Monitoring

Monitoring for Bundata workflows lets you see run history, success and failure rates, and (where supported) alerts so you can keep the Vector Catalog and downstream systems healthy and fix failures quickly.

Run history

• Run list — In the Platform, open the workflow and view past runs: time, trigger, status (success, failed, running), and duration.
• Per-run details — Drill into a run to see which step succeeded or failed, error messages, and (for extraction steps) document-level status. Use this for source lineage and debugging.
• API — Use the jobs or workflow runs API to list runs and fetch status programmatically. See API reference.

Success and failure

• Success — Run completed; extraction and delivery steps finished. Smart bites are in the Vector Catalog or destination as configured.
• Failed — One or more steps failed. Error messages and codes indicate the cause (e.g. invalid schema, connector error, rate limit). See Error codes and Error handling.
• Partial success — In batch runs, some documents may succeed and others fail. Inspect per-document or per-step status to retry only failures.

Alerts and notifications

• Configure alerts — Where the product supports it, set up notifications (email, webhook, Slack) when a workflow run fails or when failure rate exceeds a threshold. Use the Platform workflow settings or API.
• Webhooks — Send run completion or failure events to your system for custom dashboards or incident response. See Webhooks.

Troubleshooting failures

1. Read the error — Check the run detail and step error message. Common causes: invalid credentials, schema mismatch, unsupported file type, size limit, rate limit.
2. Check connectors — Ensure source and destination connectors are configured and credentials are valid. See Integrations.
3. Check schema — If extraction fails, verify the schema matches the document type and that required fields are present. See Extraction best practices.
4. Retry — For transient errors (e.g. rate limit, network), retry the run or the failed step. Use backoff when calling the API. See Rate limits and Error handling.

Quality and freshness

• Catalog freshness — If workflows run on a schedule, monitor that they’re actually succeeding so the Vector Catalog stays up to date. Failed runs mean missing or stale smart bites for search and agents. See Vector Catalog: Lineage.
• Extraction confidence — Monitor extraction confidence over time; drops may indicate schema or source issues. See Extraction runs.

Next steps

• Triggers & scheduling — When workflows run.
• Error codes — Error reference.
• Webhooks — Event notifications.