feat(providers): add ofox.ai as LLM preset and video generation provider

.env.example

@@ -97,6 +97,11 @@ SEEDANCE_BASE_URL=https://ark.cn-beijing.volces.com/api/v3
 SEEDANCE_API_KEY=
 SEEDANCE_VIDEO_MODEL=doubao-seedance-2-0-260128
 
+# OFox multi-model gateway (LLM preset + video generation share one key)
+LLM_OFOX_BASE_URL=https://api.ofox.ai/v1
+LLM_OFOX_API_KEY=
+OFOX_VIDEO_MODEL=bytedance/seedance-2.0-fast
+
 KLING_BASE_URL=https://api-singapore.klingai.com
 KLING_API_KEY=
 KLING_VIDEO_MODEL=kling-v3-omni

README.md

@@ -402,7 +402,7 @@ Clients must send `Authorization: Bearer <token>`. The current bridge is designe
 |---|---|
 | Frontend | React 19, TypeScript 6, Vite 8 |
 | Editing core | Immutable timeline state, command layer, and proposal-based application |
-| Agent | Vercel AI SDK 7 (Anthropic, OpenAI, Gemini, Kimi, Qwen, GLM, DeepSeek, MiniMax, Xiaomi MiMo, Mistral, xAI Grok by API key or SuperGrok/X Premium+ subscription sign-in, and compatible APIs), Agent Skills, MCP SDK |
+| Agent | Vercel AI SDK 7 (Anthropic, OpenAI, Gemini, Kimi, Qwen, GLM, DeepSeek, MiniMax, Xiaomi MiMo, Mistral, xAI Grok by API key or SuperGrok/X Premium+ subscription sign-in, OFox, and compatible APIs), Agent Skills, MCP SDK |
 | Preview and visuals | Remotion Player, WebGL / GLSL |
 | Server | Dual-host Vite / Electron plugins and a server-side keystore |
 | Persistence | Shared local project store under `~/.openchatcut`, IndexedDB cache, configurable local media directory, optional Cloudflare R2 |

README_ZH.md

@@ -397,7 +397,7 @@ OPENCHATCUT_EDITOR_URL=https://your-editor.example.com
 |---|---|
 | 前端 | React 19、TypeScript 6、Vite 8 |
 | 编辑核心 | 不可变时间线状态、命令层、提案式应用 |
-| Agent | Vercel AI SDK 7（Anthropic、OpenAI、Gemini、Kimi、Qwen、GLM、DeepSeek、MiniMax、小米 MiMo、Mistral、xAI Grok（API Key 或 SuperGrok/X Premium+ 订阅登录）与兼容接口）、Agent Skills、MCP SDK |
+| Agent | Vercel AI SDK 7（Anthropic、OpenAI、Gemini、Kimi、Qwen、GLM、DeepSeek、MiniMax、小米 MiMo、Mistral、xAI Grok（API Key 或 SuperGrok/X Premium+ 订阅登录）、OFox 与兼容接口）、Agent Skills、MCP SDK |
 | 预览与视觉 | Remotion Player、WebGL / GLSL |
 | 服务端 | Vite / Electron 双宿主插件、服务端密钥仓 |
 | 持久化 | `~/.openchatcut` 下的本机共享工程库、IndexedDB 缓存、可配置本地素材目录、可选 Cloudflare R2 |

