Fastest pattern
Start where the problem livesThat is usually faster than opening support immediately.
Support
Troubleshooting
Troubleshooting in opheli.ai works best when you start with the surface that owns the problem. Provider issues belong in Provider Vault. Execution issues belong in Runs and Replay. Context issues belong in Context Vault. Cost posture belongs in Billing and Provider Vault.
Guide summary
Use this page to understand the surface before you act inside it.
Use the right recovery path when provider access, runs, context, artifacts, or limits block progress.
Guide type
SupportThis guide reflects the current product workflow and surface ownership.
Sections
5Summary first, then steps, mistakes, and recovery notes.
Related guides
8Written against the current product structure and core execution workflow.
Best escalation rule
Report with contextUse the related run, mission, task, or artifact whenever possible.
Related action
Choose a specific troubleshooting guideSpecific recovery guides are more useful than a generic failure page.
Guide section
How to troubleshoot cleanly
The goal is not to click every button. The goal is to identify whether the blocker belongs to provider access, runtime execution, artifact packaging, source material, or account limits.
- Provider access issue -> Provider Vault or provider-key-invalid guide
- Run error or failure -> Run detail, Replay, or run-failed guide
- Run still moving too long -> run-stuck guide
- No deliverable -> no-artifact-generated guide
- Source material confusion -> context-not-used guide
- Cost or entitlement block -> budget-or-plan-limit guide
Guide section
When to use Help, Replay, or Support
Use Help Center when...
You need to understand product behavior or feature relationships.
Use Mission Replay when...
You need to see what actually happened during a run.
Use Contact Support when...
You still have a blocker after checking the owning surface and the relevant troubleshooting guide.
Guide section
Common diagnostic questions
What changed recently?
Provider, context, operator, blueprint, or plan posture changes often explain breakage.
Is the run actually attached to the right mission or task?
Wrong-object launches create confusing output and misleading support reports.
Did review or approval intentionally stop the run?
A blocked run is not always a broken run.
Did an artifact fail to form because the run never completed?
No artifact is often a downstream symptom, not the root cause.
Related action
Open the specific guide that matches the dominant symptom.
Guide section
Escalation posture
When you escalate, keep the report concrete and attach the most relevant object. “Run #123 failed during review after provider key rotation” is far better than “something broke”.
Tip
Context beats length
A short report with the right run or artifact attached is usually more useful than a long vague paragraph.
Guide section
Choose a guide
Provider key invalid
Credential or model access problem.
Run failed
A run hit a failure state and needs diagnosis or retry posture.
Run stuck
The run appears to be queued or processing too long.
No Artifact generated
Execution finished badly or final packaging never landed.
Context not used
Source material was not attached, not prioritized, or not scoped well.
Budget or plan limit reached
Execution is blocked by entitlement or spend posture.
Can’t find something
The issue is information architecture, navigation, or admin separation.