total deadline and a bounded body. Redirects and retries cannot hide a failed first attempt.
Model ids, protocol information, page titles, health text and SDK metadata have explicit checks. MCP tools, inference, faucet funding and explorer queries are not exercised.
Only recorded samples enter availability. Scheduled history is designed for one round every three hours. Missed runs stay missing. These observations do not prove continuous uptime or chain health.