assets/agent/openchatcut-tool-schemas.json

@@ -1922,9 +1922,10 @@
               "kling",
               "hailuo",
               "byteplus",
-              "grok-imagine-video"
+              "grok-imagine-video",
+              "ofox"
             ],
-            "description": "hailuo is MiniMax: 6 or 10s; firstFrame optional; lastFrame allowed with firstFrame; no multi-ref or multi-shot. 1080p is 6s only. byteplus is BytePlus ModelArk Seedance — same request shape/limits as seedance2. grok-imagine-video is xAI Grok Imagine: text-to-video only, 1–15s, audio track included, no references/frames."
+            "description": "hailuo is MiniMax: 6 or 10s; firstFrame optional; lastFrame allowed with firstFrame; no multi-ref or multi-shot. 1080p is 6s only. byteplus is BytePlus ModelArk Seedance — same request shape/limits as seedance2. grok-imagine-video is xAI Grok Imagine: text-to-video only, 1–15s, audio track included, no references/frames. ofox is the OFox multi-model gateway (Seedance/Wan and more behind one key): 2–30s with per-model limits enforced by the API; supports firstFrame (and optional lastFrame), or up to 9 refImages (frames and refImages are mutually exclusive); no refVideos/refAudios yet."
           },
           "prompt": {
             "type": "string",
@@ -1942,11 +1943,11 @@
                 "type": "string"
               }
             ],
-            "description": "Integer seconds, 2–15 for Seedance, 3–15 for Kling, exactly 6 or 10 for Hailuo (Hailuo 1080p → 6 only), 1–15 for grok-imagine-video."
+            "description": "Integer seconds, 2–15 for Seedance, 3–15 for Kling, exactly 6 or 10 for Hailuo (Hailuo 1080p → 6 only), 1–15 for grok-imagine-video, 2–30 for ofox (per-model limits enforced by the API)."
           },
           "ratio": {
             "type": "string",
-            "description": "Seedance: 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive. Kling: 16:9, 9:16, 1:1. grok-imagine-video: 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3. Do not send for hailuo."
+            "description": "Seedance: 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive. Kling: 16:9, 9:16, 1:1. grok-imagine-video: 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3. ofox: 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21. Do not send for hailuo."
           },
           "resolution": {
             "type": "string",
@@ -1957,7 +1958,7 @@
               "1080p",
               "4k"
             ],
-            "description": "Seedance: 480p/720p(default)/1080p/4k. Hailuo: 512p (Hailuo-02), 720p→API 768P, 1080p (6s only). Kling: pair with mode std/pro. grok-imagine-video: 480p(default)/720p/1080p."
+            "description": "Seedance: 480p/720p(default)/1080p/4k. Hailuo: 512p (Hailuo-02), 720p→API 768P, 1080p (6s only). Kling: pair with mode std/pro. grok-imagine-video: 480p(default)/720p/1080p. ofox: 480p/720p(default)/1080p, per-model support enforced by the API."
           },
           "mode": {
             "type": "string",
@@ -1973,7 +1974,7 @@
           },
           "lastFrame": {
             "type": "string",
-            "description": "Project image asset reference; requires firstFrame. Supported on seedance2, kling, and hailuo (not with multi-ref on seedance2)."
+            "description": "Project image asset reference; requires firstFrame. Supported on seedance2, kling, hailuo, and ofox (not with multi-ref on seedance2/ofox)."
           },
           "refImages": {
             "type": "array",

server/key-probes.ts

@@ -383,6 +383,13 @@ export const PROBES: Record<string, ProbeDef> = {
   'video/hailuo': minimaxProbe,
   'video/byteplus': byteplusProbe,
   'video/xai': xaiMediaProbe,
+  'video/ofox': {
+    needs: [['LLM_OFOX_API_KEY']],
+    run: (get) => fetch(`${base(get, 'LLM_OFOX_BASE_URL', 'https://api.ofox.ai/v1')}/models`, {
+      signal: t(), headers: bearer(get('LLM_OFOX_API_KEY')),
+    }),
+    models: parseModelCatalog,
+  },
   'music/mureka': {
     needs: [['MUREKA_API_KEY']],
     run: (get) => fetch(`${base(get, 'MUREKA_BASE_URL', 'https://api.mureka.ai')}/v1/account/billing`, {

