noosphere 0.7.0 → 0.9.0

package/dist/index.cjs

@@ -21,11 +21,13 @@ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: tru
 var index_exports = {};
 __export(index_exports, {
   AudioCraftProvider: () => AudioCraftProvider,
+  GoogleMediaProvider: () => GoogleMediaProvider,
   HfLocalProvider: () => HfLocalProvider,
   Noosphere: () => Noosphere,
   NoosphereError: () => NoosphereError,
   OllamaProvider: () => OllamaProvider,
   OpenAICompatProvider: () => OpenAICompatProvider,
+  OpenAIMediaProvider: () => OpenAIMediaProvider,
   PROVIDER_IDS: () => PROVIDER_IDS,
   PROVIDER_LOGOS: () => PROVIDER_LOGOS,
   WhisperLocalProvider: () => WhisperLocalProvider,
@@ -342,20 +344,22 @@ var Registry = class {
     const cached = this.modelCache.get(provider);
     return cached?.models.find((m) => m.id === modelId) ?? null;
   }
-  async syncProvider(providerId) {
+  async syncProvider(providerId, modality) {
     const provider = this.providers.get(providerId);
     if (!provider) return 0;
-    const models = await provider.listModels();
+    if (modality && !provider.modalities.includes(modality)) return 0;
+    const models = await provider.listModels(modality);
     this.modelCache.set(providerId, { models, syncedAt: Date.now() });
     return models.length;
   }
-  async syncAll() {
+  async syncAll(modality) {
     const byProvider = {};
     const errors = [];
     let synced = 0;
     for (const provider of this.providers.values()) {
+      if (modality && !provider.modalities.includes(modality)) continue;
       try {
-        const count = await this.syncProvider(provider.id);
+        const count = await this.syncProvider(provider.id, modality);
         byProvider[provider.id] = count;
         synced += count;
       } catch (err) {
@@ -1140,6 +1144,7 @@ var ComfyUIProvider = class {
                 id: `civitai-${item.id}`,
                 provider: "comfyui",
                 name: item.name ?? `CivitAI Model ${item.id}`,
+                description: item.description ? item.description.replace(/<[^>]+>/g, "").trim().slice(0, 300) || void 0 : void 0,
                 modality: "image",
                 local: true,
                 cost: { price: 0, unit: "free" },
@@ -1578,6 +1583,28 @@ async function fetchJson(url, options) {
     clearTimeout(timer);
   }
 }
+async function fetchOllamaDescriptions() {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), 5e3);
+  try {
+    const res = await fetch("https://ollama.com/library", { signal: controller.signal });
+    if (!res.ok) throw new Error(`HTTP ${res.status}`);
+    const html = await res.text();
+    const descriptions = /* @__PURE__ */ new Map();
+    const cardRegex = /href="\/library\/([^"]+)"[\s\S]*?<p[^>]*>([\s\S]*?)<\/p>/g;
+    let match;
+    while ((match = cardRegex.exec(html)) !== null) {
+      const modelName = match[1].trim();
+      const desc = match[2].replace(/<[^>]*>/g, "").replace(/&amp;/g, "&").replace(/&lt;/g, "<").replace(/&gt;/g, ">").replace(/&quot;/g, '"').replace(/&#39;/g, "'").replace(/\s+/g, " ").trim();
+      if (modelName && desc) {
+        descriptions.set(modelName, desc);
+      }
+    }
+    return descriptions;
+  } finally {
+    clearTimeout(timer);
+  }
+}
 var OllamaProvider = class {
   id = "ollama";
   name = "Ollama (Local)";
@@ -1601,10 +1628,11 @@ var OllamaProvider = class {
   }
   async listModels(_modality) {
     if (_modality && _modality !== "llm") return [];
-    const [localData, catalogData, runningData] = await Promise.all([
+    const [localData, catalogData, runningData, descriptions] = await Promise.all([
       fetchJson(`${this.baseUrl}/api/tags`, { timeoutMs: 5e3 }).catch(() => null),
       fetchJson("https://ollama.com/api/tags", { timeoutMs: 5e3 }).catch(() => null),
-      fetchJson(`${this.baseUrl}/api/ps`, { timeoutMs: 5e3 }).catch(() => null)
+      fetchJson(`${this.baseUrl}/api/ps`, { timeoutMs: 5e3 }).catch(() => null),
+      fetchOllamaDescriptions().catch(() => /* @__PURE__ */ new Map())
