Skip to main content

LLM Models

Manages LLM model settings and connection status.

📄️ Check LLM Tokenizer Availability

Report whether the agent could load ``tokenizer`` from the hub. Advice only, stored nowhere — the operator decides. Probed server-side because the badge must reflect what the AGENT pod can reach: the hub is ``HF_ENDPOINT``, the cluster's keeper mirror the browser cannot see. The agent's own load failure is permanent for the pod's lifetime, so catching a typo, a gated repo, or a not-yet-seeded keeper here — while the form is still open — is the whole value. The id is format-checked with the ``LLMModel.tokenizer`` field rules before anything goes on the wire (every path segment starts alphanumeric, so no traversal), and the probed host comes from settings, never from the caller.