feat(video): add OpenRouter and Vertex AI (Veo) video generation

Video generation was xAI-only. Adds an adapter layer under open-sse/handlers/videoProviders/ so /v1/videos/* can target OpenRouter or Google Cloud credentials. A provider with no adapter keeps the exact previous behaviour (raw body to {baseUrl}/{action}, poll {baseUrl}/{id}, verbatim passthrough), so the xAI path is unchanged. - openrouter: async job shape identical to xAI; creation POSTs to the /videos collection root (no /generations suffix) and the registry HTTP-Referer / X-Title headers are applied. Bodies pass through verbatim. - vertex: two-way translation, since Veo does not speak the OpenAI-ish videos shape. create -> :predictLongRunning { instances[], parameters{} }, poll -> :fetchPredictOperation (Veo has no REST GET poll). The operation resource name is base64url-encoded into the job id so GET /v1/videos/{id} stays a flat path. Access tokens are minted from Service Account JSON via the existing refreshVertexToken; raw API keys are rejected up front. The operation response maps back onto the { id, status, video, videos } shape clients already poll. - videoCore: the request plan is rebuilt per attempt, so the 401 -> refresh once -> retry once path picks up the refreshed token. Adapter validation errors return 400 before any upstream call, so a malformed request can never create a billable job. - videoGeneration: GET /v1/videos/{id} resolves the provider from the pinned x-connection-id connection, then ?provider=, then falls back to the xAI default. - registry: openrouter and vertex gain videoConfig, the video serviceKind and video-kind models (Veo 3.1 / 3 / 2, Sora 2 Pro, Seedance 2.0).
2026-09-10 22:05:10 +07:00
parent 807553e246
commit 3288bbc47e
9 changed files with 551 additions and 16 deletions
--- a/open-sse/handlers/videoCore.js
+++ b/open-sse/handlers/videoCore.js
@@ -2,6 +2,7 @@ import { createErrorResult } from "../utils/error.js";
 import { HTTP_STATUS } from "../config/runtimeConfig.js";
 import { refreshTokenByProvider } from "../services/tokenRefresh.js";
 import { PROVIDER_MEDIA } from "../providers/index.js";
+import { getVideoAdapter } from "./videoProviders/index.js";

 // Upstream fetch deadline for video job submission/polling (the job itself is
 // async upstream — this only bounds the HTTP round-trip, not video rendering).
@@ -94,21 +95,49 @@ export async function handleVideoProxyCore({
    return createErrorResult(HTTP_STATUS.BAD_REQUEST, `Unknown video action: ${action}`);
  }

-  const method = requestId ? "GET" : "POST";
-  const url = buildUpstreamUrl(config, action, requestId);
+  const adapter = getVideoAdapter(provider);
  const fetchSignal = combineSignals(signal, timeoutMs);

-  const doFetch = (token) =>
-    fetch(url, {
+  // Default (xAI shape) request plan; adapters override URL/method/headers/body.
+  const defaultPlan = () => {
+    const method = requestId ? "GET" : "POST";
+    return {
      method,
-      headers: buildHeaders({ token, contentType: method === "POST" ? contentType : null, idempotencyKey: method === "POST" ? idempotencyKey : null }),
+      url: buildUpstreamUrl(config, action, requestId),
+      headers: buildHeaders({
+        token: credentials?.accessToken || credentials?.apiKey,
+        contentType: method === "POST" ? contentType : null,
+        idempotencyKey: method === "POST" ? idempotencyKey : null,
+      }),
      body: method === "POST" ? rawBody : undefined,
-      signal: fetchSignal,
-    });
+    };
+  };

+  // Rebuilt per attempt so the auth retry below picks up the refreshed token.