     ]);
     const runningNames = /* @__PURE__ */ new Set();
     if (runningData?.models) {
@@ -1617,27 +1645,30 @@ var OllamaProvider = class {
     if (localData?.models) {
       for (const m of localData.models) {
         const isRunning = runningNames.has(m.name) || runningNames.has(m.model);
-        models.set(m.name, this.toModelInfo(m, isRunning ? "running" : "installed", true));
+        models.set(m.name, this.toModelInfo(m, isRunning ? "running" : "installed", true, descriptions));
       }
     }
     if (catalogData?.models) {
       for (const m of catalogData.models) {
         const name = m.name;
         if (!models.has(name)) {
-          models.set(name, this.toModelInfo(m, "available", false));
+          models.set(name, this.toModelInfo(m, "available", false, descriptions));
         }
       }
     }
     return Array.from(models.values());
   }
-  toModelInfo(m, status, isLocal) {
+  toModelInfo(m, status, isLocal, descriptions) {
     const name = m.name ?? m.model ?? "unknown";
     const family = m.details?.family;
     const logoProvider = inferLogoProvider(name, family);
+    const baseName = name.split(":")[0];
+    const description = descriptions?.get(baseName);
     return {
       id: name,
       provider: "ollama",
       name,
+      ...description ? { description } : {},
       modality: "llm",
       local: true,
       cost: { price: 0, unit: "free" },
@@ -1846,16 +1877,79 @@ var OllamaProvider = class {
   }
 };
 
+// src/utils/parse-readme.ts
+async function fetchReadmeDescription(modelId, timeoutMs = 5e3) {
+  try {
+    const controller = new AbortController();
+    const timer = setTimeout(() => controller.abort(), timeoutMs);
+    try {
+      const res = await fetch(
+        `https://huggingface.co/${modelId}/raw/main/README.md`,
+        { signal: controller.signal }
+      );
+      if (!res.ok) return void 0;
+      const text = await res.text();
+      return parseReadmeDescription(text);
+    } finally {
+      clearTimeout(timer);
+    }
+  } catch {
+    return void 0;
+  }
+}
+function parseReadmeDescription(readme) {
+  const withoutFrontmatter = readme.replace(/^---[\s\S]*?---\s*/, "");
+  const lines = withoutFrontmatter.split("\n");
+  let paragraph = "";
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (!trimmed) {
+      if (paragraph) break;
+      continue;
+    }
+    if (trimmed.startsWith("#")) {
+      if (paragraph) break;
+      continue;
+    }
+    if (/^\[?!\[/.test(trimmed) || /^</.test(trimmed)) continue;
+    if (/^\[.*\]\(.*\)$/.test(trimmed)) continue;
+    paragraph += (paragraph ? " " : "") + trimmed;
+  }
+  if (!paragraph) return void 0;
+  paragraph = paragraph.replace(/\[([^\]]+)\]\([^)]+\)/g, "$1").replace(/\*\*([^*]+)\*\*/g, "$1").replace(/`([^`]+)`/g, "$1").replace(/<[^>]+>/g, "").trim();
+  if (!paragraph) return void 0;
+  if (paragraph.length > 300) paragraph = paragraph.slice(0, 297) + "...";
+  return paragraph;
+}
+
 // src/providers/hf-local.ts
 var import_promises = require("fs/promises");
 var import_node_path = require("path");
 var import_node_os = require("os");
 var FETCH_TIMEOUT_MS3 = 5e3;
 var HF_HUB_API2 = "https://huggingface.co/api/models";
+var HF_ORG_TO_LOGO_PROVIDER = {
+  "meta-llama": "meta",
+  "facebook": "meta",
+  "google": "google",
+  "microsoft": "microsoft",
+  "nvidia": "nvidia",
+  "mistralai": "mistral",
+  "Qwen": "qwen",
+  "deepseek-ai": "deepseek",
+  "openai": "openai",
+  "CohereForAI": "cohere",
+  "rhasspy": "piper",
+  "stabilityai": "huggingface",
+  "black-forest-labs": "huggingface",
+  "tiiuae": "huggingface",
+  "allenai": "huggingface",
+  "Salesforce": "huggingface"
+};
 var PIPELINE_TAG_TO_MODALITY = {
   "text-to-image": "image",
   "text-to-video": "video",
-  "text-to-audio": "tts",
+  "text-to-audio": "music",
   "text-to-speech": "tts",
   "automatic-speech-recognition": "stt"
 };
@@ -1883,7 +1977,7 @@ async function fetchJsonTimeout(url, timeoutMs = FETCH_TIMEOUT_MS3) {
