Skip to content

Logs

Logs provides operational evidence. Use it to investigate services and subsystems after the normal product surface has identified a symptom.

Evidence sources

  • Application contains persisted Python application records with DEBUG, INFO, WARNING, ERROR, or CRITICAL severity.
  • Spine activity contains durable Chat, model, tool, approval, user, and Flow events. It does not invent a severity where none was recorded.

Search message or payload content, filter by subsystem and time, and use severity filters for Application logs. Routine application traffic is hidden by default but can be revealed. Polling runs every ten seconds and can be paused or manually refreshed.

Expand a row for timestamps, source identity, correlations, exception detail, structured fields, and the raw API record. Export filtered CSV exports a bounded newest subset and reports when more matching records exist.

Logs are not Executions

Use Executions as the authoritative normal surface for Flow outcome, stages, terminal output, provenance, and approvals. A log line without exact run linkage cannot prove which Flow snapshot ran or that it succeeded.