@contentgrowth/llm-service 1.0.2 → 1.0.3

package/dist/index.cjs

@@ -1381,6 +1381,58 @@ function sanitizeError(error, context = "general") {
 
 // src/transcription-service.js
 var import_openai2 = __toESM(require("openai"), 1);
+
+// src/utils/google-auth.js
+var jose = __toESM(require("jose"), 1);
+var GoogleAuth = class {
+  /**
+   * @param {Object} credentials - Parsed Service Account JSON
+   */
+  constructor(credentials) {
+    if (!credentials || !credentials.private_key || !credentials.client_email) {
+      throw new Error("Invalid Google credentials. Missing private_key or client_email.");
+    }
+    this.credentials = credentials;
+    this.tokenUrl = "https://oauth2.googleapis.com/token";
+    this.scopes = ["https://www.googleapis.com/auth/cloud-platform"];
+  }
+  /**
+   * Generate a signed JWT for the token exchange
+   */
+  async createSignedJWT() {
+    const alg = "RS256";
+    const privateKey = await jose.importPKCS8(this.credentials.private_key, alg);
+    const jwt = await new jose.SignJWT({
+      scope: this.scopes.join(" ")
+    }).setProtectedHeader({ alg }).setIssuedAt().setIssuer(this.credentials.client_email).setAudience(this.tokenUrl).setExpirationTime("1h").sign(privateKey);
+    return jwt;
+  }
+  /**
+   * Get an access token from Google OAuth2
+   * @returns {Promise<string>} Access Token
+   */
+  async getAccessToken() {
+    const jwt = await this.createSignedJWT();
+    const params = new URLSearchParams();
+    params.append("grant_type", "urn:ietf:params:oauth:grant-type:jwt-bearer");
+    params.append("assertion", jwt);
+    const response = await fetch(this.tokenUrl, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/x-www-form-urlencoded"
+      },
+      body: params
+    });
+    if (!response.ok) {
+      const errorText = await response.text();
+      throw new Error(`Failed to get Google Access Token: ${response.status} ${errorText}`);
+    }
+    const data = await response.json();
+    return data.access_token;
+  }
+};
+
+// src/transcription-service.js
 var TranscriptionService = class {
   /**
    * @param {Object} config
@@ -1396,6 +1448,9 @@ var TranscriptionService = class {
     this.model = config.model || "whisper-1";
     if (this.provider === "openai") {
       this.client = new import_openai2.default({ apiKey: this.apiKey });
+    } else if (this.provider === "google") {
+      this.googleAuth = new GoogleAuth(config.googleCredentials);
+      this.projectId = config.googleCredentials.project_id;
     }
   }
   /**
@@ -1413,6 +1468,8 @@ var TranscriptionService = class {
       return this._transcribeOpenAI(audioFile, model, language);
     } else if (this.provider === "generic") {
       return this._transcribeGeneric(audioFile, model, language);
+    } else if (this.provider === "google") {
+      return this._transcribeGoogle(audioFile, model, language);
     } else {
       throw new TranscriptionServiceException(
         `Unsupported transcription provider: ${this.provider}`,
@@ -1471,6 +1528,60 @@ var TranscriptionService = class {
       );
     }
   }
+  async _transcribeGoogle(audioFile, model, language) {
+    try {
+      const accessToken = await this.googleAuth.getAccessToken();
+      const arrayBuffer = await audioFile.arrayBuffer();
+      const audioBytes = Buffer.from(arrayBuffer).toString("base64");
+      let location = "global";
+      if (model === "chirp_3") {
+        location = "us";
+      } else if (model && model.includes("chirp")) {
+        location = "us-central1";
+      }
+      const recognizer = `projects/${this.projectId}/locations/${location}/recognizers/_`;
+      const apiEndpoint = location === "global" ? "https://speech.googleapis.com" : `https://${location}-speech.googleapis.com`;
+      const url = `${apiEndpoint}/v2/${recognizer}:recognize`;
+      const languageCodes = language ? [language] : ["en-US"];
+      const body = {
+        config: {
+          autoDecodingConfig: {},
+          languageCodes,
+          // Sanitize model: Strict allowlist for Google v2 models
+          model: model || "chirp_3"
+        },
+        content: audioBytes
+      };
+      const response = await fetch(url, {
+        method: "POST",
+        headers: {
+          "Authorization": `Bearer ${accessToken}`,
+          "Content-Type": "application/json",
+          "X-Goog-User-Project": this.projectId
+        },
+        body: JSON.stringify(body)
+      });
+      if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`Google Speech API Error ${response.status}: ${errorText}`);
+      }
+      const data = await response.json();
+      let transcript = "";
+      if (data.results) {
+        transcript = data.results.map((r) => {
+          var _a, _b;
+          return ((_b = (_a = r.alternatives) == null ? void 0 : _a[0]) == null ? void 0 : _b.transcript) || "";
+        }).join(" ");
+      }
+      return { text: transcript.trim() };
+    } catch (error) {
+      console.error("[TranscriptionService] Google transcription failed:", error);
+      throw new TranscriptionServiceException(
+        `Google Transcription failed: ${error.message}`,
+        500
+      );
+    }
+  }
 };
 var TranscriptionServiceException = class extends Error {
   constructor(message, statusCode = 500) {
@@ -1485,9 +1596,14 @@ function createSpeechHandler(app, getConfig) {
   app.post("/transcribe", async (c) => {
     try {
       const config = await getConfig(c);
-      if (!config || !config.apiKey) {
+      if (!config) {
         return c.json({ error: "Transcription service not configured" }, 503);
       }
+      if (config.provider === "google" && !config.googleCredentials) {
+        return c.json({ error: "Google credentials not configured" }, 503);
+      } else if (config.provider !== "google" && !config.apiKey) {
+        return c.json({ error: "API key not configured" }, 503);
+      }
       const formData = await c.req.formData();
       const audioFile = formData.get("audio");
       if (!audioFile) {