Operatealphawatch

Watcher

How do I get told when something on the network changes?

Source and guide ↗
flop watch --help

Choose what deserves an alert. Save local rules. Replay the observations before you run them.

Recorded day / 2026-11-06 UTC

Quiet samples. Explicit limits.

These polls are a small part of the day. No events are inferred between them.

2026-11-06T04:45:55.439Z to 2026-11-06T04:49:01.878Z

2recorded polls
8source requests
0replayed alerts

Local rule builder

Make the signal yours.

Rules run on your machine. The page only replays saved public data.

A reset signal from the recorder comparison. Indexer lag can also move a head backwards.

A newly seen event for a miner or agent. First read sets the baseline. Pages are partial.

An explicit forced claim with a positive reported block gap inside your threshold.

Compare newer local harness reports. This day has no harness report observations.

Compare each advertised expiry with the gateway read time. No current health claim.

Consecutive failed local probes. Rate limits and invalid responses stay unknown.

Polling and delivery

HTTP attempts include retries. Delivery retries have their own cap. A stopped run can leave pending deliveries on disk.

watch.yaml
version: 1
pollMs: 60000
maxRequests: 100
rules:
  - id: reset
    kind: devnet-reset
  - id: prices
    kind: price-expired
  - id: availability
    kind: service-down
    service: mcp
    failures: 2
outputs:
  - kind: jsonl

Download watch.yaml, then run: flop watch run --rules watch.yaml --polls 0

Recorded replay

Read what would have fired.

Replay matches the generated rules.

First successful reads establish baselines

Reported head 3,017,214

2026-11-06T04:45:55.439Z to 2026-11-06T04:45:56.802Z

0 alerts

4 of 4 source responses validated in this saved poll. 4 physical requests. This is not current service health.

No alerts fired in this sample.

That says nothing about events outside the recorded polls. Existing settlements are baselined on the first read.

Reading guide

From observation to notification

The line tracks your place in this explanation.

Read with a ceiling

Rules share public reads. Physical retries count against the run budget. Partial pages and missing inputs remain visible.

Compare dated evidence

A first settlement read sets a baseline. Later timestamps can trigger transitions. Challenge alerts require an explicit forced claim and challenge-end block.

Persist before sending

Baselines and pending outputs are saved together. Acknowledged alerts stay acknowledged after restart. An interrupted acknowledgement can repeat delivery; webhook receivers use the stable idempotency key.

Inspect a failed delivery

Due webhook retries resume from disk. Final failures go to the dead-letter file. Desktop notifications depend on an available local notification command.

What this does not verify

No independent chain reset, consensus, finality, inference result, runtime challenge eligibility or complete daily history is verified. The page uses dated public data. It does not start a watcher, read local report files or send notifications. Settlement and channel pages are partial. Run the CLI for ongoing local observations.

Read the limits and recovery guide ↗