server/key-probes.verify.ts

@@ -28,7 +28,7 @@ const EXPECTED_PAGES = [
   'image/openai', 'image/gemini', 'image/minimax', 'image/wavespeed', 'image/byteplus', 'image/xai',
   'voice/elevenlabs', 'voice/openai', 'voice/gemini', 'voice/mistral', 'voice/cartesia',
   'voice/doubao', 'voice/minimax', 'voice/inworld', 'voice/fishaudio', 'voice/speechify',
-  'video/seedance', 'video/kling', 'video/hailuo', 'video/byteplus', 'video/xai',
+  'video/seedance', 'video/kling', 'video/hailuo', 'video/byteplus', 'video/xai', 'video/ofox',
   'music/mureka', 'music/minimax', 'music/atlas', 'music/sonilo',
   'stock/pexels', 'stock/pixabay', 'stock/unsplash', 'stock/freesound',
   'transcription/assemblyai', 'transcription/openai', 'transcription/mistral',

server/keystore.ts

@@ -82,6 +82,10 @@ export const KEY_NAMES = [
   "LLM_OPENROUTER_API_KEY",
   "LLM_OPENROUTER_BASE_URL",
   "LLM_OPENROUTER_MODEL",
+  "LLM_OFOX_API_KEY",
+  "LLM_OFOX_BASE_URL",
+  "LLM_OFOX_MODEL",
+  "OFOX_VIDEO_MODEL",
   "LLM_OLLAMA_API_KEY",
   "LLM_OLLAMA_BASE_URL",
   "LLM_OLLAMA_MODEL",
@@ -237,6 +241,7 @@ export const NON_SECRET_NAMES: ReadonlySet<string> = new Set([
   "BYTEPLUS_VIDEO_MODEL",
   "XAI_IMAGE_MODEL",
   "XAI_VIDEO_MODEL",
+  "OFOX_VIDEO_MODEL",
   "INWORLD_TTS_MODEL",
   "FISHAUDIO_TTS_MODEL",
   "SPEECHIFY_TTS_MODEL",

server/keystore.verify.ts

@@ -130,7 +130,7 @@ const MODEL_ROUTING_NAMES = [
   'MINIMAX_TTS_MODEL', 'MINIMAX_VIDEO_MODEL', 'MINIMAX_MUSIC_MODEL', 'MINIMAX_IMAGE_MODEL',
   'ATLASCLOUD_API_BASE', 'ATLASCLOUD_MUSIC_MODEL',
   'WAVESPEED_IMAGE_MODEL', 'BYTEPLUS_IMAGE_MODEL', 'BYTEPLUS_VIDEO_MODEL',
-  'XAI_IMAGE_MODEL', 'XAI_VIDEO_MODEL',
+  'XAI_IMAGE_MODEL', 'XAI_VIDEO_MODEL', 'OFOX_VIDEO_MODEL',
   'INWORLD_TTS_MODEL', 'FISHAUDIO_TTS_MODEL', 'SPEECHIFY_TTS_MODEL',
   'PREFERRED_IMAGE_VENDOR', 'PREFERRED_VOICE_VENDOR', 'PREFERRED_VIDEO_VENDOR', 'PREFERRED_MUSIC_VENDOR',
   'PREFERRED_TRANSCRIPTION_PROVIDER', 'TRANSCRIPTION_LANGUAGE', 'TRANSCRIPTION_DIARIZATION', 'AUTO_TRANSCRIBE_INGEST', 'UI_SCALE',

server/plugins/index.ts

@@ -310,6 +310,15 @@ export function serverPlugins(options: { projectStoreHttp?: boolean } = {}): Plu
       get xaiVideoModel() {
         return getKey("XAI_VIDEO_MODEL") || "grok-imagine-video-1.5";
       },
+      get ofoxBaseUrl() {
+        return getKey("LLM_OFOX_BASE_URL") || "https://api.ofox.ai/v1";
+      },
+      get ofoxApiKey() {
+        return getKey("LLM_OFOX_API_KEY");
+      },
+      get ofoxVideoModel() {
+        return getKey("OFOX_VIDEO_MODEL") || "bytedance/seedance-2.0-fast";
+      },
       get byteplusModel() {
         return getKey("BYTEPLUS_VIDEO_MODEL") || "seedance-1-5-pro-251215";
       },

