Signals and errors

Spot error spikes above your logs and drill into the spans behind them.

The signals panel sits above the table in Logs. It charts your spans, errors, and behaviors over the time range you pick. Click a card, row, or cell, and the table filters to the spans behind it.

Each card compares the range with the period before it. A cell’s shade is its share of the spans in that column, and an outlined cell stands out from the rest of its row. The panel shows while Logs is grouped by Spans.

Drill into an error spike

Pick a time range

Go to Observability › Logs and pick 5m, 1h, 1d, 7d, or Custom in the panel. A spike shows as red in the Spans bars and a darker cell in the Errors row.

Open the errors view

Click the Errors card. The table filters to failed spans, and the panel lists “Incidents” first, then “Other error groups”. An error group collects failures with the same error type, provider, endpoint, and status code. Click the card again to go back to every span.

Zoom in on the spike

Drag across the columns and release, or double-click a column. Press Escape to step back, or click Reset zoom to see the whole range again.

List the spans in a cell

Click a cell. The table lists that row’s spans from that window. Click the same cell again to zoom into it.

Read the error group

Click the group’s name. The table filters to the group, and the row opens with its status code, fault, provider, and endpoint, then the sample span’s Error, Warning, and Recommendation. Errors by model splits the errors by model; switch it to Errors by customer to see who’s affected.

Open a span

Click a span in the table to see its full request and response. See Inspect spans and traces.

Triage an error group

An open error group has these actions:

• The resolution menu, which reads Unresolved until you triage the group: Resolve, Ignore, or Reopen. Resolved and ignored groups are dimmed.
• Copy fix prompt: copies a prompt for your coding agent. It asks the agent to inspect the failing spans through the Respan MCP server and propose the smallest safe fix, and it carries the fingerprint, status code, fault, time range, filters, counts, sample span ID, and error message.
• ⋯ menu: Create monitor, Copy fingerprint, Copy sample span ID, and Copy link. Create monitor makes a monitor for this group with your current filters, turned off, and opens it so you can finish it. See Monitors.

Under the details, Reach counts the customers and models affected. It’s hidden when you filter by anything other than environment.

Work an incident

An incident is a window where errors spike or drift above their usual rate. Incidents count every request in the environment, not only the spans that match your filters. An ongoing incident refreshes every minute.

Click an incident under “Incidents” to open it. It shows Status, When, Error rate, Peak, Failed, and the customers, providers, models, and endpoints it hit, followed by its error groups.

• Acknowledge tells your team someone owns it. Click Acknowledged to undo.
• Zoom to incident zooms the chart to the incident, with a little time on either side, and filters the table to its error groups. Zoom out goes back.
• The ⋯ menu has Add note, Copy incident ID, Copy link, and Copy summary. Everyone who can see the incident can read its notes, so keep secrets and customer content out of them.

Share a link

Copy link gives you a Logs URL that opens the error group or incident directly. You can also add these parameters to a Logs URL yourself: