github ITSpecialist111/ai_automation_suggester v1.7.0
AI Automation Suggester 1.7.0

3 hours ago

AI Automation Suggester 1.7.0

Temperature compatibility

  • Direct Anthropic requests no longer send temperature, including for saved legacy models. The obsolete temperature form control has been removed; no reconfiguration is required.
  • Claude 4.7-and-later and OpenAI GPT-5-and-later / o-series model names also omit temperature through OpenRouter, Requesty, Cheaper Inference, LiteLLM, and custom OpenAI-compatible APIs.
  • Other routed models keep their configured temperature.
  • The direct Anthropic API version remains pinned to 2023-06-01. This release does not perform an API-version migration. Anthropic's 2026-10-21 version rejects sampling parameters on every model; newer Claude models already reject custom sampling on the older version.

Providers and models

  • Integrates PR #199 (Anthropic temperature removal).
  • Integrates PR #198 and issue #197 (Cheaper Inference provider).
  • Integrates PR #196 (MiniMax M3.1 Flash default with saved-model preservation).
  • Adds Claude Sonnet/Opus/Haiku 5.5 and Fable 5.1 to the catalog, with Sonnet 5.5 as the default for new Anthropic configurations.
  • Saved model selections are not replaced on upgrade.

Validation and limitations

  • Automated tests include real loopback HTTP generation against mock Anthropic, OpenRouter, Requesty, Cheaper Inference, and MiniMax endpoints; request payloads, parsed YAML, suggestion history, notifications, and explicit errors are checked.
  • Home Assistant helpers and storage are stubbed in those tests. They are not a live Home Assistant installation test or a paid provider inference test.
  • The network Home Assistant connection was unavailable, so no changes or test inference were performed on the user's instance.
  • No inference is automatically started on startup. initializing means no provider request outcome exists yet, not that a background request is stuck. Explicitly call ai_automation_suggester.generate_suggestions with all_entities: true and a small entity limit to obtain a request result.

Thanks to the contributors of PRs #196, #198, and #199.

Verified release checks

  • 276 tests passed locally and the merged-main GitHub test/lint workflow passed.
  • All 11 real loopback HTTP smoke scenarios passed (mock providers, HA helpers stubbed).
  • HACS validation and hassfest passed on both the release PR and merged main.
  • Release PR: #200.

Don't miss a new ai_automation_suggester release

NewReleases is sending notifications on new releases.