+  const doFetch = async () => {
+    const plan = adapter
+      ? await adapter.buildRequest({
+          config, action, requestId, rawBody, contentType, idempotencyKey, credentials, log,
+          token: credentials?.accessToken || credentials?.apiKey,
+        })
+      : defaultPlan();
+    if (plan.error) return { planError: plan.error };
+    return {
+      response: await fetch(plan.url, {
+        method: plan.method,
+        headers: plan.headers,
+        body: plan.body,
+        signal: fetchSignal,
+      }),
+    };
+  };
+
+  const method = requestId ? "GET" : "POST";
  let upstream;
  try {
-    upstream = await doFetch(credentials?.accessToken || credentials?.apiKey);
+    const first = await doFetch();
+    if (first.planError) return createErrorResult(HTTP_STATUS.BAD_REQUEST, `[${provider}] ${first.planError}`);
+    upstream = first.response;
  } catch (error) {
    if (error?.name === "AbortError" || error?.name === "TimeoutError") {
      return createErrorResult(HTTP_STATUS.REQUEST_TIMEOUT, `[${provider}] video ${method} aborted: ${error.message}`);
@@ -136,7 +165,9 @@ export async function handleVideoProxyCore({
        await upstream.body?.cancel?.();
      } catch { /* noop */ }
      try {
-        upstream = await doFetch(credentials.accessToken || credentials.apiKey);
+        const retry = await doFetch();
+        if (retry.planError) return createErrorResult(HTTP_STATUS.BAD_REQUEST, `[${provider}] ${retry.planError}`);
+        upstream = retry.response;
      } catch (error) {
        return createErrorResult(HTTP_STATUS.BAD_GATEWAY, sanitizeSecrets(`[${provider}] video retry after refresh failed: ${error.message}`, credentials));
      }
@@ -152,13 +183,25 @@ export async function handleVideoProxyCore({
    return createErrorResult(upstream.status, `[${provider}] ${message.slice(0, 2000)}`);
  }

-  // Success: pass the upstream JSON through untouched (request_id / status / video.url).
+  // Success: pass the upstream JSON through untouched (request_id / status / video.url),
+  // unless the adapter maps a provider-native shape onto it (Vertex operations).
+  let outBody = bodyText;
+  let outType = upstream.headers.get("content-type") || "application/json";
+  if (adapter?.transformResponse) {
+    try {
+      outBody = JSON.stringify(adapter.transformResponse(JSON.parse(bodyText)));
+      outType = "application/json";
+    } catch {
+      // Non-JSON or unexpected shape — fall back to the raw upstream body.
+    }
+  }
+
  return {
    success: true,
-    response: new Response(bodyText, {
+    response: new Response(outBody, {
      status: upstream.status,
      headers: {
-        "Content-Type": upstream.headers.get("content-type") || "application/json",
+        "Content-Type": outType,
        "Access-Control-Allow-Origin": "*",
      },
    }),
--- a/open-sse/handlers/videoProviders/index.js
+++ b/open-sse/handlers/videoProviders/index.js
@@ -0,0 +1,13 @@
+// Video provider adapters.
+//
+// Default (no adapter) = xAI shape: raw body forwarded to {baseUrl}/{action},
+// polled at {baseUrl}/{id}, upstream JSON passed through verbatim.
+// A provider only needs an adapter when its wire format differs from that.
+import openrouter from "./openrouter.js";
+import vertex from "./vertex.js";
+
+const ADAPTERS = { openrouter, vertex };
+
+export function getVideoAdapter(provider) {
+  return ADAPTERS[provider] || null;
+}
--- a/open-sse/handlers/videoProviders/openrouter.js
+++ b/open-sse/handlers/videoProviders/openrouter.js
@@ -0,0 +1,39 @@
+// OpenRouter video jobs — https://openrouter.ai/docs/api/api-reference/videos
+//
+// Same async shape as xAI (POST → { id, status }, GET → status/unsigned_urls),
+// two differences only: creation POSTs to the collection root (no `/generations`
+// suffix) and the account headers come from the registry entry.
+// Response bodies are passed through verbatim.
+
+// ponytail: generations only — OpenRouter has no edits/extensions endpoint today.