server/plugins/ofox-video-provider.ts

@@ -0,0 +1,131 @@
+import { proxyDispatcher } from '../outbound-proxy.ts';
+import type { RegisterGenerationProviderTask } from './generation-jobs.ts';
+import { mediaDataUrl } from './video-media.ts';
+import type { ValidVideoRequest } from './video-validation.ts';
+
+type FetchInit = Parameters<typeof fetch>[1] & { dispatcher?: unknown };
+const fetchWithProxy = (url: RequestInfo | URL, init?: FetchInit): Promise<Response> =>
+  fetch(url, { ...init, dispatcher: proxyDispatcher() } as RequestInit);
+const FAILURES = new Set(['failed', 'error', 'cancelled', 'canceled', 'expired']);
+const SUCCESSES = new Set(['completed', 'succeeded']);
+const wait = (milliseconds: number) => new Promise((resolve) => setTimeout(resolve, milliseconds));
+
+interface OfoxVideoOptions {
+  ofoxBaseUrl: string;
+  ofoxApiKey: string;
+  ofoxVideoModel: string;
+}
+
+async function providerError(response: Response): Promise<string> {
+  const text = await response.text();
+  try {
+    const data = JSON.parse(text) as { message?: string; error?: { message?: string } };
+    return data.error?.message ?? data.message ?? `video provider failed (${response.status})`;
+  } catch {
+    return text.slice(0, 300) || `video provider failed (${response.status})`;
+  }
+}
+
+interface OfoxMediaRef {
+  type: 'image_url';
+  image_url: { url: string };
+  frame_type?: 'first_frame' | 'last_frame';
+}
+
+/** Frame anchors and image references both ride as data URLs; the OFox
+ * gateway re-hosts data URIs on its own object storage before handing the
+ * request to the upstream model (verified live), so local project media
+ * needs no public hosting. frame_images and input_references are mutually
+ * exclusive at the API level; validation enforces that before submission. */
+async function ofoxMediaInputs(input: ValidVideoRequest): Promise<{
+  frame_images?: OfoxMediaRef[];
+  input_references?: OfoxMediaRef[];
+}> {
+  if (input.firstFramePath) {
+    const frames: OfoxMediaRef[] = [{
+      type: 'image_url',
+      image_url: { url: await mediaDataUrl(input.firstFramePath) },
+      frame_type: 'first_frame',
+    }];
+    if (input.lastFramePath) {
+      frames.push({
+        type: 'image_url',
+        image_url: { url: await mediaDataUrl(input.lastFramePath) },
+        frame_type: 'last_frame',
+      });
+    }
+    return { frame_images: frames };
+  }
+  if (input.refImagePaths.length) {
+    return {
+      input_references: await Promise.all(input.refImagePaths.map(async (path) => ({
+        type: 'image_url' as const,
+        image_url: { url: await mediaDataUrl(path) },
+      }))),
+    };
+  }
+  return {};
+}
+
+/** OFox multi-model video gateway: asynchronous task id plus polling.
+ * POST /videos -> { id } ; GET /videos/{id} -> { status, unsigned_urls }.
+ * Invalid duration/resolution/model values are rejected by the API with a
+ * clear 400 before any task is created, so no per-model limits are kept here. */
+export async function generateOfoxVideo(
+  input: ValidVideoRequest,
+  options: OfoxVideoOptions,
+  registerProviderTask: RegisterGenerationProviderTask,
+  existingTaskId?: string,
+): Promise<string> {
+  const baseUrl = options.ofoxBaseUrl.replace(/\/$/, '');
+  const headers = { 'Content-Type': 'application/json', Authorization: `Bearer ${options.ofoxApiKey}` };
+  let taskId = existingTaskId;
+  if (!taskId) {
+    const startedResponse = await fetchWithProxy(`${baseUrl}/videos`, {
+      method: 'POST', headers,
+      body: JSON.stringify({
+        model: options.ofoxVideoModel, prompt: input.prompt,
+        duration: input.durationSeconds, aspect_ratio: input.ratio,
+        resolution: input.resolution ?? '720p',
+        ...(input.generateAudio !== undefined ? { generate_audio: input.generateAudio } : {}),
+        ...(input.seed !== undefined ? { seed: input.seed } : {}),
+        ...(await ofoxMediaInputs(input)),
+      }),
+      // Data-URL payloads carry whole reference images; give the submit
+      // request a proportionally larger timeout than the text-only case.
+      signal: AbortSignal.timeout(120_000),
+    });
+    if (!startedResponse.ok) throw new Error(await providerError(startedResponse));
+    const started = await startedResponse.json() as { id?: unknown };
+    taskId = String(started.id ?? '');
+    if (!taskId) throw new Error('ofox did not return a task id');
+    await registerProviderTask('ofox', taskId);
+  }
+  const deadline = Date.now() + 15 * 60_000;
+  while (Date.now() < deadline) {
+    const poll = await fetchWithProxy(`${baseUrl}/videos/${encodeURIComponent(taskId)}`, {
+      headers, signal: AbortSignal.timeout(20_000),
+    });
+    if (!poll.ok) throw new Error(await providerError(poll));
+    const current = await poll.json() as { status?: unknown; unsigned_urls?: unknown; error?: { message?: unknown } };
+    const status = String(current.status ?? '');
+    if (SUCCESSES.has(status)) {
+      // Prefer mirror_urls (persistent signed CDN addresses, present only when
+      // the upstream has mirroring enabled) and fall back to unsigned_urls
+      // (temporary upstream links that may expire within 24 hours).
+      const body = current as { mirror_urls?: unknown; unsigned_urls?: unknown };
+      for (const field of ['mirror_urls', 'unsigned_urls'] as const) {
+        const urls = Array.isArray(body[field]) ? body[field] as unknown[] : [];
+        const url = urls.find((item): item is string => typeof item === 'string' && /^https?:\/\//.test(item));
+        if (url) return url;
+      }
+      throw new Error('ofox generation succeeded without a video URL');
+    }
+    if (FAILURES.has(status)) {
+      const detail = typeof current.error?.message === 'string' ? `: ${current.error.message}` : '';
+      throw new Error(`ofox generation ${status}${detail}`);
+    }
+    await wait(3_000);
+  }
+  throw new Error('ofox generation timed out');
+}

