Alerts
Last updated on May 21, 2026
TofuPilot Alerts watch your production line and fire when something matters. Trend alerts auto-resolve when the metric recovers; event alerts stay active until a teammate clears them.
Alert paradigms
Every alert type falls under one of two paradigms.
| Paradigm | Behavior |
|---|---|
| Trends | Continuous signals. Severity comes from the metric magnitude. Auto-resolves when the metric recovers. |
| Events | Discrete incidents. Fire once with a fixed severity. Stay active until manually resolved. |
Alert types
Trends
| Type | Description |
|---|---|
| Yield drop | Procedure pass rate dropped below the threshold over the configured window. |
| Volume drop | Run volume fell sharply versus the baseline period. |
| Measurement drift | A measurement rolling mean shifted by more than N sigma from baseline. |
| API failure | API requests are failing above the configured error rate. |
| Station silent | A station produced no run in the last 24 hours. |
Events
| Type | Description |
|---|---|
| Golden sample failed | A known-good reference unit returned FAIL. |
| Failing sample passed | A known-faulty reference unit returned PASS. |
| Unit retest threshold | A unit was retested more than the threshold within 24 hours. |
| Run failed | A production run finished with a non-PASS outcome. |
Severity levels
Every alert carries a level that drives notification routing.
| Level | When to use |
|---|---|
info | Informational signals; safe to batch or skim. |
warning | Quality regressions worth investigating soon. |
critical | Escape predictors or test-system failures; act now. |
Configuring a rule
Open Alerts → Configure to create a rule. A rule pins an alert type, a scope (procedure, station, batch, or part), a level, and a destination.
- Pick the alert type and paradigm. Trends ask for a threshold and a window; events fire on the matching record.
- Scope the rule to one or more procedures, stations, batches, or parts. Empty scope means the rule runs organization-wide.
- Set the severity level the firing alert should carry.
- Toggle the email destination. The owners of the scoped resources receive the email; you can opt in or out per rule.
Destinations
Alerts deliver to the owners of the scoped resource by email. The destination toggle on each rule lets owners opt out without disabling the rule for everyone else.
Additional destinations (Slack, webhooks, Linear, Discord) are available today through Workflows, which can subscribe to the same trigger events as alerts.
Inbox and resolution
The Alerts page is the inbox for the organization. Filter by level, type, status, or scope; open an alert to see the underlying chart, the units involved, and the runs that triggered the alert.
- Trends auto-resolve when the metric recovers within the configured window.
- Events resolve manually. Click Resolve on the detail page; the resolver and timestamp are stored on the alert.
- Resolved alerts stay searchable so you can audit historical incidents.
Scope and permissions
Alerts respect the same access model as the dashboard:
- Owners, Admins, and Developers configure rules.
- Viewers see alerts for the resources they have access to.
- Operators do not receive alert emails.
In the dashboard
Open Alerts from the sidebar. The inbox lists active alerts on top, with severity bars to make magnitudes visible at a glance. The Configure tab manages rules; the Subscriptions dropdown toggles per-rule destinations.
How is this guide?
AgentPreview
Run an autonomous root cause analysis on failing units with the TofuPilot Agent: hypothesis generation, evidence collection, and statistical scoring inside the chat.
ChatPreview
Query runs, units, parts, yield, throughput, and measurement distributions with TofuPilot Chat, an AI assistant grounded in your organization's test data.