github cartertemm/AI-content-describer v2026.11.17
Version 2026.11.17

43 minutes ago

What's new

Computer use and UI viewer (alpha)

  • Let Anthropic and OpenAI models control your computer to complete one-off tasks. Risky actions ask for your approval first.
  • Show the currently visible UI as a read-only HTML page you can browse with NVDA.
  • When a computer use session starts, the target window is maximized (#111).

New models

  • Anthropic: Claude 4.8 Opus, 5 Sonnet, 5 Opus, 5 Fable, 5.1 Fable, 5.2 Fable, 5.5 Haiku, 5.5 Sonnet, and 5.5 Opus (#110). Claude 5.2 Fable and the Claude 5.5 models support computer use through the new computer toolset.
  • OpenAI: GPT-6 Astra, GPT-6 Luna, GPT-6.1 Sol, and GPT-6.1 Terra, all with computer use support.
  • Google: Gemini 3.1 Flash-Lite, 3.5 Flash, 3.5 Flash-Lite, 3.6 Flash, 3.7 Flash, 3.8 Flash, and Gemini 4 Hydrogen, Argon, and Carbon (#109, #113).
  • xAI: Grok 4.5, 4.6, 4.7, and 4.20 (non-reasoning).
  • Mistral: Mistral Large 4 (#114).
  • Moonshot: Kimi K3 and Kimi K2.6.
  • Datalab: Chandra 2, a vision and OCR model.

Other features

  • New effort setting for Claude, OpenAI GPT-5.4 and later, and Grok 4.3 and later. It applies to descriptions and computer use.
  • New commands to copy the image to the clipboard (#106).
  • If the configuration file can't be read, you now get a dialog that offers to back it up and reset it.
  • Default max tokens raised to 8192 for reasoning models (GPT-5.4, GPT-5.5, Grok 4.3, Mistral Large 4) so reasoning doesn't use up the output budget.

Removed models

These models were retired or are about to be retired by their providers:

  • OpenAI: GPT-4 turbo, GPT-4o, GPT-4.1, GPT-4.1 mini, GPT-4.1 nano, GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 chat, GPT-5.4 nano, O3, O3 pro, O3 mini, and O4 mini.
  • Anthropic: Claude 4 Sonnet, 4 Opus, 4.1 Opus, and 4.5 Haiku.
  • Google: Gemini 3 Flash Preview and Gemini 3.1 Flash-Lite Preview (replaced by the stable Gemini 3.1 Flash-Lite).
  • xAI: Grok 2 vision, Grok 4, Grok 4 Fast (reasoning), and Grok 4 Fast (non-reasoning).
  • Mistral: Pixtral Large.
  • Moonshot: Kimi K2.5 (#103).

If your saved model was removed, the add-on no longer crashes. Pick a new model in the settings.

Fixes

  • Error messages now include the reason the API gave, such as Mistral's subscription tier error.
  • Mistral reasoning replies are returned as text instead of a raw list of chunks.
  • The Gemini API key is sent in a header instead of the URL (#97).
  • Seer no longer fails silently because of a missing prompt setting.
  • The UI viewer reports unexpected errors and empty results, and no longer repeats errors.
  • Fixed a settings panel crash after a config migration.
  • Fixed an error when a dependency download fails.
  • Fixed the misspelling "incompatable" in the message shown when an old version is installed (#108).
  • The readme now explains that dependencies are downloaded on first run (#107).
  • Face positioning now describes faces that are far to the right or far above correctly.
  • Many computer use fixes: sharper screenshots, screenshots sized within each provider's limits so clicks land on target, correct handling of modifier keys, scrolling, mouse buttons, key names, and wait actions, and better handling of failed actions.

Translations

  • Updated: Chinese (Simplified), Czech, French, Italian, Polish, Russian, Serbian, Slovak, Spanish, Turkish, and Ukrainian.
  • New Italian readme. All translated readmes now use the menu names from the translated UI.
  • More strings can now be translated, including error messages and model descriptions.

Thanks

Don't miss a new AI-content-describer release

NewReleases is sending notifications on new releases.