 var HfLocalProvider = class {
   id = "hf-local";
   name = "HuggingFace Local Models";
-  modalities = ["image", "video", "tts", "stt"];
+  modalities = ["image", "video", "tts", "stt", "music"];
   isLocal = true;
   cachedModels = null;
   async ping() {
@@ -1905,8 +1999,7 @@ var HfLocalProvider = class {
   }
   async fetchCatalog() {
     const seen = /* @__PURE__ */ new Set();
-    const models = [];
-    const logo = getProviderLogo("huggingface");
+    const entries = [];
     const results = await Promise.allSettled(
       CATALOG_QUERIES.map(async (q) => {
         const params = new URLSearchParams({
@@ -1924,32 +2017,56 @@ var HfLocalProvider = class {
         const id = entry.id ?? entry.modelId;
         if (!id || seen.has(id)) continue;
         seen.add(id);
-        const pipelineTag = entry.pipeline_tag ?? "";
-        const modality = PIPELINE_TAG_TO_MODALITY[pipelineTag] ?? "image";
-        models.push({
+        entries.push({
           id,
-          provider: "hf-local",
-          name: id.split("/").pop() ?? id,
-          modality,
-          local: true,
-          cost: { price: 0, unit: "free" },
-          logo,
-          status: "available",
-          localInfo: {
-            sizeBytes: 0,
-            runtime: "huggingface",
-            family: entry.library_name
-          },
-          capabilities: {}
+          pipelineTag: entry.pipeline_tag ?? "",
+          libraryName: entry.library_name
         });
       }
     }
+    const descriptionMap = /* @__PURE__ */ new Map();
+    for (let i = 0; i < entries.length; i += 10) {
+      const batch = entries.slice(i, i + 10);
+      const descs = await Promise.allSettled(
+        batch.map(async (e) => {
+          const desc = await fetchReadmeDescription(e.id);
+          return { id: e.id, desc };
+        })
+      );
+      for (const d of descs) {
+        if (d.status === "fulfilled" && d.value.desc) {
+          descriptionMap.set(d.value.id, d.value.desc);
+        }
+      }
+    }
+    const models = [];
+    for (const e of entries) {
+      const modality = PIPELINE_TAG_TO_MODALITY[e.pipelineTag] ?? "image";
+      const org = e.id.includes("/") ? e.id.split("/")[0] : void 0;
+      const logoProvider = org ? HF_ORG_TO_LOGO_PROVIDER[org] ?? "huggingface" : "huggingface";
+      models.push({
+        id: e.id,
+        provider: "hf-local",
+        name: e.id.split("/").pop() ?? e.id,
+        modality,
+        local: true,
+        cost: { price: 0, unit: "free" },
+        logo: getProviderLogo(logoProvider),
+        description: descriptionMap.get(e.id),
+        status: "available",
+        localInfo: {
+          sizeBytes: 0,
+          runtime: "huggingface",
+          family: e.libraryName
+        },
+        capabilities: {}
+      });
+    }
     return models;
   }
   async scanLocalCache() {
     const models = [];
     const cacheDir = (0, import_node_path.join)((0, import_node_os.homedir)(), ".cache", "huggingface", "hub");
-    const logo = getProviderLogo("huggingface");
     try {
       const entries = await (0, import_promises.readdir)(cacheDir, { withFileTypes: true });
       for (const entry of entries) {
@@ -1981,6 +2098,8 @@ var HfLocalProvider = class {
           }
         }
         const modality = PIPELINE_TAG_TO_MODALITY[pipelineTag] ?? "image";
+        const org = modelId.includes("/") ? modelId.split("/")[0] : void 0;
+        const logoProvider = org ? HF_ORG_TO_LOGO_PROVIDER[org] ?? "huggingface" : "huggingface";
         models.push({
           id: modelId,
           provider: "hf-local",
@@ -1988,7 +2107,7 @@ var HfLocalProvider = class {
           modality,
           local: true,
           cost: { price: 0, unit: "free" },
-          logo,
+          logo: getProviderLogo(logoProvider),
           status: "installed",
           localInfo: {
             sizeBytes: 0,
@@ -2009,6 +2128,25 @@ var import_promises2 = require("fs/promises");
 var import_node_path2 = require("path");
