Where IOTA Watch data comes from and how often it refreshes
Official API sources, refresh intervals, sample timestamps and stale data: how to tell whether an IOTA Train at Home reading is current.
Plain text versionIOTA Watch reads the public Macrocosmos mainnet API. Device status, rewards and training statistics have separate update clocks; refreshing one does not make the others current.
Which sources feed each view?
- Device discovery: active runs and the miner roster for every active run
- Network view: runs, run occupancy, miner rosters and run progress
- Rewards: entitlement totals and reward history, fetched independently
- Training history: metrics, throughput and cumulative tokens for the selected run and period
The API base is iota-web.api.macrocosmos.ai/mainnet. These are public readings; IOTA Watch cannot read local device logs or control the official training app.
How often does the page update?
While the page is visible, device status normally checks every 30 seconds and rewards every two minutes. Network miner rosters and progress check about once a minute. Server caches last 60 seconds for miner rosters and progress, 120 seconds for occupancy, and five minutes for runs, rewards and historical metrics. Repeated page checks may reuse a valid cached response.
Manual refresh has a 15-second cooldown. It can request newer data, but still respects official rate limits and retry delays. Background polling pauses while the page is hidden.
Fetch time versus sample time
Fetch time records the last successful read of that source. Sample time is when the official API recorded a training statistic. A successful read can contain an old sample; that alone does not show a machine is offline.
What happens when a source fails?
Previous successful values remain visible with a warning and their original fetch time. Missing data stays unknown. Reward totals and reward history can fail separately, so one usable field does not imply that every reward field is current.
A device refresh interrupted for more than five minutes is a data-read warning. See refresh interrupted. A partial network view reports the number of rosters successfully read.