server/plugins/video-validation.ts

@@ -4,7 +4,7 @@ export type KlingVideoReferType = 'feature' | 'base';
 
 export interface VideoRequest {
   operationId?: string;
-  model?: 'seedance2' | 'kling' | 'hailuo' | 'byteplus' | 'grok-imagine-video';
+  model?: 'seedance2' | 'kling' | 'hailuo' | 'byteplus' | 'grok-imagine-video' | 'ofox';
   prompt?: string;
   name?: string;
   durationSeconds?: number | string;
@@ -34,7 +34,7 @@ export interface VideoRequest {
 }
 
 export interface ValidVideoRequest extends Omit<VideoRequest, 'model' | 'prompt' | 'durationSeconds' | 'ratio' | 'refImagePaths' | 'refVideoPaths' | 'refAudioPaths'> {
-  model: 'seedance2' | 'kling' | 'hailuo' | 'byteplus' | 'grok-imagine-video';
+  model: 'seedance2' | 'kling' | 'hailuo' | 'byteplus' | 'grok-imagine-video' | 'ofox';
   prompt: string;
   durationSeconds: number;
   durationSpecified: boolean;
@@ -192,14 +192,52 @@ function validateGrok(input: ValidVideoRequest): ValidVideoRequest {
   return input;
 }
 