 var import_node_os2 = require("os");
 var WHISPER_MODELS = ["tiny", "base", "small", "medium", "large", "large-v2", "large-v3", "turbo"];
+var WHISPER_HF_REPOS = {
+  "tiny": "openai/whisper-tiny",
+  "base": "openai/whisper-base",
+  "small": "openai/whisper-small",
+  "medium": "openai/whisper-medium",
+  "large": "openai/whisper-large",
+  "large-v2": "openai/whisper-large-v2",
+  "large-v3": "openai/whisper-large-v3",
+  "turbo": "openai/whisper-large-v3-turbo"
+};
+async function fetchWhisperDescriptions() {
+  const descriptions = /* @__PURE__ */ new Map();
+  const fetches = Object.entries(WHISPER_HF_REPOS).map(async ([size, repo]) => {
+    const desc = await fetchReadmeDescription(repo, 8e3);
+    if (desc) descriptions.set(size, desc);
+  });
+  await Promise.allSettled(fetches);
+  return descriptions;
+}
 function runPython(code, timeoutMs = 5e3) {
   return new Promise((resolve, reject) => {
     const proc = (0, import_node_child_process.execFile)("python3", ["-c", code], { timeout: timeoutMs }, (err, stdout) => {
@@ -2054,7 +2192,8 @@ var WhisperLocalProvider = class {
     if (_modality && _modality !== "stt") return [];
     const runtime = await this.detectRuntime();
     if (!runtime) return [];
-    const logo = getProviderLogo("huggingface");
+    const descMap = await fetchWhisperDescriptions().catch(() => /* @__PURE__ */ new Map());
+    const logo = getProviderLogo("openai");
     const models = [];
     for (const name of WHISPER_MODELS) {
       const installed = await this.isModelCached(name, runtime);
@@ -2062,6 +2201,7 @@ var WhisperLocalProvider = class {
         id: `whisper-${name}`,
         provider: "whisper-local",
         name: `Whisper ${name}`,
+        description: descMap.get(name),
         modality: "stt",
         local: true,
         cost: { price: 0, unit: "free" },
@@ -2125,6 +2265,22 @@ var AUDIOCRAFT_MODELS = [
   { id: "musicgen-melody", name: "MusicGen Melody" },
   { id: "audiogen-medium", name: "AudioGen Medium" }
 ];
+var AUDIOCRAFT_HF_REPOS = {
+  "musicgen-small": "facebook/musicgen-small",
+  "musicgen-medium": "facebook/musicgen-medium",
+  "musicgen-large": "facebook/musicgen-large",
+  "musicgen-melody": "facebook/musicgen-melody",
+  "audiogen-medium": "facebook/audiogen-medium"
+};
+async function fetchAudioCraftDescriptions() {
+  const descriptions = /* @__PURE__ */ new Map();
+  const fetches = Object.entries(AUDIOCRAFT_HF_REPOS).map(async ([id, repo]) => {
+    const desc = await fetchReadmeDescription(repo, 8e3);
+    if (desc) descriptions.set(id, desc);
+  });
+  await Promise.allSettled(fetches);
+  return descriptions;
+}
 function runPython2(code, timeoutMs = 5e3) {
   return new Promise((resolve, reject) => {
     (0, import_node_child_process2.execFile)("python3", ["-c", code], { timeout: timeoutMs }, (err, stdout) => {
@@ -2160,6 +2316,7 @@ var AudioCraftProvider = class {
   async listModels(_modality) {
     if (_modality && _modality !== "music") return [];
     if (!await this.ping()) return [];
+    const descMap = await fetchAudioCraftDescriptions().catch(() => /* @__PURE__ */ new Map());
     const logo = getProviderLogo("meta");
     const models = [];
     for (const m of AUDIOCRAFT_MODELS) {
@@ -2169,6 +2326,7 @@ var AudioCraftProvider = class {
         id: m.id,
         provider: "audiocraft",
         name: m.name,
+        description: descMap.get(m.id),
         modality: "music",
         local: true,
         cost: { price: 0, unit: "free" },
@@ -2382,6 +2540,501 @@ async function detectOpenAICompatServers() {
   return providers;
 }
 
+// src/providers/openai-media.ts
+var OPENAI_API_BASE = "https://api.openai.com/v1";
+var FETCH_TIMEOUT_MS5 = 8e3;
+var MODEL_PREFIX_MAP = [
+  { prefix: "dall-e-", modality: "image" },
+  { prefix: "gpt-image-", modality: "image" },
+  { prefix: "sora-", modality: "video" },
+  { prefix: "tts-", modality: "tts" },
+  { prefix: "whisper-", modality: "stt" }
+];
+function classifyModel(id) {
