# `PhoenixKitAI.Providers.OpenRouter`
[🔗](https://github.com/BeamLabEU/phoenix_kit_ai/blob/0.23.0/lib/phoenix_kit_ai/providers/openrouter.ex#L1)

`PhoenixKitAI.Provider` adapter for OpenRouter's unified Images API.

`POST <base_url>/images` serves both generation and editing: the prompt
plus typed options (`aspect_ratio`, `resolution`, `size`, `quality`,
`background`, `output_format`, `output_compression`, `n`, `seed`) and,
for edits, the inputs as `input_references`. Every image model on the
gateway — Gemini's, OpenAI's, Seedream, FLUX, Grok Imagine, Qwen and
the rest — sits behind the same request, and `GET /images/models` says
per model which options it accepts (see `PhoenixKitAI.Images.ImageModel`).

`transport: :chat` keeps the older chat-completions path (image parts,
`modalities: ["image","text"]`) for models that only answer there.

`provider_options` becomes OpenRouter's `provider.options` object;
`provider_routing` (`order`, `only`, `ignore`, `sort`,
`allow_fallbacks`) is merged into the same `provider` object.

---

*Consult [api-reference.md](api-reference.md) for complete listing*