agent-model-healer

Automatically switch scheduled agents to a fallback model by running a self-healing watchdog that probes model health, detects failures like 402 or 429 errors, and moves affected agents to the next healthy model. The healer restores original models when they recover.

Self-healing model availability management is a watchdog process that continuously monitors scheduled agents, probes for unhealthy model states, and dynamically switches agents to healthy fallback models to prevent downtime.

Monitor model health by applying probes via /zo/ask to detect specific failures like 402 credits, 429 rate limits, 503 unavailable errors, and timeouts across your scheduled agents, triggering automatic failover when unhealthy states are found.

Yes, you can restore agents to their original models automatically. The healer preserves a rollback path during failover and switches agents back to their preferred model once health checks confirm the original model has recovered.

Yes, the self-healing failover process maintains a per-agent switch history and state for auditing. It tracks all model switch actions and sends email notifications when failover or restoration events occur.

Yes, you need deterministic MCP calls to orchestrate end-to-end model failover across scheduled agents. The healer performs agent management via these direct calls to switch unhealthy agents and restore originals during recovery.