Merge pull request #3887 from cristinaponcela/feat/image-outputs

feat: image content
This commit is contained in:
Mario Zechner
2026-05-08 15:57:06 +02:00
committed by GitHub
13 changed files with 1138 additions and 3 deletions

View File

@@ -0,0 +1,90 @@
import { readFileSync } from "node:fs";
import { dirname, join } from "node:path";
import { fileURLToPath } from "node:url";
import { describe, expect, it } from "vitest";
import { getImageModel } from "../src/image-models.js";
import { generateImages } from "../src/images.js";
import type { ImageContent, ImagesContext, ImagesModel, ProviderImagesOptions } from "../src/types.js";
const __filename = fileURLToPath(import.meta.url);
const __dirname = dirname(__filename);
type ImagesOptionsWithExtras = ProviderImagesOptions & Record<string, unknown>;
async function basicImageGeneration<TApi extends string>(model: ImagesModel<TApi>, options?: ImagesOptionsWithExtras) {
const context: ImagesContext = {
input: [{ type: "text", text: "Generate a simple red circle on a plain white background. No text." }],
};
const response = await generateImages(model, context, options);
expect(response.stopReason, `Error: ${response.errorMessage}`).toBe("stop");
expect(response.errorMessage).toBeFalsy();
expect(response.output.some((item) => item.type === "image")).toBe(true);
expect(response.timestamp).toBeGreaterThan(0);
}
async function handleTextAndImageOutput<TApi extends string>(
model: ImagesModel<TApi>,
options?: ImagesOptionsWithExtras,
) {
if (!model.output.includes("text")) {
console.log(`Skipping text+image output test - model ${model.id} doesn't support text output`);
return;
}
const context: ImagesContext = {
input: [{ type: "text", text: "Generate a red circle and include a brief description of the image." }],
};
const response = await generateImages(model, context, options);
expect(response.stopReason, `Error: ${response.errorMessage}`).toBe("stop");
expect(response.output.some((item) => item.type === "image")).toBe(true);
expect(response.output.some((item) => item.type === "text" && item.text.trim().length > 0)).toBe(true);
}
async function handleImageInput<TApi extends string>(model: ImagesModel<TApi>, options?: ImagesOptionsWithExtras) {
if (!model.input.includes("image")) {
console.log(`Skipping image input test - model ${model.id} doesn't support image input`);
return;
}
const imagePath = join(__dirname, "data", "red-circle.png");
const imageBuffer = readFileSync(imagePath);
const imageContent: ImageContent = {
type: "image",
data: imageBuffer.toString("base64"),
mimeType: "image/png",
};
const context: ImagesContext = {
input: [{ type: "text", text: "Create a variation of this image with a blue background." }, imageContent],
};
const response = await generateImages(model, context, options);
expect(response.stopReason, `Error: ${response.errorMessage}`).toBe("stop");
expect(response.output.some((item) => item.type === "image")).toBe(true);
}
describe("Images E2E Tests", () => {
describe.skipIf(!process.env.OPENROUTER_API_KEY)(
"OpenRouter Images Provider (google/gemini-2.5-flash-image)",
() => {
const model = getImageModel("openrouter", "google/gemini-2.5-flash-image");
it("should generate a basic image", { retry: 3 }, async () => {
await basicImageGeneration(model);
});
it("should handle text plus image output", { retry: 3 }, async () => {
await handleTextAndImageOutput(model);
});
it("should handle image input", { retry: 3 }, async () => {
await handleImageInput(model);
});
},
);
});

