@umituz/web-ai 1.0.5 → 1.0.7

package/dist/providers/index.cjs

@@ -3976,6 +3976,193 @@ function resetProvider() {
   container = null;
 }
 
+// src/providers/embeddings/openai-embeddings.ts
+var DEFAULT_EMBED_TIMEOUT_MS = 6e4;
+function createOpenAiEmbeddingsAdapter(config) {
+  return {
+    async embed(req) {
+      const response = await fetch(`${config.baseUrl}/embeddings`, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${config.apiKey}`,
+          "Content-Type": "application/json"
+        },
+        body: JSON.stringify({ model: req.externalModelId, input: req.input }),
+        signal: AbortSignal.timeout(config.timeoutMs ?? DEFAULT_EMBED_TIMEOUT_MS)
+      });
+      if (!response.ok) {
+        let detail;
+        try {
+          const parsed2 = await response.json();
+          detail = typeof parsed2.error?.message === "string" ? parsed2.error.message : void 0;
+        } catch {
+          detail = void 0;
+        }
+        throw new Error(detail ?? `Embeddings upstream failed with HTTP ${response.status}`);
+      }
+      const parsed = await response.json();
+      const items = Array.isArray(parsed.data) ? parsed.data : [];
+      const byIndex = /* @__PURE__ */ new Map();
+      for (const item of items) {
+        if (typeof item.index !== "number" || !Array.isArray(item.embedding)) continue;
+        byIndex.set(item.index, item.embedding);
+      }
+      const vectors = [];
+      for (let i = 0; i < req.input.length; i += 1) {
+        const vector = byIndex.get(i);
+        if (!vector) throw new Error("Provider returned an incomplete embeddings batch");
+        vectors.push(vector);
+      }
+      const inputTokens = typeof parsed.usage?.prompt_tokens === "number" ? parsed.usage.prompt_tokens : null;
+      return { vectors, inputTokens };
+    }
+  };
+}
+
+// src/providers/audio/audio-transcription.ts
+function createOpenAiAudioTranscriptionAdapter(config) {
+  return {
+    async transcribe(req) {
+      const form = new FormData();
+      form.append("file", new Blob([req.audio], { type: req.mimeType }), req.name);
+      form.append("model", req.externalModelId);
+      form.append("response_format", "json");
+      const response = await fetch(`${config.baseUrl}/audio/transcriptions`, {
+        method: "POST",
+        headers: { Authorization: `Bearer ${config.apiKey}` },
+        body: form,
+        signal: AbortSignal.timeout(config.timeoutMs ?? 6e4)
+      });
+      if (!response.ok) {
+        let detail;
+        try {
+          const body2 = await response.json();
+          detail = body2.error?.message;
+        } catch {
+          detail = void 0;
+        }
+        throw new Error(detail ?? `Audio transcription failed with HTTP ${response.status}`);
+      }
+      let body;
+      try {
+        body = await response.json();
+      } catch {
+        throw new Error("Transcription response was not valid JSON");
+      }
+      const text = (body.text ?? "").trim();
+      if (text.length === 0) throw new Error("Transcription returned no text");
+      return { text };
+    }
+  };
+}
+
+// src/providers/audio/speech-synthesis.ts
+var SPEECH_VOICES = [
+  "tara",
+  "leah",
+  "jess",
+  "julian",
+  "annie",
+  "steve",
+  "zac",
+  "bella"
+];
+var SPEECH_INPUT_MAX = 4e3;
+function createOpenAiSpeechAdapter(config) {
+  return {
+    async speak(req) {
+      const response = await fetch(`${config.baseUrl}/audio/speech`, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${config.apiKey}`,
+          "Content-Type": "application/json"
+        },
+        body: JSON.stringify({
+          model: req.externalModelId,
+          input: req.input,
+          voice: req.voice,
+          response_format: "wav"
+        }),
+        signal: AbortSignal.timeout(config.timeoutMs ?? 6e4)
+      });
+      if (!response.ok) {
+        let detail;
+        try {
+          const body = await response.json();
