@mixio-pro/kalaasetu-mcp 1.0.12 → 1.0.14

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mixio-pro/kalaasetu-mcp",
-  "version": "1.0.12",
+  "version": "1.0.14",
   "description": "A powerful Model Context Protocol server providing AI tools for content generation and analysis",
   "type": "module",
   "module": "src/index.ts",

package/src/storage/gcs.ts

@@ -1,6 +1,10 @@
 import { GoogleAuth } from "google-auth-library";
 import type { StorageProvider } from "./interface";
 import * as path from "path";
+import {
+  getGoogleAuthClient,
+  getGoogleAccessToken,
+} from "../utils/google-auth";
 
 export class GCSStorageProvider implements StorageProvider {
   private bucket: string;
@@ -8,7 +12,7 @@ export class GCSStorageProvider implements StorageProvider {
 
   constructor(bucket: string) {
     this.bucket = bucket;
-    this.auth = new GoogleAuth({
+    this.auth = getGoogleAuthClient({
       scopes: ["https://www.googleapis.com/auth/cloud-platform"],
     });
   }
@@ -26,12 +30,12 @@ export class GCSStorageProvider implements StorageProvider {
   }
 
   private async getAccessToken(): Promise<string> {
-    const client = await this.auth.getClient();
-    const token = await client.getAccessToken();
-    if (!token.token) {
-      throw new Error("Failed to get GCS access token");
-    }
-    return token.token;
+    // Use the centralized helper which includes gcloud fallback if needed,
+    // although GCSStorageProvider primarily relies on the initialized GoogleAuth.
+    // However, if we want to support the same fallback logic as tools:
+    return getGoogleAccessToken({
+      scopes: ["https://www.googleapis.com/auth/cloud-platform"],
+    });
   }
 
   async readFile(filePath: string): Promise<Buffer> {

package/src/tools/image-to-video.ts

@@ -1,4 +1,3 @@
-import * as fs from "fs";
 import { GoogleAuth } from "google-auth-library";
 import { exec } from "child_process";
 import * as path from "path";
@@ -6,50 +5,12 @@ import { z } from "zod";
 import { getStorage } from "../storage";
 import { generateTimestampedFilename } from "../utils/filename";
 
+import { getGoogleAccessToken } from "../utils/google-auth";
+
 async function wait(ms: number): Promise<void> {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 
-async function fetchAccessToken(): Promise<string> {
-  try {
-    const auth = new GoogleAuth({
-      scopes: ["https://www.googleapis.com/auth/cloud-platform"],
-    });
-    const client = await auth.getClient();
-    const token = await client.getAccessToken();
-    if (!token || typeof token !== "string") {
-      throw new Error("No token from GoogleAuth");
-    }
-    return token;
-  } catch (e) {
-    // Fallback to gcloud
-    return await new Promise((resolve, reject) => {
-      exec("gcloud auth print-access-token", (err, stdout, stderr) => {
-        if (err) {
-          reject(
-            new Error(
-              `Failed to fetch an access token (ADC and gcloud): ${
-                stderr || err.message
-              }`
-            )
-          );
-          return;
-        }
-        const t = (stdout || "").trim();
-        if (!t) {
-          reject(
-            new Error(
-              "Failed to fetch an access token: empty token from gcloud"
-            )
-          );
-          return;
-        }
-        resolve(t);
-      });
-    });
-  }
-}
-
 async function fileToBase64(
   filePath: string
 ): Promise<{ data: string; mimeType: string }> {
@@ -99,7 +60,9 @@ export const imageToVideo = {
     duration_seconds: z
       .string()
       .optional()
-      .describe("Video duration in seconds: '4', '6', or '8' (default: '6')"),
+      .describe(
+        "Video duration in seconds. MUST be one of: '4', '6', or '8' (default: '6'). Other values will be rejected by Vertex AI."
+      ),
     resolution: z
       .string()
       .optional()
@@ -164,7 +127,33 @@ export const imageToVideo = {
     const location = args.location_id || "us-central1";
     const modelId = args.model_id || "veo-3.1-fast-generate-preview";
 
