vibesurf 0.1.37__tar.gz → 0.1.39__tar.gz

{vibesurf-0.1.37 → vibesurf-0.1.39}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vibesurf
-Version: 0.1.37
+Version: 0.1.39
 Summary: VibeSurf: A powerful browser assistant for vibe surfing
 Author: WarmShao
 License: Apache-2.0

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.1.37'
-__version_tuple__ = version_tuple = (0, 1, 37)
+__version__ = version = '0.1.39'
+__version_tuple__ = version_tuple = (0, 1, 39)
 
-__commit_id__ = commit_id = 'g8865ef13e'
+__commit_id__ = commit_id = 'g60f699d0a'

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/agents/prompts/vibe_surf_prompt.py

@@ -78,7 +78,7 @@ When using Composio tools (those with `cpo.{toolkit_name}.{tool_name}` prefix):
 - **Special Cases**: `skill_deep_research` only returns guidelines only, then follow guidelines to conduct actual research
 - **Execution Policy**: Skill actions execute only once (no need to retry if errors occur), and all results - whether successful or failed - should be presented to users in structured markdown format.
 - **Follow-up Operations**: When users input skill operations without specifying additional tasks, do not automatically perform subsequent operations. Only perform additional tool operations when users specifically request actions like saving results to files or writing reports.
-- After `/search` completion, NEVER use browser agent to deeply investigate search result (unless explicitly emphasized by the user). The user usually only need the search results. Just return the search results.
+- **Search Skill Usage**: `/skill_search` should ONLY be used when users want to quickly obtain specific information or news and user specify `/skill_search` in request. Please analyze user intent carefully - if the request contains other browser tasks or requires more complex web operations, you should generally execute browser tasks instead of using skill search.
 
 ## Language Adaptability
 

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/agents/report_writer_agent.py

@@ -163,19 +163,19 @@ class ReportWriterAgent:
                     You must ALWAYS respond with a valid JSON in this exact format:
                     {{
                       "thinking": "A structured <think>-style reasoning.",
-                      "action":[{{"task_done": {{ }}, // ... more actions in sequence]
+                      "action":[{{"<action_name>": {{<action_params>}}]
                     }}
         
-                    Action list should NEVER be empty.
+                    Action list should NEVER be empty and Each step can only output one action. If multiple actions are output, only the first one will be executed.
                     """
                 else:
                     report_system_prompt += """
                     You must ALWAYS respond with a valid JSON in this exact format:
                     {{
-                      "action":[{{"task_done": {{ }}, // ... more actions in sequence]
+                      "action":[{{"<action_name>": {{<action_params>}}]
                     }}
     
-                    Action list should NEVER be empty.
+                    Action list should NEVER be empty and Each step can only output one action. If multiple actions are output, only the first one will be executed.
                     """
                 self.message_history.append(SystemMessage(content=report_system_prompt))
 
@@ -234,7 +234,7 @@ Please analyze the task, determine if you need to read any additional files, the
                 results = []
                 time_start = time.time()
 
-                for i, action in enumerate(actions):
+                for i, action in enumerate(actions[:1]):
                     action_data = action.model_dump(exclude_unset=True)
                     action_name = next(iter(action_data.keys())) if action_data else 'unknown'
                     logger.info(f"🛠️ Executing action {i + 1}/{len(actions)}: {action_name}")

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/agents/vibe_surf_agent.py

@@ -393,6 +393,8 @@ async def _vibesurf_agent_node_impl(state: VibeSurfState) -> VibeSurfState:
             context_info.append(f"Generated Report: ❌ Failed - {state.generated_report_result.msg}\nPath: {state.generated_report_result.report_path}\n")
 
     context_str = "\n".join(context_info) if context_info else "No additional context available."
+    logger.debug("VibeSurf State Message:\n")
+    logger.debug(context_str)
     vibesurf_agent.message_history.append(UserMessage(content=context_str))
 
     try:
@@ -427,7 +429,7 @@ async def _vibesurf_agent_node_impl(state: VibeSurfState) -> VibeSurfState:
         if hasattr(parsed, 'thinking') and parsed.thinking:
             await log_agent_activity(state, agent_name, "thinking", parsed.thinking)
 
