Add image generation model and update provider to support it

[milindmore22]

Description of the Change

This PR adds support for Image generation model based on experimental feature of ollama please refer blog

This pull request adds support for image generation models to the Ollama provider integration. The most significant changes include the introduction of a new OllamaImageGenerationModel class, updates to model metadata detection to identify image generation capabilities, and enhancements to provider logic to instantiate the correct model type based on capabilities.

Image Generation Model Support:

• Added a new OllamaImageGenerationModel class that implements image generation using the Ollama /api/generate endpoint, including prompt extraction, response parsing, and custom timeout handling.
• Updated the provider logic in OllamaProvider to instantiate OllamaImageGenerationModel when a model supports image generation, based on its capabilities. [1] [2]

Model Metadata Detection:

• Enhanced OllamaModelMetadataDirectory to detect image generation models by checking for the 'image' capability, and to construct appropriate ModelMetadata for such models with relevant options and output types. [1] [2] [3] [4]

Tooling and Static Analysis:

• Updated phpstan.neon.dist to temporarily ignore the new OllamaImageGenerationModel.php file for static analysis, consistent with other files depending on unreleased WordPress core AI Client classes.

Warning

Supported only on MacOS (because experiments are only avilable on MacOS) Tested with x/flux2-klein and x/z-image-turbo

Screenshots

1. Verify the model is ready by visiting Settings->Ollama Settings

2. Visit Media->Generate Image, Enter prompt

3. Generate Image

4. Image in Media Gallery After saving it.

Closes #

How to test the Change

Download Model

ollama pull x/flux2-klein

Test Model

ollama run x/flux2-klein "a cat holding a sign that says hello world"

Test Curl

curl http://localhost:11434/api/generate -d '{
  "model": "x/flux2-klein",
  "prompt": "a cat holding a sign that says hello world",
  "stream": false
}'

Now you can try generate Image in WordPress

Changelog Entry

Added - support Image generation model

Credits

Props @milindmore22

Checklist:

• I agree to follow this project's Code of Conduct.
• I have updated the documentation accordingly.
• I have added Critical Flows, Test Cases, and/or End-to-End Tests to cover my change.
• All new and existing tests pass.

[dkotter]

@milindmore22 Thanks for the PR. We use develop as our base branch and main for our release branch. Can you open this PR against develop?

[dkotter]

Few minor comments but otherwise this looks good to me and tests well!

[sentry]

Codecov Report

❌ Patch coverage is 99.03846% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@4550eeb). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
includes/Metadata/OllamaModelMetadataDirectory.php	96.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop      #30   +/-   ##
==========================================
  Coverage           ?   61.68%           
  Complexity         ?      100           
==========================================
  Files              ?        7           
  Lines              ?      462           
  Branches           ?        0           
==========================================
  Hits               ?      285           
  Misses             ?      177           
  Partials           ?        0           

Flag	Coverage Δ
unit	61.68% <99.03%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[dkotter]

Would also be great to get unit tests added to cover these new changes

[TylerB24890]

All looks good to me!

Is there ever a case where the model would have the image capability but not be an image generation model? If not, I would propose updating the name of the isLikelyImageGenerationModel() method to isImageGenerationModel() -- less misleading and it matches the variable name it's being applied to

[milindmore22]

Renamed isLikelyImageGenerationModel to isImageGenerationModel, but unable to test locally due to an Ollama environment error (failing to initialize MLX due to missing libmlxc.dylib). with latest version of Ollama, l need to resolve that environment issue before I can verify the change.

[Update]
ollama/ollama#15471

[dkotter]

Renamed isLikelyImageGenerationModel to isImageGenerationModel, but unable to test locally due to an Ollama environment error (failing to initialize MLX due to missing libmlxc.dylib). with latest version of Ollama, l need to resolve that environment issue before I can verify the change.

[Update] ollama/ollama#15471

Yeah, I'm getting the same error now. Going to hold on merging this until they get that resolved (hopefully soon)

[jeffpaul]

According to ollama/ollama#15433 the related PR ollama/ollama#15443 was in the v0.20.5 release so perhaps a local Ollama update will see this working again?

[dkotter]

According to ollama/ollama#15433 the related PR ollama/ollama#15443 was in the v0.20.5 release so perhaps a local Ollama update will see this working again?

Yeah, I pulled the latest version (0.27.0) and still seeing the same issue. Reading through that thread, sounds like most people are still having the same problem (and would be ideal if the issue was opened back up). There is a "fix" detailed on the thread for those running it through Homebrew but I have it installed as a normal package so wasn't able to test that.

[milindmore22]

The issue has been resolved in the latest Ollama version 0.21.0, as reported in the pull request ollama/ollama#15588.
I have successfully tested the changes and they are working as expected at my end. 👍🏼