Models And Routes

A fresh HLVM configuration uses auto. The core resolves each request to one eligible concrete agent/model route based on discovered availability, authentication, capabilities, privacy, and the user's configuration.

Inspect Routes

hlvm model list
hlvm doctor

Use exact IDs from the live list because vendor catalogs and installed runtimes change.

Automatic Selection

hlvm model set auto

auto resolves before the turn begins. It does not silently change engines in the middle of a turn.

Subscription-backed Agents

hlvm model login <backend>
hlvm model usage <backend>

An installed, already authenticated agent is immediately eligible. If a supported installed backend needs authentication, HLVM can expose its focused login flow.

Provider API Keys

hlvm model key list
hlvm model key set <provider>
hlvm model key unset <provider>

Keys are stored in the platform credential store where available.

Pin An Exact Route

hlvm model set '<exact-ready-route-from-model-list>'

Use a fixed route for reproducibility, billing, privacy, or a specific native harness. An unavailable explicit route fails clearly instead of silently becoming another engine.

Optional Existing Ollama

HLVM can use an Ollama installation that the user already owns and runs. HLVM never bundles, installs, starts, or repairs Ollama and never forces a model download.

hlvm model pull ollama/qwen3:8b
hlvm model set ollama/qwen3:8b
hlvm model remove ollama/qwen3:8b

These commands are explicit BYO-Ollama operations. Ollama Cloud is treated as another discovered provider route, not a special onboarding requirement.