+  for (const { prefix, modality } of MODEL_PREFIX_MAP) {
+    if (id.startsWith(prefix)) return modality;
+  }
+  return null;
+}
+var OpenAIMediaProvider = class {
+  constructor(apiKey) {
+    this.apiKey = apiKey;
+  }
+  id = "openai-media";
+  name = "OpenAI (Image, Video, TTS, STT)";
+  modalities = ["image", "video", "tts", "stt"];
+  isLocal = false;
+  modelsCache = null;
+  async ping() {
+    try {
+      const controller = new AbortController();
+      const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS5);
+      try {
+        const res = await fetch(`${OPENAI_API_BASE}/models`, {
+          headers: { Authorization: `Bearer ${this.apiKey}` },
+          signal: controller.signal
+        });
+        return res.ok;
+      } finally {
+        clearTimeout(timer);
+      }
+    } catch {
+      return false;
+    }
+  }
+  async listModels(modality) {
+    if (this.modelsCache) {
+      return modality ? this.modelsCache.filter((m) => m.modality === modality) : this.modelsCache;
+    }
+    try {
+      const controller = new AbortController();
+      const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS5);
+      let data;
+      try {
+        const res = await fetch(`${OPENAI_API_BASE}/models`, {
+          headers: { Authorization: `Bearer ${this.apiKey}` },
+          signal: controller.signal
+        });
+        if (!res.ok) return [];
+        data = await res.json();
+      } finally {
+        clearTimeout(timer);
+      }
+      const entries = data?.data ?? [];
+      const logo = getProviderLogo("openai");
+      const models = [];
+      for (const entry of entries) {
+        const mod = classifyModel(entry.id);
+        if (!mod) continue;
+        const info = {
+          id: entry.id,
+          provider: "openai-media",
+          name: entry.id,
+          modality: mod,
+          local: false,
+          cost: { price: 0, unit: "per_request" },
+          logo,
+          description: entry.description,
+          capabilities: this.getCapabilities(entry.id, mod)
+        };
+        models.push(info);
+      }
+      this.modelsCache = models;
+      return modality ? models.filter((m) => m.modality === modality) : models;
+    } catch {
+      return [];
+    }
+  }
+  async image(options) {
+    const model = options.model ?? "gpt-image-1";
+    const width = options.width ?? 1024;
+    const height = options.height ?? 1024;
+    const start = Date.now();
+    const isGptImage = model.startsWith("gpt-image-");
+    const body = {
+      model,
+      prompt: options.prompt,
+      n: 1,
+      size: `${width}x${height}`
+    };
+    if (!isGptImage) {
+      body.response_format = "url";
+    }
+    const res = await fetch(`${OPENAI_API_BASE}/images/generations`, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        Authorization: `Bearer ${this.apiKey}`
+      },
+      body: JSON.stringify(body)
+    });
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`OpenAI image generation failed (${res.status}): ${errorBody}`);
+    }
+    const data = await res.json();
+    const item = data?.data?.[0];
+    const result = {
+      provider: "openai-media",
+      model,
+      modality: "image",
+      latencyMs: Date.now() - start,
+      usage: {
+        cost: 0,
+        unit: "per_image"
+      },
+      media: {
+        width,
+        height,
+        format: "png"
+      }
+    };
+    if (item?.b64_json) {
+      result.buffer = Buffer.from(item.b64_json, "base64");
+    } else if (item?.url) {
+      result.url = item.url;
+    }
+    return result;
+  }
+  async speak(options) {
+    const model = options.model ?? "tts-1";
+    const voice = options.voice ?? "alloy";
+    const format = options.format ?? "mp3";
+    const speed = options.speed ?? 1;
+    const start = Date.now();
+    const body = {
+      model,
+      input: options.text,
+      voice,
+      response_format: format,
+      speed
+    };