+const SUPPORTED_ACTIONS = new Set(["generations"]);
+
+function headers(config, token) {
+  return {
+    Accept: "application/json",
+    ...(config.headers || {}),
+    ...(token ? { Authorization: `Bearer ${token}` } : {}),
+  };
+}
+
+export default {
+  buildRequest({ config, action, requestId, rawBody, contentType, token }) {
+    const base = config.baseUrl.replace(/\/$/, "");
+
+    if (requestId) {
+      return { method: "GET", url: `${base}/${encodeURIComponent(requestId)}`, headers: headers(config, token) };
+    }
+    if (!SUPPORTED_ACTIONS.has(action)) {
+      return { error: `OpenRouter video supports 'generations' only (got '${action}')` };
+    }
+    if (contentType && !contentType.includes("application/json")) {
+      return { error: "OpenRouter video requires an application/json body" };
+    }
+    return {
+      method: "POST",
+      url: base,
+      headers: { ...headers(config, token), "Content-Type": "application/json" },
+      body: rawBody,
+    };
+  },
+};
--- a/open-sse/handlers/videoProviders/vertex.js
+++ b/open-sse/handlers/videoProviders/vertex.js
@@ -0,0 +1,145 @@
+// Vertex AI (Veo) video jobs.
+//
+// Vertex does NOT speak the OpenAI-ish /v1/videos shape, so unlike OpenRouter
+// this adapter translates both directions:
+//   create → POST {model}:predictLongRunning  { instances[], parameters{} } → { name }
+//   poll   → POST {model}:fetchPredictOperation { operationName } → { done, response }
+// Docs: https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/veo-video-generation
+//
+// The operation name is a resource path (contains "/"), so it is base64url-encoded
+// into the job id returned to the client — GET /v1/videos/{id} stays a flat path.
+import { parseVertexSaJson, refreshVertexToken } from "../../services/tokenRefresh.js";
+
+const DEFAULT_LOCATION = "us-central1";
+
+const encodeJobId = (name) => Buffer.from(name, "utf8").toString("base64url");
+const decodeJobId = (id) => Buffer.from(id, "base64url").toString("utf8");
+
+async function resolveAuth(credentials, log) {
+  const saJson = parseVertexSaJson(credentials?.apiKey);
+  const projectId =
+    saJson?.project_id ||
+    credentials?.projectId ||
+    credentials?.providerSpecificData?.projectId;
+  const location = credentials?.providerSpecificData?.location || DEFAULT_LOCATION;
+
+  if (!projectId) {
+    return { error: "Vertex video requires a project_id — use Service Account JSON or set providerSpecificData.projectId" };
+  }
+
+  let token = credentials?.accessToken;
+  if (saJson) {
+    const minted = await refreshVertexToken(saJson, log);
+    if (!minted?.accessToken) return { error: "Vertex video: failed to mint access token from service account JSON" };
+    token = minted.accessToken;
+  }
+  if (!token) return { error: "Vertex video requires Service Account JSON or an OAuth access token (raw API keys are not supported)" };
+
+  return { token, projectId, location };
+}
+
+/** OpenAI-ish video body → Vertex predictLongRunning body. */
+function toVertexBody(body) {
+  const instance = { prompt: body.prompt };
+  // Image-to-video: accept the Vertex-native shape or a bare data URL / base64 string.
+  const image = body.image ?? body.image_url;
+  if (image && typeof image === "object") {
+    instance.image = image;
+  } else if (typeof image === "string") {
+    const match = image.match(/^data:([^;]+);base64,(.*)$/s);
+    instance.image = match
+      ? { bytesBase64Encoded: match[2], mimeType: match[1] }
+      : { gcsUri: image };
+  }
+  if (body.video && typeof body.video === "object") instance.video = body.video;
+
+  const parameters = {};
+  if (body.n != null) parameters.sampleCount = Number(body.n);
+  if (body.duration != null) parameters.durationSeconds = Number(body.duration);
+  if (body.aspect_ratio) parameters.aspectRatio = body.aspect_ratio;
+  if (body.resolution) parameters.resolution = body.resolution;
+  if (body.seed != null) parameters.seed = body.seed;
+  if (body.negative_prompt) parameters.negativePrompt = body.negative_prompt;
+  // Without storageUri Vertex returns inline base64 bytes; a GCS bucket keeps
+  // the poll response small and is what production callers want.
