Configuring Hermes Model Routing Without Surprises

Model routing decides which provider Rocky uses for normal chat, coding, image work, and auxiliary tasks. Treat it as an operations policy, not a hidden magic layer.

Official provider docs: https://hermes-agent.nousresearch.com/docs/integrations/providers

What to define first

The default model for everyday work.
Premium models for difficult coding, creative, or client-facing tasks.
Cheap models for summaries, background reports, and low-risk cron jobs.
Provider-specific credentials and fallback behavior.
Which tools or jobs are allowed to use media generation.

Routing checklist

Use the Hermes model picker or configuration commands documented by Hermes.
Keep API keys in the configured secret store or environment path, never in prompts.
Test a simple request after adding a provider.
Record model policies in project docs or skills when they are durable.
Re-check routing after profile changes because profiles can isolate configuration.

Pitfalls

Assuming one provider supports every modality.
Falling back from a preferred image provider without user approval.
Spending premium model credits on routine cron summaries.
Using a cheap model for high-stakes public/client copy without review.
Reporting a provider is configured without a live test.

Verification steps

Run hermes model or hermes config from a trusted terminal when changing settings.
Confirm the target provider appears in hermes status or a simple chat test.
For image workflows, verify the actual backend that generated the file.
For cron jobs, check the job’s configured provider/model override when applicable.