+    const res = await fetch(`${OPENAI_API_BASE}/audio/speech`, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        Authorization: `Bearer ${this.apiKey}`
+      },
+      body: JSON.stringify(body)
+    });
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`OpenAI TTS failed (${res.status}): ${errorBody}`);
+    }
+    const arrayBuffer = await res.arrayBuffer();
+    const buffer = Buffer.from(arrayBuffer);
+    return {
+      buffer,
+      provider: "openai-media",
+      model,
+      modality: "tts",
+      latencyMs: Date.now() - start,
+      usage: {
+        cost: 0,
+        input: options.text.length,
+        unit: "per_1k_chars"
+      },
+      media: {
+        format
+      }
+    };
+  }
+  async video(options) {
+    const model = options.model ?? "sora-2";
+    const start = Date.now();
+    const body = {
+      model,
+      prompt: options.prompt,
+      n: 1
+    };
+    if (options.duration) body.duration = options.duration;
+    if (options.width && options.height) body.size = `${options.width}x${options.height}`;
+    const res = await fetch(`${OPENAI_API_BASE}/videos/generations`, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        Authorization: `Bearer ${this.apiKey}`
+      },
+      body: JSON.stringify(body)
+    });
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`OpenAI video generation failed (${res.status}): ${errorBody}`);
+    }
+    const data = await res.json();
+    const videoUrl = data?.data?.[0]?.url;
+    return {
+      url: videoUrl,
+      provider: "openai-media",
+      model,
+      modality: "video",
+      latencyMs: Date.now() - start,
+      usage: {
+        cost: 0,
+        unit: "per_video"
+      },
+      media: {
+        duration: options.duration,
+        width: options.width,
+        height: options.height
+      }
+    };
+  }
+  getCapabilities(id, modality) {
+    if (modality === "image") {
+      return {
+        maxWidth: id.startsWith("dall-e-3") ? 1792 : 1024,
+        maxHeight: id.startsWith("dall-e-3") ? 1792 : 1024
+      };
+    }
+    if (modality === "tts") {
+      return {
+        voices: ["alloy", "ash", "coral", "echo", "fable", "onyx", "nova", "sage", "shimmer"]
+      };
+    }
+    if (modality === "video") {
+      return {
+        maxDuration: id.includes("pro") ? 20 : 10,
+        supportsStreaming: false
+      };
+    }
+    if (modality === "stt") {
+      return {
+        languages: ["en", "zh", "de", "es", "ru", "ko", "fr", "ja", "pt", "tr", "pl", "ca", "nl", "ar", "sv", "it", "id", "hi", "fi", "vi", "he", "uk", "el", "ms", "cs", "ro", "da", "hu", "ta", "no", "th", "ur", "hr", "bg", "lt", "la", "mi", "ml", "cy", "sk", "te", "fa", "lv", "bn", "sr", "az", "sl", "kn", "et", "mk", "br", "eu", "is", "hy", "ne", "mn", "bs", "kk", "sq", "sw", "gl", "mr", "pa", "si", "km", "sn", "yo", "so", "af", "oc", "ka", "be", "tg", "sd", "gu", "am", "yi", "lo", "uz", "fo", "ht", "ps", "tk", "nn", "mt", "sa", "lb", "my", "bo", "tl", "mg", "as", "tt", "haw", "ln", "ha", "ba", "jw", "su"]
+      };
+    }
+    return void 0;
+  }
+};
+
+// src/providers/google-media.ts
+var GOOGLE_API_BASE = "https://generativelanguage.googleapis.com/v1beta";
+var FETCH_TIMEOUT_MS6 = 8e3;
+var GOOGLE_TTS_VOICES = [
+  "Aoede",
+  "Charon",
+  "Fenrir",
+  "Kore",
+  "Puck",
+  "Leda",
+  "Orus",
+  "Perseus",
+  "Zephyr",
+  "Callirrhoe"
+];
+function classifyGoogleModel(model) {
+  const name = (model.name ?? "").replace("models/", "");
+  const methods = model.supportedGenerationMethods ?? [];
+  if (name.startsWith("imagen") && methods.includes("predict")) return "image";
+  if (name.startsWith("veo") && methods.includes("predictLongRunning")) return "video";
+  if (name.includes("-tts") && methods.includes("generateContent")) return "tts";
+  return null;
+}
+var GoogleMediaProvider = class {
+  constructor(apiKey) {
+    this.apiKey = apiKey;
+  }
+  id = "google-media";
+  name = "Google (Image, Video, TTS)";
+  modalities = ["image", "video", "tts"];
+  isLocal = false;
+  modelsCache = null;
+  async ping() {