+  if (body.storage_uri) parameters.storageUri = body.storage_uri;
+  if (body.generate_audio != null) parameters.generateAudio = !!body.generate_audio;
+
+  return { instances: [instance], ...(Object.keys(parameters).length ? { parameters } : {}) };
+}
+
+/** Vertex operation → the async-job shape 9Router clients already poll for. */
+function fromVertexOperation(json) {
+  if (!json?.name) return json;
+  const id = encodeJobId(json.name);
+  if (json.error) {
+    return { id, request_id: id, status: "failed", error: json.error };
+  }
+  if (!json.done) {
+    return { id, request_id: id, status: "pending" };
+  }
+  const samples =
+    json.response?.videos ||
+    json.response?.generateVideoResponse?.generatedSamples ||
+    [];
+  const videos = samples.map((s) => ({
+    url: s.gcsUri || s.video?.uri || s.uri || null,
+    b64_json: s.bytesBase64Encoded || s.video?.bytesBase64Encoded || null,
+    mime_type: s.mimeType || s.video?.mimeType || "video/mp4",
+  }));
+  return { id, request_id: id, status: "completed", video: videos[0] || null, videos };
+}
+
+export default {
+  async buildRequest({ config, action, requestId, rawBody, contentType, credentials, log }) {
+    if (contentType && !contentType.includes("application/json")) {
+      return { error: "Vertex video requires an application/json body" };
+    }
+
+    const auth = await resolveAuth(credentials, log);
+    if (auth.error) return { error: auth.error };
+    const { token, projectId, location } = auth;
+    const base = (config.baseUrl || "https://aiplatform.googleapis.com").replace(/\/$/, "");
+    const headers = { Accept: "application/json", "Content-Type": "application/json", Authorization: `Bearer ${token}` };
+
+    if (requestId) {
+      let operationName;
+      try {
+        operationName = decodeJobId(requestId);
+      } catch {
+        return { error: "Invalid Vertex video job id" };
+      }
+      const modelPath = operationName.split("/operations/")[0];
+      if (!modelPath || modelPath === operationName) return { error: "Invalid Vertex video job id" };
+      return {
+        method: "POST",
+        url: `${base}/v1/${modelPath}:fetchPredictOperation`,
+        headers,
+        body: JSON.stringify({ operationName }),
+      };
+    }
+
+    if (action !== "generations") {
+      // ponytail: Veo extend/edit go through generations with `video`/`image` in the body.
+      return { error: `Vertex video supports 'generations' only (got '${action}')` };
+    }
+
+    let body;
+    try {
+      body = JSON.parse(typeof rawBody === "string" ? rawBody : rawBody.toString("utf8"));
+    } catch {
+      return { error: "Invalid JSON body" };
+    }
+    if (!body.model) return { error: "Vertex video requires a model (e.g. vertex/veo-3.1-generate-preview)" };
+    if (!body.prompt && !body.image && !body.image_url) return { error: "Vertex video requires a prompt or an image" };
+
+    return {
+      method: "POST",
+      url: `${base}/v1/projects/${projectId}/locations/${location}/publishers/google/models/${body.model}:predictLongRunning`,
+      headers,
+      body: JSON.stringify(toVertexBody(body)),
+    };
+  },
+
+  transformResponse: fromVertexOperation,
+};
--- a/open-sse/providers/registry/openrouter.js
+++ b/open-sse/providers/registry/openrouter.js
@@ -40,8 +40,11 @@ export default {
    { id: "openai/gpt-image-1", name: "GPT Image 1 (via OpenRouter)", params: ["n","size","quality","response_format"], kind: "image" },
    { id: "google/imagen-3.0-generate-002", name: "Imagen 3 (via OpenRouter)", params: ["n","size"], kind: "image" },
    { id: "black-forest-labs/FLUX.1-schnell", name: "FLUX.1 Schnell (via OpenRouter)", params: ["n","size"], kind: "image" },
+    { id: "google/veo-3.1", name: "Veo 3.1 (via OpenRouter)", params: ["duration","aspect_ratio","resolution"], kind: "video" },
+    { id: "openai/sora-2-pro", name: "Sora 2 Pro (via OpenRouter)", params: ["duration","aspect_ratio","resolution"], kind: "video" },
+    { id: "bytedance/seedance-2.0", name: "Seedance 2.0 (via OpenRouter)", params: ["duration","aspect_ratio","resolution"], kind: "video" },
  ],
-  serviceKinds: ["llm","embedding","tts","imageToText"],
+  serviceKinds: ["llm","embedding","tts","imageToText","video"],
  ttsConfig: {
    baseUrl: "https://openrouter.ai/api/v1/chat/completions",
    defaultModel: "openai/gpt-4o-mini-tts",
@@ -57,6 +60,12 @@ export default {
    baseUrl: "https://openrouter.ai/api/v1/images/generations",
    headers: {"HTTP-Referer":"https://endpoint-proxy.local","X-Title":"Endpoint Proxy"},
  },
+  // Async video jobs (POST /videos → { id, status }, GET /videos/{id} polls).
