Home Assistant Podcast

Home Assistant AI Deep Dive: LLM Vision, Copilot, OpenClaw & More

28 snips
Mar 10, 2026
They dig into using local and private LLM models to power Home Assistant automation and routing. They talk about using Copilot to generate YAML and HACS add‑ons like AI Agent and AI Suggester. Vision models on doorbells and cameras for alerts and token-saving heuristics get practical coverage. They also wrestle with OpenClaw integration, OAuth quirks, and when agent-style automation actually helps.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Private Cloud Models For Cheaper AI Routing

  • Rohan is building a private AI foundation that routes requests to preferred models to keep costs down and control data flow.
  • He uses Ollama Cloud plus LiteLLM and plans fallbacks to paid models to avoid per-token API charges.
ADVICE

Let Copilot Generate YAML Automations

  • Use GitHub Copilot in VS Code to generate and complete YAML automations by writing clear IDs, names, and descriptions first.
  • Phil writes detailed descriptions so Copilot can prefill triggers, conditions, and entity IDs for full automations.
ADVICE

Add AI Agent In HACS For In-UI Chat

  • Install the HACS AI Agent to get a chat UI inside Home Assistant that queries your instance and returns YAML you can copy.
  • It supports multiple providers but requires separate API keys specific to the integration.
Get the Snipd Podcast app to discover more snips from this episode
Get the app