View File

@@ -0,0 +1,140 @@
import { beforeEach, describe, expect, it, vi } from "vitest";
import { generateImages } from "../src/images.js";
import type { ImagesContext, ImagesModel } from "../src/types.js";
const mockState = vi.hoisted(() => ({
lastParams: undefined as unknown,
lastRequestOptions: undefined as unknown,
}));
vi.mock("openai", () => {
class FakeOpenAI {
chat = {
completions: {
create: (params: unknown, requestOptions?: unknown) => {
mockState.lastParams = params;
mockState.lastRequestOptions = requestOptions;
const signal = (requestOptions as { signal?: AbortSignal } | undefined)?.signal;
if (signal?.aborted) {
const error = new Error("Request aborted");
return {
withResponse: async () => {
throw error;
},
};
}
const response = {
id: "img-1",
usage: {
prompt_tokens: 12,
completion_tokens: 34,
prompt_tokens_details: { cached_tokens: 0 },
},
choices: [
{
message: {
content: "Here is your image.",
images: [{ image_url: "data:image/png;base64,ZmFrZS1wbmc=" }],
},
},
],
};
const promise = Promise.resolve(response) as Promise<typeof response> & {
withResponse: () => Promise<{
data: typeof response;
response: { status: number; headers: Headers };
}>;
};
promise.withResponse = async () => ({
data: response,
response: { status: 200, headers: new Headers() },
});
return promise;
},
},
};
}
return { default: FakeOpenAI };
});
describe("openrouter images", () => {
beforeEach(() => {
mockState.lastParams = undefined;
mockState.lastRequestOptions = undefined;
});
it("returns text plus images in final output", async () => {
const model: ImagesModel<"openrouter-images"> = {
id: "google/gemini-3.1-flash-image-preview",
name: "Gemini 3.1 Flash Image Preview",
api: "openrouter-images",
provider: "openrouter",
baseUrl: "https://openrouter.ai/api/v1",
input: ["text", "image"],
output: ["text", "image"],
cost: { input: 0.015, output: 0.03, cacheRead: 0, cacheWrite: 0 },
headers: { "HTTP-Referer": "https://example.com" },
};
const context: ImagesContext = {
input: [{ type: "text", text: "Generate a dog" }],
};
const output = await generateImages(model, context, { apiKey: "test" });
expect(output.stopReason).toBe("stop");
expect(output.responseId).toBe("img-1");
expect(output.output[0]).toMatchObject({ type: "text", text: "Here is your image." });
expect(output.output[1]).toMatchObject({ type: "image", mimeType: "image/png", data: "ZmFrZS1wbmc=" });
const params = mockState.lastParams as {
stream?: boolean;
modalities?: string[];
messages?: [{ content?: [{ type: string; text?: string }] }];
};
expect(params.stream).toBe(false);
expect(params.modalities).toEqual(["image", "text"]);
expect(params.messages?.[0]?.content?.[0]).toMatchObject({ type: "text", text: "Generate a dog" });
});
it("passes through abort signal and returns aborted result", async () => {
const model: ImagesModel<"openrouter-images"> = {
id: "black-forest-labs/flux.2-pro",
name: "FLUX.2 Pro",
api: "openrouter-images",
provider: "openrouter",
baseUrl: "https://openrouter.ai/api/v1",
input: ["text", "image"],
output: ["image"],
cost: { input: 0.015, output: 0.03, cacheRead: 0, cacheWrite: 0 },
};
const context: ImagesContext = {
input: [{ type: "text", text: "Generate a dog" }],
};
const controller = new AbortController();
controller.abort();
const output = await generateImages(model, context, { apiKey: "test", signal: controller.signal });
expect(output.stopReason).toBe("aborted");
expect(output.errorMessage).toBe("Request aborted");
expect(mockState.lastRequestOptions).toMatchObject({ signal: controller.signal });
});
it("generateImages resolves the final assistant images result", async () => {
const model: ImagesModel<"openrouter-images"> = {
id: "black-forest-labs/flux.2-pro",
name: "FLUX.2 Pro",
api: "openrouter-images",
provider: "openrouter",
baseUrl: "https://openrouter.ai/api/v1",
input: ["text", "image"],
output: ["image"],
cost: { input: 0.015, output: 0.03, cacheRead: 0, cacheWrite: 0 },
};
const context: ImagesContext = {
input: [{ type: "text", text: "Generate a dog" }],
};
const output = await generateImages(model, context, { apiKey: "test" });
expect(output.output.some((item) => item.type === "image")).toBe(true);
});
});