+  // Docs: https://openrouter.ai/docs/api/api-reference/videos
+  videoConfig: {
+    baseUrl: "https://openrouter.ai/api/v1/videos",
+    headers: {"HTTP-Referer":"https://endpoint-proxy.local","X-Title":"Endpoint Proxy"},
+  },
  modelsFetcher: { url: "https://openrouter.ai/api/v1/models", type: "openrouter-free" },
  passthroughModels: true,
 };
--- a/open-sse/providers/registry/vertex.js
+++ b/open-sse/providers/registry/vertex.js
@@ -27,6 +27,13 @@ export default {
    { id: "gemini-3.1-flash-lite-preview", name: "Gemini 3.1 Flash Lite Preview" },
    { id: "gemini-3-flash-preview", name: "Gemini 3 Flash Preview" },
    { id: "gemini-2.5-flash", name: "Gemini 2.5 Flash" },
+    { id: "veo-3.1-generate-preview", name: "Veo 3.1 (Preview)", params: ["duration","aspect_ratio","resolution","negative_prompt","seed","storage_uri","generate_audio"], kind: "video" },
+    { id: "veo-3.1-fast-generate-preview", name: "Veo 3.1 Fast (Preview)", params: ["duration","aspect_ratio","resolution","negative_prompt","seed","storage_uri","generate_audio"], kind: "video" },
+    { id: "veo-3.0-generate-001", name: "Veo 3", params: ["duration","aspect_ratio","resolution","negative_prompt","seed","storage_uri","generate_audio"], kind: "video" },
+    { id: "veo-2.0-generate-001", name: "Veo 2", params: ["duration","aspect_ratio","negative_prompt","seed","storage_uri"], kind: "video" },
  ],
-  serviceKinds: ["llm","imageToText"],
+  serviceKinds: ["llm","imageToText","video"],
+  // Veo via predictLongRunning + fetchPredictOperation (adapter: handlers/videoProviders/vertex.js).
+  // Docs: https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/veo-video-generation
+  videoConfig: { baseUrl: "https://aiplatform.googleapis.com" },
 };
--- a/src/sse/handlers/videoGeneration.js
+++ b/src/sse/handlers/videoGeneration.js
@@ -5,7 +5,7 @@ import {
  extractApiKey,
  isValidApiKey,
 } from "../services/auth.js";
-import { getSettings } from "@/lib/localDb";
+import { getSettings, getProviderConnectionById } from "@/lib/localDb";
 import { getModelInfo } from "../services/model.js";
 import { handleVideoProxyCore, getVideoConfig, sanitizeSecrets } from "open-sse/handlers/videoCore.js";
 import { errorResponse, unavailableResponse } from "open-sse/utils/error.js";
@@ -17,6 +17,21 @@ import * as log from "../utils/logger.js";
 // (bare model id, or multipart bodies we deliberately don't parse) land here.
 const DEFAULT_VIDEO_PROVIDER = "xai";

+/**
+ * Poll requests carry no model, so the provider comes from the pinned
+ * connection (`x-connection-id`, returned on create) or an explicit
+ * `?provider=` — falling back to the historical xAI default.
+ */
+async function resolveGetProvider(request, connectionId) {
+  if (connectionId) {
+    const conn = await getProviderConnectionById(connectionId).catch(() => null);
+    if (conn?.provider && getVideoConfig(conn.provider)) return conn.provider;
+  }
+  const queried = new URL(request.url).searchParams.get("provider");
+  if (queried && getVideoConfig(queried)) return queried;
+  return DEFAULT_VIDEO_PROVIDER;
+}
+
 // Creation POSTs are billable jobs — only rotate to another account for
 // errors that upstream rejects BEFORE creating a job (auth/quota). A 5xx may
 // have created the job, so it is returned to the caller instead of re-sent.