-    const token = await fetchAccessToken();
+    // Validate and parse duration_seconds - snap to nearest 4, 6, or 8
+    let durationSeconds = parseInt(args.duration_seconds || "6");
+    if (isNaN(durationSeconds)) durationSeconds = 6;
+
+    const validDurations = [4, 6, 8];
+    // Find nearest valid duration
+    durationSeconds = validDurations.reduce((prev, curr) => {
+      return Math.abs(curr - durationSeconds) < Math.abs(prev - durationSeconds)
+        ? curr
+        : prev;
+    });
+
+    // Tie-breaking: if equidistant (e.g. 5), the reduce above keeps the first one (4) because < is strict.
+    // However, user requested "nearest duration with the ceil", effectively meaning round up if equidistant.
+    // Let's explicitly handle the equidistant cases or just use a custom finder.
+    // 5 -> equidistant to 4 and 6. "With ceil" implies 6.
+    // 7 -> equidistant to 6 and 8. "With ceil" implies 8.
+
+    // Simpler logic for these specific values:
+    if (durationSeconds === 4 && parseInt(args.duration_seconds || "6") === 5) {
+      durationSeconds = 6;
+    }
+    if (durationSeconds === 6 && parseInt(args.duration_seconds || "6") === 7) {
+      durationSeconds = 8;
+    }
+
+    const token = await getGoogleAccessToken();
 
     const url = `https://${location}-aiplatform.googleapis.com/v1/projects/${projectId}/locations/${location}/publishers/google/models/${modelId}:predictLongRunning`;
 
@@ -242,7 +231,7 @@ export const imageToVideo = {
 
     const parameters: any = {
       aspectRatio: args.aspect_ratio || "9:16",
-      durationSeconds: parseInt(args.duration_seconds || "6") || 6,
+      durationSeconds: durationSeconds,
       resolution: args.resolution || "720p",
       negativePrompt: args.negative_prompt,
       generateAudio: args.generate_audio || false,
@@ -302,14 +291,15 @@ export const imageToVideo = {
       [];
     const saveVideo = async (base64: string, index: number) => {
       if (!base64) return;
-      
+
       // Use provided output path or generate default with timestamp
       let filePath: string;
       if (args.output_path) {
         // User provided path - use as-is for first video, add index for subsequent
-        filePath = index === 0
-          ? args.output_path
-          : args.output_path.replace(/\.mp4$/i, `_${index}.mp4`);
+        filePath =
+          index === 0
+            ? args.output_path
+            : args.output_path.replace(/\.mp4$/i, `_${index}.mp4`);
       } else {
         // No path provided - generate timestamped default
         const defaultName = `video_output${index > 0 ? `_${index}` : ""}.mp4`;

package/src/utils/google-auth.ts

@@ -0,0 +1,83 @@
+import { GoogleAuth } from "google-auth-library";
+import { exec } from "child_process";
+
+export interface GoogleAuthOptions {
+  scopes?: string | string[];
+}
+
+/**
+ * Returns a GoogleAuth client, prioritizing GOOGLE_SERVICE_ACCOUNT_JSON env var.
+ */
+export function getGoogleAuthClient(
+  options: GoogleAuthOptions = {}
+): GoogleAuth {
+  const opts: any = {
+    scopes: options.scopes || [
+      "https://www.googleapis.com/auth/cloud-platform",
+    ],
+  };
+
+  if (process.env.GOOGLE_SERVICE_ACCOUNT_JSON) {
+    try {
+      const credentials = JSON.parse(process.env.GOOGLE_SERVICE_ACCOUNT_JSON);
+      opts.credentials = credentials;
+      if (credentials.project_id) {
+        opts.projectId = credentials.project_id;
+      }
+    } catch (error) {
+      console.warn(
+        "Failed to parse GOOGLE_SERVICE_ACCOUNT_JSON, falling back to other methods:",
+        error
+      );
+    }
+  } else if (process.env.GOOGLE_SERVICE_ACCOUNT_JSON_PATH) {
+    // GoogleAuth natively supports keyFile/keyFilename
+    opts.keyFile = process.env.GOOGLE_SERVICE_ACCOUNT_JSON_PATH;
+  }
+
+  return new GoogleAuth(opts);
+}
+
+/**
+ * Returns an access token, trying GoogleAuth first (JSON > ADC), then gcloud CLI.
+ */
+export async function getGoogleAccessToken(
+  options: GoogleAuthOptions = {}
+): Promise<string> {
+  try {
+    const auth = getGoogleAuthClient(options);
+    const client = await auth.getClient();
+    const token = await client.getAccessToken();
+    if (token.token) {
+      return token.token;
+    }
+    throw new Error("No token returned from GoogleAuth");
+  } catch (e) {
+    // Fallback to gcloud
+    console.warn("GoogleAuth failed, falling back to gcloud CLI:", e);
+    return await new Promise((resolve, reject) => {
+      exec("gcloud auth print-access-token", (err, stdout, stderr) => {
+        if (err) {
+          reject(
+            new Error(
+              `Failed to fetch an access token (Auth Library and gcloud): ${
+                stderr || err.message
+              }`
+            )
+          );
+          return;
+        }
+        const t = (stdout || "").trim();
+        if (!t) {
+          reject(
+            new Error(
+              "Failed to fetch an access token: empty token from gcloud"
+            )
+          );
+          return;
+        }
+        resolve(t);
+      });
+    });
+  }
+}