+    try {
+      const controller = new AbortController();
+      const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS6);
+      try {
+        const res = await fetch(`${GOOGLE_API_BASE}/models?key=${this.apiKey}`, {
+          signal: controller.signal
+        });
+        return res.ok;
+      } finally {
+        clearTimeout(timer);
+      }
+    } catch {
+      return false;
+    }
+  }
+  async listModels(modality) {
+    if (this.modelsCache) {
+      return modality ? this.modelsCache.filter((m) => m.modality === modality) : this.modelsCache;
+    }
+    try {
+      const controller = new AbortController();
+      const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS6);
+      let data;
+      try {
+        const res = await fetch(`${GOOGLE_API_BASE}/models?key=${this.apiKey}`, {
+          signal: controller.signal
+        });
+        if (!res.ok) return [];
+        data = await res.json();
+      } finally {
+        clearTimeout(timer);
+      }
+      const entries = data?.models ?? [];
+      const logo = getProviderLogo("google");
+      const models = [];
+      for (const entry of entries) {
+        const modality2 = classifyGoogleModel(entry);
+        if (!modality2) continue;
+        const fullName = entry.name ?? "";
+        const modelId = fullName.startsWith("models/") ? fullName.slice("models/".length) : fullName;
+        const info = {
+          id: modelId,
+          provider: "google-media",
+          name: entry.displayName ?? modelId,
+          modality: modality2,
+          local: false,
+          cost: { price: 0, unit: modality2 === "video" ? "per_video" : "per_image" },
+          logo,
+          description: entry.description,
+          capabilities: modality2 === "video" ? { maxDuration: 8, supportsStreaming: false } : modality2 === "tts" ? { voices: GOOGLE_TTS_VOICES } : void 0
+        };
+        models.push(info);
+      }
+      this.modelsCache = models;
+      return modality ? models.filter((m) => m.modality === modality) : models;
+    } catch {
+      return [];
+    }
+  }
+  async image(options) {
+    const model = options.model ?? "imagen-4.0-generate-001";
+    const start = Date.now();
+    const body = {
+      instances: [{ prompt: options.prompt }],
+      parameters: {
+        sampleCount: 1
+      }
+    };
+    const res = await fetch(
+      `${GOOGLE_API_BASE}/models/${model}:predict?key=${this.apiKey}`,
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(body)
+      }
+    );
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`Google image generation failed (${res.status}): ${errorBody}`);
+    }
+    const data = await res.json();
+    const base64 = data?.predictions?.[0]?.bytesBase64Encoded ?? data?.generatedImages?.[0]?.image?.imageBytes;
+    if (!base64) {
+      throw new Error("Google image generation returned no image data");
+    }
+    const buffer = Buffer.from(base64, "base64");
+    return {
+      buffer,
+      provider: "google-media",
+      model,
+      modality: "image",
+      latencyMs: Date.now() - start,
+      usage: {
+        cost: 0,
+        unit: "per_image"
+      },
+      media: {
+        format: "png"
+      }
+    };
+  }
+  async speak(options) {
+    const model = options.model ?? "gemini-2.5-flash-preview-tts";
+    const voice = options.voice ?? "Kore";
+    const start = Date.now();
+    const body = {
+      contents: [{ parts: [{ text: options.text }] }],
+      generationConfig: {
+        response_modalities: ["AUDIO"],
+        speech_config: {
+          voiceConfig: {
+            prebuiltVoiceConfig: { voiceName: voice }
+          }
+        }
+      }
+    };
+    const res = await fetch(
+      `${GOOGLE_API_BASE}/models/${model}:generateContent?key=${this.apiKey}`,
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(body)
+      }
+    );
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`Google TTS failed (${res.status}): ${errorBody}`);
+    }
+    const data = await res.json();