+const OFOX_UNSUPPORTED_ARK_KEYS = ['cameraFixed', 'watermark', 'returnLastFrame', 'executionExpiresAfter', 'priority'] as const;
+
+/** OFox multi-model video gateway: the API validates duration, resolution and
+ * vendor per model with a clear 400 before any task is created, so only
+ * cross-provider limits and unsupported options are enforced here. Text,
+ * first/last-frame and image-reference modes are wired; frame anchors and
+ * references are mutually exclusive at the API level (400 references_conflict),
+ * enforced locally before any paid submission. */
+function validateOfox(input: ValidVideoRequest): ValidVideoRequest {
+  if (!input.prompt || input.prompt.length > 4000) throw new Error('ofox prompt is required and must be at most 4000 characters');
+  if (input.durationSeconds < 2 || input.durationSeconds > 30) throw new Error('ofox durationSeconds must be between 2 and 30 (per-model limits are enforced by the API)');
+  if (!['16:9', '9:16', '1:1', '4:3', '3:4', '3:2', '2:3', '21:9', '9:21'].includes(input.ratio)) {
+    throw new Error(`ofox does not support ratio ${input.ratio}`);
+  }
+  if (input.resolution && !['480p', '720p', '1080p'].includes(input.resolution)) throw new Error('ofox resolution must be 480p, 720p, or 1080p (per-model support is enforced by the API)');
+  if (input.lastFramePath && !input.firstFramePath) throw new Error('lastFrame requires firstFrame');
+  if ((input.firstFramePath || input.lastFramePath) && input.refImagePaths.length) {
+    throw new Error('ofox frame anchors (firstFrame/lastFrame) cannot be combined with refImages');
+  }
+  if (input.refImagePaths.length > 9) throw new Error('ofox accepts at most 9 refImages');
+  if (input.refVideoPaths.length || input.refAudioPaths.length) {
+    throw new Error('ofox refVideos/refAudios are not wired in this integration yet; use refImages or firstFrame/lastFrame');
+  }
+  if (input.mode || input.shotType || input.multiPrompts?.length || input.refVideoMode) {
+    throw new Error('multi-shot and editing options are not supported by ofox');
+  }
+  for (const key of OFOX_UNSUPPORTED_ARK_KEYS) {
+    if (input[key] !== undefined) throw new Error(`${key} is supported by seedance2/byteplus only`);
+  }
+  if (input.generateAudio !== undefined && typeof input.generateAudio !== 'boolean') throw new Error('generateAudio must be a boolean');
+  if (input.seed !== undefined && !Number.isSafeInteger(input.seed)) throw new Error('seed must be a safe integer');
+  if (input.promptOptimizer !== undefined || input.fastPretreatment !== undefined) {
+    throw new Error('promptOptimizer/fastPretreatment are supported by hailuo only');
+  }
+  return input;
+}
+
 export function validateVideoRequest(input: VideoRequest): ValidVideoRequest {
-  if (input.model !== 'seedance2' && input.model !== 'kling' && input.model !== 'hailuo' && input.model !== 'byteplus' && input.model !== 'grok-imagine-video') {
-    throw new Error('model must be seedance2, kling, hailuo, byteplus, or grok-imagine-video');
+  if (input.model !== 'seedance2' && input.model !== 'kling' && input.model !== 'hailuo' && input.model !== 'byteplus' && input.model !== 'grok-imagine-video' && input.model !== 'ofox') {
+    throw new Error('model must be seedance2, kling, hailuo, byteplus, grok-imagine-video, or ofox');
   }
   if (input.model === 'hailuo' && input.ratio !== undefined) throw new Error('hailuo does not accept ratio; framing follows the first frame when present');
   const normalized = common(input, input.model);
   if (normalized.model === 'hailuo') return validateHailuo(normalized);
   if (normalized.model === 'kling') return validateKling(normalized);
   if (normalized.model === 'grok-imagine-video') return validateGrok(normalized);
+  if (normalized.model === 'ofox') return validateOfox(normalized);
   return validateSeedance(normalized);
 }