Troubleshooting failed generations

Generations fail sometimes. This page is the order you work through it. Most failures are a single page and clear on a regenerate; the rest come down to intake quality or a missing upstream stage.

Where a failure shows up

A failed page shows in two places:

• In the agent workspace, the deliverable grid marks the page with a Failed status badge.
• In Monitoring, a permanently failed run appears in the dead-letter table.

Work it in order

1. Regenerate the single page. From the agent workspace deliverable grid, use Regenerate on the failed page. Transient failures usually clear on the first retry. Re-open the preview and read the page once it comes back.

2. Check the intake. If the page keeps failing or comes back thin, the source is often the intake. Open the client's intake and confirm it is complete and the answers are sensible. A sparse or contradictory profile gives the agent little to work with. See intake__reviewing-the-profile.html.

3. If the whole run failed, check the upstream stage. Runs are gated: Sam needs a completed intake, Alex needs Sam published, Riley needs Alex published. If a full run will not start or fails immediately, confirm the previous stage is actually published, not just generated. The prelaunch modal checks readiness, so read what it tells you.

4. Escalate persistent failures. If a page or run fails repeatedly after a clean intake and a published upstream stage, escalate it rather than republishing a broken pack. A repeatable failure is a platform issue, not a content one.

Related

• workshop-pack__regenerating.html
• intake__reviewing-the-profile.html
• operations__monitoring-runs.html