vibesurf 0.1.25__py3-none-any.whl → 0.1.27__py3-none-any.whl

vibe_surf/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.1.25'
-__version_tuple__ = version_tuple = (0, 1, 25)
+__version__ = version = '0.1.27'
+__version_tuple__ = version_tuple = (0, 1, 27)
 
 __commit_id__ = commit_id = None

vibe_surf/agents/vibe_surf_agent.py

@@ -479,17 +479,16 @@ async def _vibesurf_agent_node_impl(state: VibeSurfState) -> VibeSurfState:
                     llm=vibesurf_agent.llm,
                     file_system=vibesurf_agent.file_system,
                 )
-                if action_name in ["skill_search", "skill_crawl", "skill_summary", "skill_finance"]:
+                if action_name.startswith("skill_"):
                     state.current_step = "END"
                     # Format final response
-                    final_response = f"{result.extracted_content}"
-                    await log_agent_activity(state, agent_name, "result", final_response)
+                    final_response = f"{result.extracted_content}" or f"{result.error}"
                     state.final_response = final_response
                     logger.debug(final_response)
                     state.is_complete = True
-                    return state
+                else:
+                    state.current_step = "vibesurf_agent"
 
-                state.current_step = "vibesurf_agent"
                 if result.extracted_content:
                     vibesurf_agent.message_history.append(
                         UserMessage(content=f'Action result:\n{result.extracted_content}'))

vibe_surf/backend/api/task.py

@@ -104,7 +104,7 @@ async def submit_task(
             logger.info("Using default empty MCP server configuration")
 
         # DEBUG: Log the type and content of mcp_server_config
-        logger.info(f"mcp_server_config type: {type(mcp_server_config)}, value: {mcp_server_config}")
+        logger.debug(f"mcp_server_config type: {type(mcp_server_config)}, value: {mcp_server_config}")
 
         # Create initial task record in database
         from ..database.queries import TaskQueries

vibe_surf/backend/database/queries.py

@@ -486,13 +486,13 @@ class TaskQueries:
                 return existing_task
             else:
                 # DEBUG: Log the type and content of mcp_server_config before saving
-                logger.info(
+                logger.debug(
                     f"Creating task with mcp_server_config type: {type(mcp_server_config)}, value: {mcp_server_config}")
 
                 # Serialize mcp_server_config to JSON string if it's a dict
                 if isinstance(mcp_server_config, dict):
                     mcp_server_config_json = json.dumps(mcp_server_config)
-                    logger.info(f"Converted dict to JSON string: {mcp_server_config_json}")
+                    logger.debug(f"Converted dict to JSON string: {mcp_server_config_json}")
                 else:
                     mcp_server_config_json = mcp_server_config
 

vibe_surf/backend/utils/llm_factory.py

@@ -58,7 +58,7 @@ def create_llm_from_profile(llm_profile) -> BaseChatModel:
             "deepseek": ["temperature"],
             "aws_bedrock": ["temperature"],
             "anthropic_bedrock": ["temperature"],
-            "openai_compatible": ["temperature"]
+            "openai_compatible": ["temperature", "max_tokens"]
         }
 
         # Build common parameters based on provider support

vibe_surf/browser/agent_browser_session.py

@@ -384,6 +384,32 @@ class AgentBrowserSession(BrowserSession):
                 )
             ]
 
+    def model_post_init(self, __context) -> None:
+        """Register event handlers after model initialization."""
+        # Check if handlers are already registered to prevent duplicates
+
+        from browser_use.browser.watchdog_base import BaseWatchdog
+
+        start_handlers = self.event_bus.handlers.get('BrowserStartEvent', [])
+        start_handler_names = [getattr(h, '__name__', str(h)) for h in start_handlers]
+
+        if any('on_BrowserStartEvent' in name for name in start_handler_names):
+            raise RuntimeError(
+                '[BrowserSession] Duplicate handler registration attempted! '
+                'on_BrowserStartEvent is already registered. '
+                'This likely means BrowserSession was initialized multiple times with the same EventBus.'
+            )
+
+        BaseWatchdog.attach_handler_to_session(self, BrowserStartEvent, self.on_BrowserStartEvent)
+        BaseWatchdog.attach_handler_to_session(self, BrowserStopEvent, self.on_BrowserStopEvent)
+        BaseWatchdog.attach_handler_to_session(self, NavigateToUrlEvent, self.on_NavigateToUrlEvent)
+        BaseWatchdog.attach_handler_to_session(self, SwitchTabEvent, self.on_SwitchTabEvent)
+        BaseWatchdog.attach_handler_to_session(self, TabCreatedEvent, self.on_TabCreatedEvent)
+        BaseWatchdog.attach_handler_to_session(self, TabClosedEvent, self.on_TabClosedEvent)
+        BaseWatchdog.attach_handler_to_session(self, AgentFocusChangedEvent, self.on_AgentFocusChangedEvent)
+        # BaseWatchdog.attach_handler_to_session(self, FileDownloadedEvent, self.on_FileDownloadedEvent)
+        BaseWatchdog.attach_handler_to_session(self, CloseTabEvent, self.on_CloseTabEvent)
+
     async def attach_all_watchdogs(self) -> None:
         """Initialize and attach all watchdogs EXCEPT AboutBlankWatchdog to disable DVD animation."""
         # Prevent duplicate watchdog attachment

