Troubleshooting
An upload produced no candidates
Section titled “An upload produced no candidates”- Check the job in the Jobs panel — it may still be running, or it may have a warning.
- Scanned or image-only PDFs extract poorly; a text-based version works far better.
- A file that’s mostly boilerplate or that doesn’t describe the architecture won’t produce much — that’s expected, not a failure.
- The Jobs panel distinguishes a completed job that found nothing from a job that errored — check which one you’re looking at.
A phase won’t advance
Section titled “A phase won’t advance”The gate tells you what’s missing. Common causes:
- Unreviewed candidates in the queue (Phase 1).
- An open conflict record — resolve it before the phase can close.
- A critical gap without a complete closure chain (Phase 3).
- Missing sign-offs (Phase 5 — the dual sign-off is the hard gate; the other closure-readiness checks are informational).
The gate panel for the phase lists exactly which condition isn’t met.
A candidate won’t accept — “name collision”
Section titled “A candidate won’t accept — “name collision””Another element with that name is already approved. Merge the candidate into the existing element, rename one of them, or reject the candidate. The message names the existing element and who approved it.
A deliverable generated with an “incomplete” annotation
Section titled “A deliverable generated with an “incomplete” annotation”The underlying data was less than complete when you generated it, and you chose to proceed. The annotation is permanent for that generated file — regenerate after filling the data gaps to get a clean one.
An export failed validation
Section titled “An export failed validation”ArchiMate exports are schema-validated (and SHACL-checked) before you get the file. A validation failure usually means the model has a structure the target schema doesn’t allow — the error lists which elements or relationships are the problem, so fix those in the model and re-export.
The AI features aren’t responding
Section titled “The AI features aren’t responding”If the deployment has no language-model key configured, eavantage runs in manual mode — every AI feature falls back to human input. If a key is set and AI features still fail, it’s a provider or connectivity issue for whoever runs the deployment to check (Admin → LLM Usage will show whether calls are getting through).
A score is locked and I need to change it
Section titled “A score is locked and I need to change it”Submit an unlock request; a platform administrator grants it (Admin → Score Unlock Requests). Same pattern for unlocking a gap.