AI Automation Suggester 1.5.0
This foundation release modernizes provider/model handling and makes suggestions easier to review, retain, and display in Home Assistant.
Highlights
- Current model defaults and model capability metadata for OpenAI GPT-5-style models, Azure OpenAI deployments, Claude Sonnet/Opus 4.x, Gemini 2.5/3 preview, Groq, Mistral, Perplexity, OpenRouter, and local/OpenAI-compatible providers.
- OpenAI Responses API support for GPT-5-style reasoning models with safer
max_output_tokensand reasoning effort handling. - Structured output support with JSON-first parsing, fenced YAML fallback, YAML validation warnings, and truncation warnings.
- Persistent suggestion history stored in Home Assistant storage.
- New backend endpoints for the included Lovelace card:
GET /api/ai_automation_suggester/suggestionsPOST /api/ai_automation_suggester/{accept|decline|dismiss}/{suggestion_id}
- New services:
ai_automation_suggester.clear_historyai_automation_suggester.update_suggestion
- Persistent custom prompt, exclusion filters, request timeout, and history retention options.
- Tests, changelog, issue templates, and CI scaffolding for safer future releases.
HACS Updates
HACS detects updates from GitHub releases. After the 1.5.0 tag and GitHub release are published, HACS users should see the update in Home Assistant and can upgrade from HACS. A Home Assistant restart is still required after updating a custom integration.
Notes For Upgraders
- Existing custom model names remain supported.
- Some stale defaults were refreshed for new configurations. Existing configured model values are preserved unless the user changes them in options.
- Suggestions are still review-only. The integration does not automatically create or modify automations.
- Cloud providers may receive entity names, states, attributes, areas, device context, and automation metadata. Use local providers for local-only processing.