vibe_surf/cli.py

@@ -325,7 +325,7 @@ def start_backend(port: int) -> None:
         console.print("[yellow]📝 Press Ctrl+C to stop the server[/yellow]\n")
         
         # Run the server
-        uvicorn.run(app, host="127.0.0.1", port=port, log_level="info")
+        uvicorn.run(app, host="127.0.0.1", port=port, log_level="error")
         
     except KeyboardInterrupt:
         console.print("\n[yellow]🛑 Server stopped by user[/yellow]")

vibe_surf/llm/openai_compatible.py

@@ -76,7 +76,7 @@ class ChatOpenAICompatible(ChatOpenAI):
     The class automatically detects the model type and applies appropriate fixes.
     """
 
-    max_completion_tokens: int | None = 16000
+    max_completion_tokens: int | None = 8192
 
     def _is_gemini_model(self) -> bool:
         """Check if the current model is a Gemini model."""

vibe_surf/tools/browser_use_tools.py

@@ -6,6 +6,9 @@ import enum
 import base64
 import mimetypes
 import datetime
+import aiohttp
+import re
+import urllib.parse
 from pathvalidate import sanitize_filename
 from typing import Optional, Type, Callable, Dict, Any, Union, Awaitable, TypeVar
 from pydantic import BaseModel
@@ -40,7 +43,7 @@ from browser_use.browser.views import BrowserError
 from browser_use.mcp.client import MCPClient
 