-        for i, action in enumerate(actions):
+        for i, action in enumerate(actions[:1]):
             action_data = action.model_dump(exclude_unset=True)
             action_name = next(iter(action_data.keys())) if action_data else 'unknown'
             logger.info(f"🛠️ Processing VibeSurf action {i + 1}/{len(actions)}: {action_name}")
@@ -1651,19 +1653,19 @@ Please continue with your assigned work, incorporating this guidance only if it'
 You must ALWAYS respond with a valid JSON in this exact format:
 {{
   "thinking": "A structured <think>-style reasoning.",
-  "action":[{{"task_done": {{ }}, // ... more actions in sequence]
+  "action":[{{"<action_name>": {{<action_params>}}]
 }}
 
-Action list should NEVER be empty.
+Action list should NEVER be empty and Each step can only output one action. If multiple actions are output, only the first one will be executed.
                     """
                 else:
                     vibesurf_system_prompt += """
 You must ALWAYS respond with a valid JSON in this exact format:
 {{
-  "action":[{{"task_done": {{ }}, // ... more actions in sequence]
+  "action":[{{"<action_name>": {{<action_params>}}]
 }}
 
-Action list should NEVER be empty.
+Action list should NEVER be empty and Each step can only output one action. If multiple actions are output, only the first one will be executed.
                     """
                 self.message_history.append(SystemMessage(content=vibesurf_system_prompt))
 
@@ -1728,7 +1730,7 @@ Action list should NEVER be empty.
             completion_event = VibeSurfAgentTelemetryEvent(
                 version=vibe_surf.__version__,
                 action='task_completed',
-                task_description=task[:200] if task else None,
+                task_description=task if task else None,
                 model=getattr(self.llm, 'model_name', None),
                 model_provider=getattr(self.llm, 'provider', None),
                 duration_seconds=duration,

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/backend/llm_config.py

@@ -69,6 +69,12 @@ LLM_PROVIDERS = {
     "anthropic_bedrock": [
     ],
     "openai_compatible": [
+    ],
+    "lm_studio":[
+        "qwen/qwen3-vl-8b",
+        "qwen/qwen3-vl-30b",
+        "qwen/qwen3-14b",
+        "openai/gpt-oss-20b"
     ]
 }
 
@@ -167,15 +173,23 @@ PROVIDER_METADATA = {
     "qwen": {
         "display_name": "Qwen",
         "requires_api_key": True,
-        "requires_base_url": True,
+        "requires_base_url": False,
         "supports_tools": True,
-        "supports_vision": True,
+        "supports_vision": False,
         "default_model": ""
     },
     "kimi": {
         "display_name": "Kimi",
         "requires_api_key": True,
-        "requires_base_url": True,
+        "requires_base_url": False,
+        "supports_tools": True,
+        "supports_vision": False,
+        "default_model": ""
+    },
+    "lm_studio": {
+        "display_name": "LM Studio",
+        "requires_api_key": False,
+        "requires_base_url": False,
         "supports_tools": True,
         "supports_vision": True,
         "default_model": ""

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/backend/utils/llm_factory.py

@@ -193,6 +193,14 @@ def create_llm_from_profile(llm_profile) -> BaseChatModel:
                 **common_params
             )
 
+        elif provider == "lm_studio":
+            return ChatOpenAI(
+                model=model,
+                base_url="http://localhost:1234/v1" or base_url,
+                api_key="lm_studio",
+                **common_params
+            )
+
         elif provider == "openai_compatible":
             if not base_url:
                 raise ValueError("OpenAI Compatible provider requires base_url")

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibe_surf/browser/agent_browser_session.py

