Documentation

Run Lifecycle & Troubleshooting

Every prompt execution produces a run. Runs store the raw AI response, analysis output, citations, and diagnostic metadata that help you understand what happened behind the scenes.

Statuses

Status Meaning
Queued Run is scheduled and will start shortly
Running Chatobserver is currently collecting answers
Succeeded Responses were captured and processed
Failed The run ended with an error

Analysis status

After a successful run, Chatobserver executes enrichment jobs (ranking, sentiment, etc.). Analysis statuses include:

  • Pending – Enrichment job scheduled
  • Complete – Structured data ready
  • Failed – Enrichment could not finish; raw markdown is still available

Inspecting a run

  1. Open the Runs tab.
  2. Click any row to slide open the Run details sheet.
  3. Review:
    • Rendered markdown – The AI response as delivered to end users.
    • Sources – Normalized citations, including title, domain, and excerpt.
    • Timeline – When the run was queued, started, completed, and synced.
    • Errors – Any message returned while processing the run.

Use the “Copy prompt” action to recreate the run’s prompt in a new tab for testing.

Common failure reasons

  • Platform throttling – The AI provider rejected the request. Retry after the cooling-off period.
  • Expired credentials – Update any third-party credentials you use to fetch AI answers on behalf of Chatobserver.
  • Credit hold – The workspace exhausted its monthly credit allotment. Visit Plan & usage to add credits.
  • Prompt removed – If a prompt is deleted while a run is in progress, the run will fail.

Retrying runs

  • Click Retry run inside the details sheet. This creates a new run with the same prompt configuration.
  • If the issue persists, confirm platform credentials and credit balance before retrying again.

Exporting data

  • Use the table export action to download the current page of runs as CSV. The export includes prompt label, run status, start/completion timestamps, and trigger.
  • For programmatic access, the External API exposes a GET /prompt-runs endpoint with the same fields.

Best practices

  • Keep an eye on the analysis status pill to confirm downstream jobs finish.
  • Leverage groups to triage failures by campaign—filter the Runs tab via the Prompts tab before drilling in.
  • Document your retry policy so teammates know when it’s safe to rerun without inflating credit usage unexpectedly.