[Inference Providers] deepinfra: add text-to-image support

[ovuruska]

What

Adds DeepInfraTextToImageTask so the deepinfra provider supports the text-to-image task.

DeepInfra exposes an OpenAI-compatible image endpoint, so this mirrors the existing together/nscale text-to-image helpers:

• Route: v1/openai/images/generations
• preparePayload: forwards prompt, sets response_format from outputType (url vs base64), and applies prompt/model after caller parameters so neither can be overridden.
• getResponse: handles url and b64_json responses, honoring json/url/dataUrl/blob output types.

Wired in getProviderHelper.ts (deepinfra['text-to-image']). Existing deepinfra tasks are untouched.

Test

Added a textToImage integration case to the DeepInfra suite in InferenceClient.spec.ts (same skip-gated pattern as the other deepinfra tasks).

🤖 Generated with Claude Code

---

Note

Low Risk
Additive provider helper for an existing inference client; no auth, routing, or existing DeepInfra task changes. Risk is limited to image payload/response mapping for this provider.

Overview
Adds text-to-image for the deepinfra provider via a new DeepInfraTextToImageTask wired in getProviderHelper.

Calls DeepInfra’s OpenAI-compatible v1/openai/images/generations endpoint. Payload maps inputs to prompt, sets response_format from outputType, and applies prompt/model last so callers cannot override them. Responses support json, url, dataUrl, and blob (from url or b64_json).

Includes a skip-gated DeepInfra textToImage integration test using stabilityai/sdxl-turbo.

^{Reviewed by Cursor Bugbot for commit ca24da6. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit ca24da6. Configure here.}

[cursor]

Unsupported URL image output

Medium Severity

preparePayload sets response_format to url when outputType is url, but DeepInfra’s OpenAI-compatible images API only supports b64_json. That request can fail, or getResponse may return a Blob for a url request when only b64_json comes back. Other b64-only providers such as nscale reject url up front.

Additional Locations (1)

• packages/inference/src/providers/deepinfra.ts#L352-L363

 

^{Reviewed by Cursor Bugbot for commit ca24da6. Configure here.}