@minded-ai/mindedjs 1.0.0-ec2-beta-5 → 1.0.0-ec2-beta-6

package/dist/browserTask/executeBrowserTask.py

@@ -22,6 +22,7 @@ from browser_use.llm import ChatOpenAI
 import os
 import sys
 import logging
+import base64
 from datetime import datetime
 from pathlib import Path
 from dotenv import load_dotenv
@@ -104,17 +105,47 @@ class ScreenshotCapture:
             self.step_counter += 1
             
             # Take screenshot - browser-use takes viewport by default
-            screenshot_bytes = await page.screenshot()
+            screenshot_data = await page.screenshot()
+            
+            # Ensure we have bytes - browser-use might return different format
+            if isinstance(screenshot_data, bytes):
+                screenshot_bytes = screenshot_data
+            elif isinstance(screenshot_data, str):
+                # If it's base64 encoded string
+                screenshot_bytes = base64.b64decode(screenshot_data)
+            else:
+                # Try to get bytes from whatever format it is
+                screenshot_bytes = bytes(screenshot_data)
+            
+            logger.debug(f"Screenshot data type: {type(screenshot_data)}, size: {len(screenshot_bytes)} bytes")
             
             # Generate filename with metadata
             timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")[:-3]
             filename = f"{self.session_id}/step_{step_number:04d}_{timestamp}.png"
             
+            # Optional: Save to temp file for debugging (like Lambda does)
+            if os.getenv('DEBUG_SCREENSHOTS') == 'true':
+                temp_dir = Path("/tmp/screenshots")
+                temp_dir.mkdir(exist_ok=True, parents=True)
+                temp_path = temp_dir / f"step_{step_number:04d}_{timestamp}.png"
+                with open(temp_path, "wb") as f:
+                    f.write(screenshot_bytes)
+                logger.debug(f"💾 Debug: Saved screenshot to: {temp_path}")
+            
             # Upload to S3
             s3_key = f"{self.s3_prefix}{filename}"
             
             try:
-                # Upload with KMS encryption if required by bucket
+                # Verify we have valid PNG data before uploading
+                if not screenshot_bytes or len(screenshot_bytes) == 0:
+                    logger.error(f"❌ Screenshot bytes are empty!")
+                    return
+                
+                # Check for PNG header (89 50 4E 47 = PNG signature)
+                if len(screenshot_bytes) > 4 and screenshot_bytes[:4] != b'\x89PNG':
+                    logger.warning(f"⚠️ Screenshot may not be valid PNG format. First 4 bytes: {screenshot_bytes[:4].hex()}")
+                
+                # Upload with KMS encryption if required by bucket (matching Lambda implementation)
                 self.s3_client.put_object(
                     Bucket=self.s3_bucket,
                     Key=s3_key,
@@ -122,10 +153,11 @@ class ScreenshotCapture:
                     ContentType='image/png',
                     ServerSideEncryption='aws:kms',
                     Tagging='retention=30d'
+                    # Using bucket's default KMS key
                 )
                 
                 s3_url = f"s3://{self.s3_bucket}/{s3_key}"
-                logger.info(f"✅ Screenshot uploaded: {s3_url}")
+                logger.info(f"✅ Screenshot uploaded: {s3_url} (size: {len(screenshot_bytes)} bytes)")
                 
             except self.ClientError as e:
                 logger.error(f"❌ Failed to upload screenshot to S3: {str(e)}")

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@minded-ai/mindedjs",
-  "version": "1.0.0-ec2-beta-5",
+  "version": "1.0.0-ec2-beta-6",
   "description": "MindedJS is a TypeScript library for building agents.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/src/browserTask/executeBrowserTask.py

@@ -22,6 +22,7 @@ from browser_use.llm import ChatOpenAI
 import os
 import sys
 import logging
+import base64
 from datetime import datetime
 from pathlib import Path
 from dotenv import load_dotenv
@@ -104,17 +105,47 @@ class ScreenshotCapture:
             self.step_counter += 1
             
             # Take screenshot - browser-use takes viewport by default
-            screenshot_bytes = await page.screenshot()
+            screenshot_data = await page.screenshot()
+            
+            # Ensure we have bytes - browser-use might return different format
+            if isinstance(screenshot_data, bytes):
+                screenshot_bytes = screenshot_data
+            elif isinstance(screenshot_data, str):
+                # If it's base64 encoded string
+                screenshot_bytes = base64.b64decode(screenshot_data)
+            else:
+                # Try to get bytes from whatever format it is
+                screenshot_bytes = bytes(screenshot_data)
+            
+            logger.debug(f"Screenshot data type: {type(screenshot_data)}, size: {len(screenshot_bytes)} bytes")
             
             # Generate filename with metadata
             timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")[:-3]
             filename = f"{self.session_id}/step_{step_number:04d}_{timestamp}.png"
             
+            # Optional: Save to temp file for debugging (like Lambda does)
+            if os.getenv('DEBUG_SCREENSHOTS') == 'true':
+                temp_dir = Path("/tmp/screenshots")
+                temp_dir.mkdir(exist_ok=True, parents=True)
+                temp_path = temp_dir / f"step_{step_number:04d}_{timestamp}.png"
+                with open(temp_path, "wb") as f:
+                    f.write(screenshot_bytes)
+                logger.debug(f"💾 Debug: Saved screenshot to: {temp_path}")
+            
             # Upload to S3
             s3_key = f"{self.s3_prefix}{filename}"
             
             try:
-                # Upload with KMS encryption if required by bucket
+                # Verify we have valid PNG data before uploading
+                if not screenshot_bytes or len(screenshot_bytes) == 0:
+                    logger.error(f"❌ Screenshot bytes are empty!")
+                    return
+                
+                # Check for PNG header (89 50 4E 47 = PNG signature)
+                if len(screenshot_bytes) > 4 and screenshot_bytes[:4] != b'\x89PNG':
+                    logger.warning(f"⚠️ Screenshot may not be valid PNG format. First 4 bytes: {screenshot_bytes[:4].hex()}")
+                
+                # Upload with KMS encryption if required by bucket (matching Lambda implementation)
                 self.s3_client.put_object(
                     Bucket=self.s3_bucket,
                     Key=s3_key,
@@ -122,10 +153,11 @@ class ScreenshotCapture:
                     ContentType='image/png',
                     ServerSideEncryption='aws:kms',
                     Tagging='retention=30d'
+                    # Using bucket's default KMS key
                 )
                 
                 s3_url = f"s3://{self.s3_bucket}/{s3_key}"
-                logger.info(f"✅ Screenshot uploaded: {s3_url}")
+                logger.info(f"✅ Screenshot uploaded: {s3_url} (size: {len(screenshot_bytes)} bytes)")
                 
             except self.ClientError as e:
                 logger.error(f"❌ Failed to upload screenshot to S3: {str(e)}")