 from vibe_surf.browser.agent_browser_session import AgentBrowserSession
-from vibe_surf.tools.views import HoverAction, ExtractionAction, FileExtractionAction
+from vibe_surf.tools.views import HoverAction, ExtractionAction, FileExtractionAction, DownloadMediaAction
 from vibe_surf.tools.mcp_client import CustomMCPClient
 from vibe_surf.tools.file_system import CustomFileSystem
 from vibe_surf.logger import get_logger
@@ -501,3 +504,167 @@ class BrowserUseTools(Tools, VibeSurfTools):
                 error_msg = f'❌ Failed to take screenshot: {str(e)}'
                 logger.error(error_msg)
                 return ActionResult(error=error_msg)
+
+        @self.registry.action(
+            'Download media from URL and save to filesystem downloads folder',
+            param_model=DownloadMediaAction
+        )
+        async def download_media(params: DownloadMediaAction, file_system: FileSystem):
+            """Download media from URL with automatic file format detection"""
+            try:
+                # Get file system directory path (Path type)
+                fs_dir = file_system.get_dir()
+                
+                # Create downloads directory if it doesn't exist
+                downloads_dir = fs_dir / "downloads"
+                downloads_dir.mkdir(exist_ok=True)
+                
+                # Download the file and detect format
+                async with aiohttp.ClientSession() as session:
+                    async with session.get(params.url) as response:
+                        if response.status != 200:
+                            raise Exception(f"HTTP {response.status}: Failed to download from {params.url}")
+                        
+                        # Get content
+                        content = await response.read()
+                        
+                        # Detect file format and extension
+                        file_extension = await self._detect_file_format(params.url, response.headers, content)
+                        
+                        # Generate filename
+                        if params.filename:
+                            # Use provided filename, add extension if missing
+                            filename = params.filename
+                            if not filename.endswith(file_extension):
+                                filename = f"{filename}{file_extension}"
+                        else:
+                            # Generate filename from URL or timestamp
+                            url_path = urllib.parse.urlparse(params.url).path
+                            url_filename = os.path.basename(url_path)
+                            
+                            if url_filename and not url_filename.startswith('.'):
+                                # Use URL filename, ensure correct extension
+                                filename = url_filename
+                                if not filename.endswith(file_extension):
+                                    base_name = os.path.splitext(filename)[0]
+                                    filename = f"{base_name}{file_extension}"
+                            else:
+                                # Generate timestamp-based filename
+                                timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+                                filename = f"media_{timestamp}{file_extension}"
+                        
+                        # Sanitize filename
+                        filename = sanitize_filename(filename)
+                        filepath = downloads_dir / filename
+                        
+                        # Save file
+                        with open(filepath, "wb") as f:
+                            f.write(content)
+                        
+                        # Calculate file size for display
+                        file_size = len(content)
+                        size_str = self._format_file_size(file_size)
+                        
+                        msg = f'📥 Downloaded media to: {str(filepath.relative_to(fs_dir))} ({size_str})'
+                        logger.info(msg)
+                        return ActionResult(
+                            extracted_content=msg,
+                            include_in_memory=True,
+                            long_term_memory=f'Downloaded media from {params.url} to {str(filepath.relative_to(fs_dir))}',
+                        )
+                        
+            except Exception as e:
+                error_msg = f'❌ Failed to download media: {str(e)}'
+                logger.error(error_msg)
+                return ActionResult(error=error_msg)
+
+        async def _detect_file_format(self, url: str, headers: dict, content: bytes) -> str:
+            """Detect file format from URL, headers, and content"""
+            
+            # Try Content-Type header first
+            content_type = headers.get('content-type', '').lower()
+            if content_type:
+                # Common image formats
+                if 'image/jpeg' in content_type or 'image/jpg' in content_type:
+                    return '.jpg'
+                elif 'image/png' in content_type:
+                    return '.png'
+                elif 'image/gif' in content_type:
+                    return '.gif'
+                elif 'image/webp' in content_type:
+                    return '.webp'
+                elif 'image/svg' in content_type:
+                    return '.svg'
+                elif 'image/bmp' in content_type:
+                    return '.bmp'
+                elif 'image/tiff' in content_type:
+                    return '.tiff'
+                # Video formats
+                elif 'video/mp4' in content_type:
+                    return '.mp4'
+                elif 'video/webm' in content_type:
+                    return '.webm'
+                elif 'video/avi' in content_type:
+                    return '.avi'
+                elif 'video/mov' in content_type or 'video/quicktime' in content_type:
+                    return '.mov'
+                # Audio formats
+                elif 'audio/mpeg' in content_type or 'audio/mp3' in content_type:
+                    return '.mp3'
+                elif 'audio/wav' in content_type:
+                    return '.wav'
+                elif 'audio/ogg' in content_type:
+                    return '.ogg'
+                elif 'audio/webm' in content_type:
+                    return '.webm'
+            
+            # Try magic number detection
+            if len(content) >= 8:
+                # JPEG
+                if content.startswith(b'\xff\xd8\xff'):
+                    return '.jpg'
+                # PNG
+                elif content.startswith(b'\x89PNG\r\n\x1a\n'):
+                    return '.png'
+                # GIF
+                elif content.startswith(b'GIF87a') or content.startswith(b'GIF89a'):
+                    return '.gif'
+                # WebP
+                elif content[8:12] == b'WEBP':
+                    return '.webp'
+                # BMP
+                elif content.startswith(b'BM'):
+                    return '.bmp'
+                # TIFF
+                elif content.startswith(b'II*\x00') or content.startswith(b'MM\x00*'):
+                    return '.tiff'
+                # MP4
+                elif b'ftyp' in content[4:12]:
+                    return '.mp4'
+                # PDF
+                elif content.startswith(b'%PDF'):
+                    return '.pdf'
+            
+            # Try URL path extension
+            url_path = urllib.parse.urlparse(url).path
+            if url_path:
+                ext = os.path.splitext(url_path)[1].lower()
+                if ext in ['.jpg', '.jpeg', '.png', '.gif', '.webp', '.svg', '.bmp', '.tiff',
+                          '.mp4', '.webm', '.avi', '.mov', '.wmv', '.flv',
+                          '.mp3', '.wav', '.ogg', '.aac', '.flac',
+                          '.pdf', '.doc', '.docx', '.txt']:
+                    return ext
+            
+            # Default fallback
+            return '.bin'
+
+        def _format_file_size(self, size_bytes: int) -> str:
+            """Format file size in human readable format"""
+            if size_bytes == 0:
+                return "0 B"
+            size_names = ["B", "KB", "MB", "GB", "TB"]
+            i = 0
+            while size_bytes >= 1024.0 and i < len(size_names) - 1:
+                size_bytes /= 1024.0
+                i += 1
+            return f"{size_bytes:.1f} {size_names[i]}"