+          detail = body.error?.message;
+        } catch {
+          detail = void 0;
+        }
+        throw new Error(detail ?? `Speech synthesis failed with HTTP ${response.status}`);
+      }
+      const audio = new Uint8Array(await response.arrayBuffer());
+      if (audio.length === 0) throw new Error("Speech synthesis returned no audio");
+      return {
+        audio,
+        mimeType: response.headers.get("content-type")?.split(";")[0] ?? "audio/wav"
+      };
+    }
+  };
+}
+
+// src/providers/images/openai-images.ts
+var IMAGE_SIZES = ["1024x1024", "1536x1024", "1024x1536"];
+var IMAGE_QUALITIES = ["auto", "low", "medium", "high", "xhigh", "max"];
+var IMAGE_B64_LIMIT = 12e6;
+function createOpenAiImageAdapter(config) {
+  return {
+    async generate(req) {
+      const response = await fetch(`${config.baseUrl}/images/generations`, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${config.apiKey}`,
+          "Content-Type": "application/json"
+        },
+        body: JSON.stringify({
+          model: req.externalModelId,
+          prompt: req.prompt,
+          n: 1,
+          size: req.size,
+          quality: req.quality
+        }),
+        signal: AbortSignal.timeout(config.timeoutMs ?? 6e4)
+      });
+      if (!response.ok) {
+        let detail;
+        try {
+          const body = await response.json();
+          detail = body.error?.message;
+        } catch {
+          detail = void 0;
+        }
+        throw new Error(detail ?? `Image provider failed with HTTP ${response.status}`);
+      }
+      let payload;
+      try {
+        payload = await response.json();
+      } catch {
+        throw new Error("The image provider returned an unreadable response.");
+      }
+      const images = (payload.data ?? []).map((item) => item.b64_json).filter((value) => typeof value === "string" && value.length > 0);
+      if (images.length === 0) {
+        throw new Error("The image provider returned no image data.");
+      }
+      if (images.some((value) => value.length > IMAGE_B64_LIMIT)) {
+        throw new Error("The generated image exceeds the storage limit.");
+      }
+      const usage = payload.usage;
+      return {
+        images,
+        usage: usage ? {
+          inputTokens: usage.input_tokens ?? 0,
+          outputTokens: usage.output_tokens ?? 0,
+          totalTokens: (usage.input_tokens ?? 0) + (usage.output_tokens ?? 0)
+        } : null
+      };
+    }
+  };
+}
+
 exports.ANTHROPIC_API_VERSION = ANTHROPIC_API_VERSION;
 exports.ANTHROPIC_DEFAULT_MAX_TOKENS = ANTHROPIC_DEFAULT_MAX_TOKENS;
 exports.ANTHROPIC_MODELS = ANTHROPIC_MODELS;
@@ -4002,9 +4189,14 @@ exports.GroqConfigStore = GroqConfigStore;
 exports.GroqHttpClient = GroqHttpClient;
 exports.GroqHttpTransport = GroqHttpTransport;
 exports.GroqProvider = GroqProvider;
+exports.IMAGE_B64_LIMIT = IMAGE_B64_LIMIT;
+exports.IMAGE_QUALITIES = IMAGE_QUALITIES;
+exports.IMAGE_SIZES = IMAGE_SIZES;
 exports.ProviderError = ProviderError;
 exports.ProviderFactory = ProviderFactory;
 exports.PrunaProvider = PrunaProvider;
+exports.SPEECH_INPUT_MAX = SPEECH_INPUT_MAX;
+exports.SPEECH_VOICES = SPEECH_VOICES;
 exports.TIMEOUTS = TIMEOUTS;
 exports.TextGenerationService = TextGenerationService;
 exports.TimeoutError = TimeoutError;
@@ -4018,6 +4210,10 @@ exports.createFalProvider = createFalProvider;
 exports.createGeminiProvider = createGeminiProvider;
 exports.createGroqProvider = createGroqProvider;
 exports.createMessage = createMessage;
+exports.createOpenAiAudioTranscriptionAdapter = createOpenAiAudioTranscriptionAdapter;
+exports.createOpenAiEmbeddingsAdapter = createOpenAiEmbeddingsAdapter;
+exports.createOpenAiImageAdapter = createOpenAiImageAdapter;
+exports.createOpenAiSpeechAdapter = createOpenAiSpeechAdapter;
 exports.createProviderFactory = createProviderFactory;
 exports.createPrunaProvider = createPrunaProvider;
 exports.createSystemMessage = createSystemMessage;