Find problems and collect diagnostics
When a connection fails, a task stalls, or a feature reports an error, locate the failing layer before changing settings. This guide narrows the problem and prepares useful support details. For restart steps, see Desktop service maintenance.
Start with the symptom
| Symptom | Check first | Next step |
|---|---|---|
| App will not start | Installer source, version, startup error | Read Update the app and retain the error |
| Window opens but several features fail | Background services, health diagnostics, technical details in Settings > Advanced | Address service readiness |
| Only one model fails | Connection, model ID, quota, request error | Check Model connections and availability |
| Only one plugin fails | Enabled state, account, resource permissions | Check Manage plugins and access |
| Task waits or errors with an unknown outcome | Progress, output files, destination system | Establish the outcome before retrying |
| Saved setting has no effect | Effective source, locks, read errors | Read Settings sources and locks |
A working window, a ready service, and a completed task are separate results. A health check observes one point in time; it does not configure ongoing alerts.
Preserve the error and run a minimal check
Note the time, full error, model or plugin, and recent settings changes. Reproduce with a small permitted sample while keeping other configuration unchanged. Change one identified factor at a time and compare results.
If one model fails while others work, inspect its account, quota, and model ID rather than resetting every connection. If a plugin's authentication fails while services are ready, repair authentication rather than repeatedly restarting.
Do not repeat sending, publishing, or payment to test a failure. Check the destination and follow Recover after a failed task.
Collect diagnostics
Open Settings > Advanced for service technical details and health diagnostics. Use Refresh, Open logs folder, and Copy diagnostics in service details. Retain records corresponding to the failure time. An old successful report does not establish current health; retain a report-read failure as an error.
When seeking support, include:
- App version, operating system, and failure time.
- Expected action and actual result.
- Minimal reproduction steps, model or plugin, and relevant task identifier.
- Changes or restarts already attempted and their individual outcomes.
- Relevant errors and redacted diagnostics.
Review Logs and sharing redaction first. Do not send keys, authorization headers, full environment files, or unreviewed personal data. For request-level failures, include identifiers from Request tracing.
Verify recovery
Repeat the original small task and inspect its actual output and location. If services are ready but the feature still fails, investigate that feature. If the error disappears without output, reconcile the task result. Desktop users do not need standalone Evolver CLI diagnostics; self-hosted runtimes use Nodes and runtime instances in Developers.