@@ -185,8 +200,8 @@ export async function handleVideoGet(request, requestId) {

  if (!requestId) return errorResponse(HTTP_STATUS.BAD_REQUEST, "Missing video request id");

-  const provider = DEFAULT_VIDEO_PROVIDER;
  const preferredConnectionId = request.headers.get("x-connection-id") || null;
+  const provider = await resolveGetProvider(request, preferredConnectionId);

  const credentials = await getProviderCredentials(provider, null, null, { preferredConnectionId });
  if (!credentials || credentials.allRateLimited) {
--- a/tests/unit/video-providers.test.js
+++ b/tests/unit/video-providers.test.js
@@ -0,0 +1,263 @@
+/**
+ * Unit tests for the OpenRouter + Vertex (Veo) video adapters.
+ *
+ * Covers:
+ *  - registry wiring (videoConfig, video serviceKind, video-kind models)
+ *  - OpenRouter: POST to the collection root, GET poll, verbatim passthrough
+ *  - Vertex: predictLongRunning body translation, fetchPredictOperation polling,
+ *    operation-name round-trip through the job id, response mapping
+ *  - xAI default path is unchanged by the adapter hook
+ */
+
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+
+vi.mock("open-sse/services/tokenRefresh.js", async (importOriginal) => {
+  const actual = await importOriginal();
+  return { ...actual, refreshTokenByProvider: vi.fn(), refreshVertexToken: vi.fn() };
+});
+
+import { handleVideoProxyCore, getVideoConfig } from "open-sse/handlers/videoCore.js";
+import { refreshVertexToken } from "open-sse/services/tokenRefresh.js";
+import { PROVIDER_MEDIA, PROVIDER_MODELS } from "open-sse/providers/index.js";
+
+const originalFetch = global.fetch;
+const jsonResponse = (body, status = 200) =>
+  new Response(JSON.stringify(body), { status, headers: { "Content-Type": "application/json" } });
+
+// Vertex operation names are resource paths; the adapter base64url-encodes them.
+const OPERATION_NAME =
+  "projects/proj-1/locations/us-central1/publishers/google/models/veo-3.1-generate-preview/operations/op-abc";
+const JOB_ID = Buffer.from(OPERATION_NAME, "utf8").toString("base64url");
+
+describe("registry wiring", () => {
+  it("exposes videoConfig + video serviceKind for openrouter and vertex", () => {
+    expect(getVideoConfig("openrouter").baseUrl).toBe("https://openrouter.ai/api/v1/videos");
+    expect(getVideoConfig("vertex").baseUrl).toBe("https://aiplatform.googleapis.com");
+    expect(PROVIDER_MEDIA.openrouter.serviceKinds).toContain("video");
+    expect(PROVIDER_MEDIA.vertex.serviceKinds).toContain("video");
+  });
+
+  it("registers video-kind models on both providers", () => {
+    const or = PROVIDER_MODELS.openrouter.find((m) => m.id === "google/veo-3.1");
+    const vx = PROVIDER_MODELS.vertex.find((m) => m.id === "veo-3.1-generate-preview");
+    expect(or?.kind).toBe("video");
+    expect(vx?.kind).toBe("video");
+  });
+});
+
+describe("openrouter video adapter", () => {
+  beforeEach(() => { global.fetch = vi.fn(); });
+  afterEach(() => { global.fetch = originalFetch; });
+
+  it("POSTs creation to the collection root (no /generations suffix)", async () => {
+    global.fetch.mockResolvedValueOnce(jsonResponse({ id: "job-1", status: "pending" }));
+
+    const raw = '{"model":"google/veo-3.1","prompt":"a paper boat"}';
+    const result = await handleVideoProxyCore({
+      provider: "openrouter",
+      action: "generations",
+      rawBody: raw,
