Workbooks troubleshooting

Symptom	Cause and fix
"No online synthetic agents available"	The web/API/probe node needs an ONLINE agent. Check the agent is connected and, for private targets, that it's the one selected in the node's Assigned Agents field.
APM Observe finds 0 spans	Traces hadn't arrived yet, or the service name is wrong. Increase Collection Delay / Max Retries and verify Service Name matches what the service reports (the config panel checks this live).
AI Evaluate: "judge returned no parseable result"	The judge model returned an unusable answer (often a weak/free model) or hit a quota. Make your default AI provider a capable model — Model Compare uses it as the one independent judge, so a candidate never grades itself.
AI Probe (Chatbot UI) captures an empty answer	End the steps with a Read step on the answer element; its text becomes the captured answer. For internal apps VerOps can't reach directly, use Chatbot UI mode with a private synthetic agent instead of Agent API mode.
DB Collector auth error (MySQL 8)	Create the database user with the `mysql_native_password` authentication plugin, and store the password as a secret.
A node keeps retrying	Retries apply to `error` (infra) only. A repeated `failed` is semantic (assertion/eval) — fix the condition or the system under test, not the retry.
Quality scores don't appear on the trend card	Scores appear per criterion on the executions page after an AI Evaluate node runs; ensure the node actually ran (not skipped by a branch).

Still stuck? Open the run timeline and read each node's outputs and the AI narration — the failing node's error message usually names the exact cause.