+    const inlineData = data?.candidates?.[0]?.content?.parts?.[0]?.inlineData;
+    if (!inlineData?.data) {
+      throw new Error("Google TTS returned no audio data");
+    }
+    const buffer = Buffer.from(inlineData.data, "base64");
+    return {
+      buffer,
+      provider: "google-media",
+      model,
+      modality: "tts",
+      latencyMs: Date.now() - start,
+      usage: {
+        cost: 0,
+        input: options.text.length,
+        unit: "per_1k_chars"
+      },
+      media: {
+        format: "wav"
+        // Google returns PCM L16, essentially WAV
+      }
+    };
+  }
+  async video(options) {
+    const model = options.model ?? "veo-3.0-generate-001";
+    const start = Date.now();
+    const body = {
+      instances: [{ prompt: options.prompt }],
+      parameters: {
+        sampleCount: 1
+      }
+    };
+    if (options.duration) body.parameters.durationSeconds = options.duration;
+    const res = await fetch(
+      `${GOOGLE_API_BASE}/models/${model}:predictLongRunning?key=${this.apiKey}`,
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(body)
+      }
+    );
+    if (!res.ok) {
+      const errorBody = await res.text();
+      throw new Error(`Google video generation failed (${res.status}): ${errorBody}`);
+    }
+    const operation = await res.json();
+    const operationName = operation?.name;
+    if (!operationName) {
+      throw new Error("Google video generation returned no operation name");
+    }
+    const deadline = Date.now() + 3e5;
+    while (Date.now() < deadline) {
+      await new Promise((r) => setTimeout(r, 5e3));
+      const pollRes = await fetch(
+        `${GOOGLE_API_BASE}/${operationName}?key=${this.apiKey}`
+      );
+      if (!pollRes.ok) continue;
+      const status = await pollRes.json();
+      if (status.done) {
+        const videoBase64 = status.response?.generatedSamples?.[0]?.video?.bytesBase64Encoded;
+        if (videoBase64) {
+          return {
+            buffer: Buffer.from(videoBase64, "base64"),
+            provider: "google-media",
+            model,
+            modality: "video",
+            latencyMs: Date.now() - start,
+            usage: { cost: 0, unit: "per_video" },
+            media: { format: "mp4", duration: options.duration }
+          };
+        }
+        const videoUrl = status.response?.generatedSamples?.[0]?.video?.uri;
+        return {
+          url: videoUrl,
+          provider: "google-media",
+          model,
+          modality: "video",
+          latencyMs: Date.now() - start,
+          usage: { cost: 0, unit: "per_video" },
+          media: { format: "mp4", duration: options.duration }
+        };
+      }
+    }
+    throw new Error(`Google video generation timed out after 5 minutes`);
+  }
+};
+
 // src/noosphere.ts
 var Noosphere = class {
   config;
@@ -2564,9 +3217,9 @@ var Noosphere = class {
     if (!this.initialized) await this.init();
     return this.registry.getModel(provider, modelId);
   }
-  async syncModels() {
+  async syncModels(modality) {
     if (!this.initialized) await this.init();
-    return this.registry.syncAll();
+    return this.registry.syncAll(modality);
   }
   // --- Tracking Methods ---
   getUsage(options) {
@@ -2624,6 +3277,12 @@ var Noosphere = class {
     if (hasAnyLLMKey) {
       this.registry.addProvider(new PiAiProvider(llmKeys));
     }
+    if (keys.openai) {
+      this.registry.addProvider(new OpenAIMediaProvider(keys.openai));
+    }
+    if (keys.google) {
+      this.registry.addProvider(new GoogleMediaProvider(keys.google));
+    }
     if (keys.fal) {
       this.registry.addProvider(new FalProvider(keys.fal));
     }
@@ -2803,11 +3462,13 @@ var Noosphere = class {
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   AudioCraftProvider,
+  GoogleMediaProvider,
   HfLocalProvider,
   Noosphere,
   NoosphereError,
   OllamaProvider,
   OpenAICompatProvider,
+  OpenAIMediaProvider,
   PROVIDER_IDS,
   PROVIDER_LOGOS,
   WhisperLocalProvider,