+      contentType: "application/json",
+      credentials: { apiKey: "sk-or-key" },
+    });
+
+    expect(result.success).toBe(true);
+    const [url, init] = global.fetch.mock.calls[0];
+    expect(url).toBe("https://openrouter.ai/api/v1/videos");
+    expect(init.method).toBe("POST");
+    expect(init.body).toBe(raw); // verbatim
+    expect(init.headers.Authorization).toBe("Bearer sk-or-key");
+    expect(init.headers["HTTP-Referer"]).toBe("https://endpoint-proxy.local");
+    expect(await result.response.json()).toEqual({ id: "job-1", status: "pending" });
+  });
+
+  it("polls GET /videos/{id} and passes the payload through verbatim", async () => {
+    const payload = { id: "job-1", status: "completed", unsigned_urls: ["https://cdn/v.mp4"] };
+    global.fetch.mockResolvedValueOnce(jsonResponse(payload));
+
+    const result = await handleVideoProxyCore({
+      provider: "openrouter",
+      requestId: "job-1",
+      credentials: { apiKey: "sk-or-key" },
+    });
+
+    const [url, init] = global.fetch.mock.calls[0];
+    expect(url).toBe("https://openrouter.ai/api/v1/videos/job-1");
+    expect(init.method).toBe("GET");
+    expect(await result.response.json()).toEqual(payload);
+  });
+
+  it("rejects unsupported actions before any upstream call (no billable job)", async () => {
+    const result = await handleVideoProxyCore({
+      provider: "openrouter",
+      action: "extensions",
+      rawBody: "{}",
+      contentType: "application/json",
+      credentials: { apiKey: "sk-or-key" },
+    });
+
+    expect(result.success).toBe(false);
+    expect(result.status).toBe(400);
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+});
+
+describe("vertex (veo) video adapter", () => {
+  beforeEach(() => {
+    global.fetch = vi.fn();
+    refreshVertexToken.mockReset();
+  });
+  afterEach(() => { global.fetch = originalFetch; });
+
+  const saJson = JSON.stringify({
+    type: "service_account",
+    client_email: "sa@proj-1.iam.gserviceaccount.com",
+    private_key: "-----BEGIN PRIVATE KEY-----\nx\n-----END PRIVATE KEY-----\n",
+    project_id: "proj-1",
+  });
+
+  it("translates the create body to predictLongRunning and returns a poll-able job id", async () => {
+    refreshVertexToken.mockResolvedValueOnce({ accessToken: "vertex-tok" });
+    global.fetch.mockResolvedValueOnce(jsonResponse({ name: OPERATION_NAME }));
+
+    const result = await handleVideoProxyCore({
+      provider: "vertex",
+      action: "generations",
+      rawBody: JSON.stringify({
+        model: "veo-3.1-generate-preview",
+        prompt: "a neon city",
+        duration: 8,
+        aspect_ratio: "16:9",
+        resolution: "720p",
+        n: 1,
+      }),
+      contentType: "application/json",
+      credentials: { apiKey: saJson },
+    });
+
+    expect(result.success).toBe(true);
+    const [url, init] = global.fetch.mock.calls[0];
+    expect(url).toBe(
+      "https://aiplatform.googleapis.com/v1/projects/proj-1/locations/us-central1/publishers/google/models/veo-3.1-generate-preview:predictLongRunning"
+    );
+    expect(init.headers.Authorization).toBe("Bearer vertex-tok");
+    expect(JSON.parse(init.body)).toEqual({
+      instances: [{ prompt: "a neon city" }],
+      parameters: { sampleCount: 1, durationSeconds: 8, aspectRatio: "16:9", resolution: "720p" },
+    });
+
+    // Response is mapped onto the async-job shape clients already poll.
