Janus v0.1.0

First tagged release of Janus — an AI safety gatekeeper between any LLM and Home Assistant: the model proposes, deterministic code decides. Dangerous operations always confirm; ambiguity asks; vague intent becomes a suggestion, never a silent action.

• HA conversation agent (install via HACS), CLI REPL, and a headless /v1 service
• Local (Ollama) or cloud (Anthropic) LLM backends
• Registry curation, disambiguation, intent inference, decision audit, dangerous-op PIN
• One-command full stack (deploy/compose) and a HAOS add-on (deploy/addon)