LLM Engine

Available models

API Keys

{% if api_keys %} {% for key in api_keys %} {% endfor %}

Name	Source	Value
`{{ key.name }}`	{% if key.source == 'env' %} `{{ key.env_var }}` {% else %} encrypted {% endif %}	`{{ key.masked }}`	{% if key.source != 'env' %} {% else %} env override {% endif %}

{% else %}

No API keys configured. Add one below to enable cloud providers.

{% endif %}

Notifications

Push Notifications	{% if notify_enabled %} Enabled {% else %} Disabled {% endif %}	Toggle via `fw notify enable\|disable`
Alert Dispatcher	{% if dispatcher_available %} Found {% else %} Missing {% endif %}	skills-manager alert dispatcher

Notifications alert the human when the agent is blocked (Tier 0), tasks need review, or audits fail. Run fw notify setup for prerequisites.

How it works

Ollama — Local models. No API key needed. Models must be pulled first (ollama pull model).
OpenRouter — Cloud gateway to 200+ models (Claude, GPT, Gemini, Llama, etc). Requires an API key from openrouter.ai/keys.
Key storage — Keys are encrypted with Fernet (machine-bound via /etc/machine-id). Environment variables override stored keys.
Hot-switching — Changing the provider takes effect immediately, no restart needed.

Settings