Switchyard Local-first Routing (Beta)

From its catalog listing
Routes Hermes between a local and a cloud model with Verification-Gated Routing (VGR): each turn is tried on the local model, checked, and escalated to the cloud model only when the check fails. Either tier can be any URL or a third-party model provider set up in Hermes; each provider's terms and conditions apply. Presets are selectable as switchyard/<preset> models, with a Desktop page and dashboard tab showing each decision live.
Good to know before installing
installs nemo-switchyard from PyPI; runs a loopback sidecar on 127.0.0.1:4110 that keeps running after Hermes quits (`hermes switchyard stop`), and Open in browser starts Hermes's web dashboard; reads every model request and response to route it; for a provider tier it takes that provider's credential (sign-in tokens included) from Hermes on each call and sends it only to that provider; stores a generated sidecar token and any tier API keys in Hermes's .env through Hermes's config API; setup makes switchyard/<preset> the default model and sends titles and smart approvals to the local model; copies its provider profile to <HERMES_HOME>/plugins/model-providers/switchyard/ on load; logs and routing records stay in <HERMES_HOME>/switchyard/; after removal it restores the replaced models, stops its sidecar and deletes its provider copy; it never rewrites saved conversations. No telemetry.
From its catalog listing.
How to install
- Install it. Hermes asks for any API keys it needs, then asks whether to turn it on.
hermes plugins install nvidia-switchyard - Said no? Turn it on later:
hermes plugins enable nvidia-switchyard - Restart Hermes so it loads the plugin.
Some plugins also need a sign-in or settings: see Full setup & settings. Memory plugins ask whether to become your memory provider instead. Steps follow the official plugins guide (opens in a new tab).
- Category
- AI models & providers
- Maintained by
- NVIDIA (Verified)
- Needs
- No API keys declared
- Adds hooks
pre_api_requestpost_api_requestpre_llm_calltransform_llm_outputtransform_api_error_classification- Works on
- linux, macos, windows
- Version
- 0.1.0
- In the catalog
- 2026-11-10
Using the Hermes desktop app? See where plugins live

Full setup & settings ↗Source code ↗
More in ai models & providers
All ai models & providers plugins
New Hermes Agent updates every day on X. Follow @HermesWatcher