+    expect(await result.response.json()).toEqual({
+      id: JOB_ID,
+      request_id: JOB_ID,
+      status: "pending",
+    });
+  });
+
+  it("maps a data-URL image onto the Vertex image instance", async () => {
+    refreshVertexToken.mockResolvedValueOnce({ accessToken: "vertex-tok" });
+    global.fetch.mockResolvedValueOnce(jsonResponse({ name: OPERATION_NAME }));
+
+    await handleVideoProxyCore({
+      provider: "vertex",
+      action: "generations",
+      rawBody: JSON.stringify({
+        model: "veo-3.1-generate-preview",
+        prompt: "animate this",
+        image: "data:image/png;base64,AAAB",
+      }),
+      contentType: "application/json",
+      credentials: { apiKey: saJson },
+    });
+
+    expect(JSON.parse(global.fetch.mock.calls[0][1].body).instances[0].image).toEqual({
+      bytesBase64Encoded: "AAAB",
+      mimeType: "image/png",
+    });
+  });
+
+  it("polls via fetchPredictOperation and maps a completed operation", async () => {
+    refreshVertexToken.mockResolvedValueOnce({ accessToken: "vertex-tok" });
+    global.fetch.mockResolvedValueOnce(
+      jsonResponse({
+        name: OPERATION_NAME,
+        done: true,
+        response: { videos: [{ gcsUri: "gs://bucket/v.mp4", mimeType: "video/mp4" }] },
+      })
+    );
+
+    const result = await handleVideoProxyCore({
+      provider: "vertex",
+      requestId: JOB_ID,
+      credentials: { apiKey: saJson },
+    });
+
+    const [url, init] = global.fetch.mock.calls[0];
+    expect(url).toBe(
+      "https://aiplatform.googleapis.com/v1/projects/proj-1/locations/us-central1/publishers/google/models/veo-3.1-generate-preview:fetchPredictOperation"
+    );
+    expect(init.method).toBe("POST"); // Vertex polls with POST, not GET
+    expect(JSON.parse(init.body)).toEqual({ operationName: OPERATION_NAME });
+
+    expect(await result.response.json()).toEqual({
+      id: JOB_ID,
+      request_id: JOB_ID,
+      status: "completed",
+      video: { url: "gs://bucket/v.mp4", b64_json: null, mime_type: "video/mp4" },
+      videos: [{ url: "gs://bucket/v.mp4", b64_json: null, mime_type: "video/mp4" }],
+    });
+  });
+
+  it("maps a failed operation to status failed", async () => {
+    refreshVertexToken.mockResolvedValueOnce({ accessToken: "vertex-tok" });
+    global.fetch.mockResolvedValueOnce(
+      jsonResponse({ name: OPERATION_NAME, done: true, error: { code: 3, message: "bad prompt" } })
+    );
+
+    const result = await handleVideoProxyCore({
+      provider: "vertex",
+      requestId: JOB_ID,
+      credentials: { apiKey: saJson },
+    });
+
+    const body = await result.response.json();
+    expect(body.status).toBe("failed");
+    expect(body.error.message).toBe("bad prompt");
+  });
+
+  it("rejects missing project id and raw API keys before any upstream call", async () => {
+    const noProject = await handleVideoProxyCore({
+      provider: "vertex",
+      action: "generations",
+      rawBody: JSON.stringify({ model: "veo-3.1-generate-preview", prompt: "x" }),
+      contentType: "application/json",
+      credentials: { apiKey: "AIzaRawKey" },
+    });
+    expect(noProject.success).toBe(false);
+    expect(noProject.status).toBe(400);
+
+    const noToken = await handleVideoProxyCore({
+      provider: "vertex",
+      action: "generations",
+      rawBody: JSON.stringify({ model: "veo-3.1-generate-preview", prompt: "x" }),
+      contentType: "application/json",
+      credentials: { apiKey: "AIzaRawKey", providerSpecificData: { projectId: "proj-1" } },
+    });
+    expect(noToken.success).toBe(false);
+    expect(noToken.status).toBe(400);
+
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+
+  it("rejects an invalid job id without calling upstream", async () => {
+    refreshVertexToken.mockResolvedValue({ accessToken: "vertex-tok" });
+    const result = await handleVideoProxyCore({
+      provider: "vertex",
+      requestId: Buffer.from("not-an-operation", "utf8").toString("base64url"),
+      credentials: { apiKey: saJson },
+    });
+    expect(result.success).toBe(false);
+    expect(result.status).toBe(400);
+    expect(global.fetch).not.toHaveBeenCalled();
+  });
+});
--- a/tests/unit/xai-video-handler.test.js
+++ b/tests/unit/xai-video-handler.test.js
@@ -29,6 +29,7 @@ vi.mock("@/sse/services/auth.js", () => authMocks);
 vi.mock("@/sse/services/tokenRefresh.js", () => tokenMocks);
 vi.mock("@/lib/localDb", () => ({
  getSettings: vi.fn(async () => ({ requireApiKey: false })),
+  getProviderConnectionById: vi.fn(async () => ({ id: "conn-5", provider: "xai" })),
  getComboByName: vi.fn(async () => null),
  getModelAliases: vi.fn(async () => ({})),
  getProviderNodes: vi.fn(async () => []),