@mixio-pro/kalaasetu-mcp 1.0.8 → 1.0.10

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mixio-pro/kalaasetu-mcp",
-  "version": "1.0.8",
+  "version": "1.0.10",
   "description": "A powerful Model Context Protocol server providing AI tools for content generation and analysis",
   "type": "module",
   "module": "src/index.ts",

package/src/tools/gemini.ts

@@ -10,6 +10,7 @@ import * as os from "os";
 import * as wav from "wav";
 import { PassThrough } from "stream";
 import { getStorage } from "../storage";
+import { generateTimestampedFilename } from "../utils/filename";
 
 const ai = new GoogleGenAI({
   apiKey: process.env.GEMINI_API_KEY || "",
@@ -166,30 +167,44 @@ async function processVideoInput(
 export const geminiTextToImage = {
   name: "generateImage",
   description:
-    "Generate images from text prompts using Gemini 2.5 Flash Image model",
+    "Generate images from text prompts using Gemini image models with optional reference images",
   parameters: z.object({
     prompt: z.string().describe("Text description of the image to generate"),
     aspect_ratio: z
       .string()
       .optional()
-      .describe("Aspect ratio: 1:1, 3:4, 4:3, 9:16, or 16:9"),
+      .describe("Aspect ratio: 1:1, 3:4, 4:3, 9:16, or 16:9 (default 9:16)"),
     output_path: z
       .string()
       .optional()
       .describe("File path to save the generated image"),
+    reference_images: z
+      .array(z.string())
+      .optional()
+      .describe("Optional reference image file paths to guide generation"),
   }),
   execute: async (args: {
     prompt: string;
     aspect_ratio?: string;
     output_path?: string;
+    reference_images?: string[];
   }) => {
     try {
+      const contents: any[] = [args.prompt];
+
+      if (args.reference_images && Array.isArray(args.reference_images)) {
+        for (const refPath of args.reference_images) {
+          contents.push(await fileToGenerativePart(refPath));
+        }
+      }
+
       const response = await ai.models.generateContent({
-        model: "gemini-2.5-flash-image",
-        contents: args.prompt,
+        model: "gemini-3-pro-image-preview",
+        contents: contents,
         config: {
+          responseModalities: ["TEXT", "IMAGE"],
           imageConfig: {
-            aspectRatio: args.aspect_ratio || "1:1",
+            aspectRatio: args.aspect_ratio || "9:16",
           },
         },
       });
@@ -205,13 +220,16 @@ export const geminiTextToImage = {
             const imageData = part.inlineData.data;
             if (args.output_path) {
               const storage = getStorage();
+              const timestampedPath = generateTimestampedFilename(
+                args.output_path
+              );
               const url = await storage.writeFile(
-                args.output_path,
+                timestampedPath,
                 Buffer.from(imageData, "base64")
               );
               images.push({
                 url,
-                filename: args.output_path,
+                filename: timestampedPath,
                 mimeType: "image/png",
               });
             }
@@ -238,7 +256,7 @@ export const geminiTextToImage = {
 export const geminiEditImage = {
   name: "editImage",
   description:
-    "Edit existing images with text instructions using Gemini 2.5 Flash Image Preview",
+    "Edit existing images with text instructions using Gemini 3 Pro Image Preview",
   parameters: z.object({
     image_path: z.string().describe("Path to the source image file"),
     prompt: z.string().describe("Text instructions for editing the image"),
@@ -268,7 +286,7 @@ export const geminiEditImage = {
       }
 
       const response = await ai.models.generateContent({
-        model: "gemini-2.5-flash-image-preview",
+        model: "gemini-3-pro-image-preview",
         contents: contents,
       });
 
@@ -283,13 +301,16 @@ export const geminiEditImage = {
             const imageData = part.inlineData.data;
             if (args.output_path) {
               const storage = getStorage();
+              const timestampedPath = generateTimestampedFilename(
+                args.output_path
+              );
               const url = await storage.writeFile(
-                args.output_path,
+                timestampedPath,
                 Buffer.from(imageData, "base64")
               );
               images.push({
                 url,
-                filename: args.output_path,
+                filename: timestampedPath,
                 mimeType: "image/png",
               });
             }
@@ -425,10 +446,11 @@ export const geminiSingleSpeakerTts = {
       const audioBuffer = Buffer.from(data, "base64");
 
       // Generate output filename if not provided
-      const outputPath = args.output_path || `voice_output_${Date.now()}.wav`;
+      const outputPath = args.output_path || "voice_output.wav";
+      const timestampedPath = generateTimestampedFilename(outputPath);
 
       const storage = getStorage();
-      const url = await storage.writeFile(outputPath, audioBuffer);
+      const url = await storage.writeFile(timestampedPath, audioBuffer);
 
       return JSON.stringify({
         audio: {

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

@@ -4,6 +4,7 @@ import { exec } from "child_process";
 import * as path from "path";
 import { z } from "zod";
 import { getStorage } from "../storage";
+import { generateTimestampedFilename } from "../utils/filename";
 
 async function wait(ms: number): Promise<void> {
   return new Promise((resolve) => setTimeout(resolve, ms));
@@ -287,11 +288,13 @@ export const imageToVideo = {
       [];
     const saveVideo = async (base64: string, index: number) => {
       if (!base64) return;
-      const filePath = args.output_path
+      const baseFilename = args.output_path
         ? index === 0
           ? args.output_path
           : args.output_path.replace(/\.mp4$/i, `_${index}.mp4`)
-        : `video_output_${Date.now()}${index === 0 ? "" : "_" + index}.mp4`;
+        : `video_output${index > 0 ? `_${index}` : ""}.mp4`;
+
+      const filePath = generateTimestampedFilename(baseFilename);
 
       const buf = Buffer.from(base64, "base64");
       const storage = getStorage();

package/src/utils/filename.ts

@@ -0,0 +1,22 @@
+/**
+ * Generate a timestamped filename to avoid conflicts
+ * Format: YYYYMMDD_HHmmss_filename.ext
+ */
+export function generateTimestampedFilename(basename: string): string {
+  const now = new Date();
+  const timestamp = now
+    .toISOString()
+    .replace(/[-:]/g, "")
+    .replace(/\.\d{3}Z$/, "")
+    .replace("T", "_");
+
+  // Extract extension if present
+  const lastDot = basename.lastIndexOf(".");
+  if (lastDot > 0) {
+    const name = basename.substring(0, lastDot);
+    const ext = basename.substring(lastDot);
+    return `${timestamp}_${name}${ext}`;
+  }
+
+  return `${timestamp}_${basename}`;
+}