@@ -5,7 +5,7 @@ import os
 import pdb
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Literal, Self, Union, cast, Optional
-
+from cdp_use.cdp.target import AttachedToTargetEvent, SessionID, TargetID
 from browser_use.browser.session import BrowserSession, CDPSession
 from pydantic import Field
 from browser_use.browser.events import (
@@ -717,6 +717,78 @@ class AgentBrowserSession(BrowserSession):
             self.logger.error(f'Concurrent screenshot failed: {type(e).__name__}: {e}')
             raise
 
+    async def get_or_create_cdp_session(
+            self, target_id: TargetID | None = None, focus: bool = True, new_socket: bool | None = None
+    ) -> CDPSession:
+        """Get or create a CDP session for a target.
+
+        Args:
+                target_id: Target ID to get session for. If None, uses current agent focus.
+                focus: If True, switches agent focus to this target. If False, just returns session without changing focus.
+                new_socket: If True, create a dedicated WebSocket connection. If None (default), creates new socket for new targets only.
+
+        Returns:
+                CDPSession for the specified target.
+        """
+        assert self.cdp_url is not None, 'CDP URL not set - browser may not be configured or launched yet'
+        assert self._cdp_client_root is not None, 'Root CDP client not initialized - browser may not be connected yet'
+        assert self.agent_focus is not None, 'CDP session not initialized - browser may not be connected yet'
+
+        # If no target_id specified, use the current target_id
+        if target_id is None:
+            target_id = self.agent_focus.target_id
+
+        # Check if we already have a session for this target in the pool
+        if target_id in self._cdp_session_pool:
+            session = self._cdp_session_pool[target_id]
+            if focus and self.agent_focus.target_id != target_id:
+                self.logger.debug(
+                    f'[get_or_create_cdp_session] Switching agent focus from {self.agent_focus.target_id} to {target_id}'
+                )
+                self.agent_focus = session
+            if focus:
+                await session.cdp_client.send.Target.activateTarget(params={'targetId': session.target_id})
+                await session.cdp_client.send.Runtime.runIfWaitingForDebugger(session_id=session.session_id)
+            # else:
+            # self.logger.debug(f'[get_or_create_cdp_session] Reusing existing session for {target_id} (focus={focus})')
+            return session
+
+        # If it's the current focus target, return that session
+        if self.agent_focus.target_id == target_id:
+            self._cdp_session_pool[target_id] = self.agent_focus
+            return self.agent_focus
+
+        # Create new session for this target
+        # Default to True for new sessions (each new target gets its own WebSocket)
+        should_use_new_socket = True if new_socket is None else new_socket
+        self.logger.debug(
+            f'[get_or_create_cdp_session] Creating new CDP session for target {target_id} (new_socket={should_use_new_socket})'
+        )
+        session = await CDPSession.for_target(
+            self._cdp_client_root,
+            target_id,
+            new_socket=should_use_new_socket,
+            cdp_url=self.cdp_url if should_use_new_socket else None,
+        )
+        self._cdp_session_pool[target_id] = session
+        # log length of _cdp_session_pool
+        self.logger.debug(f'[get_or_create_cdp_session] new _cdp_session_pool length: {len(self._cdp_session_pool)}')
+
+        # Only change agent focus if requested
+        if focus:
+            self.logger.debug(
+                f'[get_or_create_cdp_session] Switching agent focus from {self.agent_focus.target_id} to {target_id}'
+            )
+            self.agent_focus = session
+            await session.cdp_client.send.Target.activateTarget(params={'targetId': session.target_id})
+            await session.cdp_client.send.Runtime.runIfWaitingForDebugger(session_id=session.session_id)
+        else:
+            self.logger.debug(
+                f'[get_or_create_cdp_session] Created session for {target_id} without changing focus (still on {self.agent_focus.target_id})'
+            )
+
+        return session
+
     async def get_html_content(self, target_id: Optional[str] = None) -> str:
         """
         Get html content of current page

{vibesurf-0.1.37 → vibesurf-0.1.39}/vibesurf.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vibesurf
-Version: 0.1.37
+Version: 0.1.39
 Summary: VibeSurf: A powerful browser assistant for vibe surfing
 Author: WarmShao
 License: Apache-2.0