squidclaw 0.8.2 → 0.8.3

package/lib/tools/image-gen.js

@@ -1,6 +1,6 @@
 /**
  * 🦑 Image Generation Tool
- * Supports: OpenAI DALL-E, Google Gemini/Imagen
+ * Supports: OpenAI DALL-E 3, Google Gemini Flash Image
  */
 
 import { logger } from '../core/logger.js';
@@ -13,13 +13,14 @@ export class ImageGenTool {
   async generate(prompt, provider) {
     const providers = this.config.ai?.providers || {};
 
-    // Auto-select provider
     if (!provider) {
       if (providers.openai?.key) provider = 'openai';
       else if (providers.google?.key) provider = 'google';
       else throw new Error('No image generation API key configured');
     }
 
+    logger.info('image-gen', `Generating with ${provider}: ${prompt.slice(0, 80)}...`);
+
     if (provider === 'openai') return this.generateOpenAI(prompt, providers.openai.key);
     if (provider === 'google') return this.generateGemini(prompt, providers.google.key);
     throw new Error('Unsupported provider: ' + provider);
@@ -37,52 +38,30 @@ export class ImageGenTool {
   }
 
   async generateGemini(prompt, apiKey) {
-    // Gemini Imagen 3 via generateImages endpoint
-    const res = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/imagen-3.0-generate-002:predict?key=${apiKey}`, {
+    // Try gemini-2.5-flash-image (native image generation)
+    const res = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-preview-image:generateContent?key=${apiKey}`, {
       method: 'POST',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({
-        instances: [{ prompt }],
-        parameters: { sampleCount: 1, aspectRatio: '1:1' },
+        contents: [{ parts: [{ text: `Generate an image: ${prompt}` }] }],
+        generationConfig: { responseModalities: ['TEXT', 'IMAGE'] },
       }),
     });
     const data = await res.json();
-    
-    if (data.error) {
-      // Fallback: use Gemini 2.0 Flash native image generation
-      return this.generateGeminiFlash(prompt, apiKey);
-    }
 
-    if (data.predictions?.[0]?.bytesBase64Encoded) {
-      return { base64: data.predictions[0].bytesBase64Encoded, mimeType: 'image/png' };
+    if (data.error) {
+      logger.error('image-gen', `Gemini error: ${data.error.message}`);
+      throw new Error(data.error.message);
     }
 
-    // Fallback to Gemini Flash
-    return this.generateGeminiFlash(prompt, apiKey);
-  }
-
-  async generateGeminiFlash(prompt, apiKey) {
-    const res = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp:generateContent?key=${apiKey}`, {
-      method: 'POST',
-      headers: { 'Content-Type': 'application/json' },
-      body: JSON.stringify({
-        contents: [{ parts: [{ text: `Generate an image: ${prompt}` }] }],
-        generationConfig: { responseModalities: ['TEXT', 'IMAGE'] },
-      }),
-    });
-    const data = await res.json();
-    
-    if (data.error) throw new Error(data.error.message);
-    
-    // Find image part in response
     const parts = data.candidates?.[0]?.content?.parts || [];
     for (const part of parts) {
       if (part.inlineData) {
+        logger.info('image-gen', `Image generated! ${part.inlineData.mimeType}, ${part.inlineData.data?.length} bytes`);
         return { base64: part.inlineData.data, mimeType: part.inlineData.mimeType || 'image/png' };
       }
     }
 
-    // Text-only response
     const text = parts.find(p => p.text)?.text;
     throw new Error(text || 'Gemini could not generate an image');
   }

package/lib/tools/router.js

@@ -86,22 +86,23 @@ export class ToolRouter {
           break;
 
         case 'imagine':
-      case 'draw':
-      case 'image': {
-        try {
-          const { ImageGenTool } = await import('./image-gen.js');
-          const gen = new ImageGenTool(this.config);
-          const result = await gen.generate(args);
-          if (result.url) {
-            return { toolUsed: true, toolName: 'image', toolResult: result.url, imageUrl: result.url };
-          } else if (result.base64) {
-            return { toolUsed: true, toolName: 'image', toolResult: '[Image generated]', imageBase64: result.base64, mimeType: result.mimeType };
+        case 'draw':
+        case 'image': {
+          try {
+            const { ImageGenTool } = await import('./image-gen.js');
+            const gen = new ImageGenTool(this.config);
+            const imgResult = await gen.generate(toolArg);
+            if (imgResult.url) {
+              return { toolUsed: true, toolName: 'image', toolResult: imgResult.url, imageUrl: imgResult.url, cleanResponse };
+            } else if (imgResult.base64) {
+              return { toolUsed: true, toolName: 'image', toolResult: '[Image generated]', imageBase64: imgResult.base64, mimeType: imgResult.mimeType, cleanResponse };
+            }
+          } catch (err) {
+            toolResult = 'Image generation failed: ' + err.message;
           }
-        } catch (err) {
-          return { toolUsed: true, toolName: 'image', toolResult: 'Image generation failed: ' + err.message };
+          break;
         }
-      }
-      case 'read':
+        case 'read':
           const page = await this.browser.readPage(toolArg, 3000);
           toolResult = `Title: ${page.title}\n\n${page.content}`;
           break;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "squidclaw",
-  "version": "0.8.2",
+  "version": "0.8.3",
   "description": "\ud83e\udd91 AI agent platform \u2014 human-like agents for WhatsApp, Telegram & more",
   "main": "lib/engine.js",
   "bin": {