code-puppy 0.0.348__py3-none-any.whl → 0.0.361__py3-none-any.whl

code_puppy/tools/__init__.py

@@ -55,10 +55,32 @@ from code_puppy.tools.browser.browser_workflows import (
     register_read_workflow,
     register_save_workflow,
 )
+from code_puppy.tools.browser.terminal_command_tools import (
+    register_run_terminal_command,
+    register_send_terminal_keys,
+    register_wait_terminal_output,
+)
+from code_puppy.tools.browser.terminal_screenshot_tools import (
+    register_load_image,
+    register_terminal_compare_mockup,
+    register_terminal_read_output,
+    register_terminal_screenshot,
+)
+
+# Terminal automation tools
+from code_puppy.tools.browser.terminal_tools import (
+    register_check_terminal_server,
+    register_close_terminal,
+    register_open_terminal,
+    register_start_api_server,
+)
 from code_puppy.tools.command_runner import (
     register_agent_run_shell_command,
     register_agent_share_your_reasoning,
 )
+from code_puppy.tools.display import (
+    display_non_streamed_result as display_non_streamed_result,
+)
 from code_puppy.tools.file_modifications import register_delete_file, register_edit_file
 from code_puppy.tools.file_operations import (
     register_grep,
@@ -121,12 +143,26 @@ TOOL_REGISTRY = {
     "browser_wait_for_element": register_wait_for_element,
     "browser_highlight_element": register_browser_highlight_element,
     "browser_clear_highlights": register_browser_clear_highlights,
-    # Browser Screenshots and VQA
+    # Browser Screenshots
     "browser_screenshot_analyze": register_take_screenshot_and_analyze,
     # Browser Workflows
     "browser_save_workflow": register_save_workflow,
     "browser_list_workflows": register_list_workflows,
     "browser_read_workflow": register_read_workflow,
+    # Terminal Connection Tools
+    "terminal_check_server": register_check_terminal_server,
+    "terminal_open": register_open_terminal,
+    "terminal_close": register_close_terminal,
+    "start_api_server": register_start_api_server,
+    # Terminal Command Execution Tools
+    "terminal_run_command": register_run_terminal_command,
+    "terminal_send_keys": register_send_terminal_keys,
+    "terminal_wait_output": register_wait_terminal_output,
+    # Terminal Screenshot Tools
+    "terminal_screenshot_analyze": register_terminal_screenshot,
+    "terminal_read_output": register_terminal_read_output,
+    "terminal_compare_mockup": register_terminal_compare_mockup,
+    "load_image_for_analysis": register_load_image,
 }
 
 

code_puppy/tools/agent_tools.py

@@ -7,6 +7,7 @@ import pickle
 import re
 import traceback
 from datetime import datetime
+from functools import partial
 from pathlib import Path
 from typing import List, Set
 
@@ -28,12 +29,13 @@ from code_puppy.messaging import (
     SubAgentResponseMessage,
     emit_error,
     emit_info,
+    emit_success,
     get_message_bus,
     get_session_context,
     set_session_context,
 )
-from code_puppy.model_factory import ModelFactory, make_model_settings
 from code_puppy.tools.common import generate_group_id
+from code_puppy.tools.subagent_context import subagent_context
 
 # Set to track active subagent invocation tasks
 _active_subagent_tasks: Set[asyncio.Task] = set()
@@ -413,6 +415,9 @@ def register_invoke_agent(agent):
             session_id = f"{session_id}-{hash_suffix}"
         # else: continuing existing session, use session_id as-is
 
+        # Lazy imports to avoid circular dependency
+        from code_puppy.agents.subagent_stream_handler import subagent_stream_handler
+
         # Emit structured invocation message via MessageBus
         bus = get_message_bus()
         bus.emit(
@@ -429,7 +434,27 @@ def register_invoke_agent(agent):
         previous_session_id = get_session_context()
         set_session_context(session_id)
 
+        # Set terminal session for browser-based terminal tools
+        # This uses contextvars which properly propagate through async tasks
+        from code_puppy.tools.browser.terminal_tools import (
+            _terminal_session_var,
+            set_terminal_session,
+        )
+
+        terminal_session_token = set_terminal_session(f"terminal-{session_id}")
+
+        # Set browser session for Camoufox browser tools (qa-kitten, etc.)
+        # This allows parallel agent invocations to each have their own browser
+        from code_puppy.tools.browser.camoufox_manager import (
+            set_browser_session,
+        )
+
+        browser_session_token = set_browser_session(f"browser-{session_id}")
+
         try:
+            # Lazy import to break circular dependency with messaging module
+            from code_puppy.model_factory import ModelFactory, make_model_settings
+
             # Load the specified agent config
             agent_config = load_agent(agent_name)
 
@@ -483,9 +508,6 @@ def register_invoke_agent(agent):
                 manager = get_mcp_manager()
                 mcp_servers = manager.get_servers_for_agent()
 
-            # Get the event_stream_handler for streaming output
-            from code_puppy.agents.event_stream_handler import event_stream_handler
-
             if get_use_dbos():
                 from pydantic_ai.durable_exec.dbos import DBOSAgent
 
@@ -507,11 +529,10 @@ def register_invoke_agent(agent):
                 agent_tools = agent_config.get_available_tools()
                 register_tools_for_agent(temp_agent, agent_tools)
 
-                # Wrap with DBOS - pass event_stream_handler for streaming output
+                # Wrap with DBOS - no streaming for sub-agents
                 dbos_agent = DBOSAgent(
                     temp_agent,
                     name=subagent_name,
-                    event_stream_handler=event_stream_handler,
                 )
                 temp_agent = dbos_agent
 
@@ -540,43 +561,54 @@ def register_invoke_agent(agent):
             # Run the temporary agent with the provided prompt as an asyncio task
             # Pass the message_history from the session to continue the conversation
             workflow_id = None  # Track for potential cancellation
-            if get_use_dbos():
-                # Generate a unique workflow ID for DBOS - ensures no collisions in back-to-back calls
-                workflow_id = _generate_dbos_workflow_id(group_id)
 
-                # Add MCP servers to the DBOS agent's toolsets
-                # (temp_agent is discarded after this invocation, so no need to restore)
-                if subagent_mcp_servers:
-                    temp_agent._toolsets = temp_agent._toolsets + subagent_mcp_servers
+            # Always use subagent_stream_handler to silence output and update console manager
+            # This ensures all sub-agent output goes through the aggregated dashboard
+            stream_handler = partial(subagent_stream_handler, session_id=session_id)
+
+            # Wrap the agent run in subagent context for tracking
+            with subagent_context(agent_name):
+                if get_use_dbos():
+                    # Generate a unique workflow ID for DBOS - ensures no collisions in back-to-back calls
+                    workflow_id = _generate_dbos_workflow_id(group_id)
+
+                    # Add MCP servers to the DBOS agent's toolsets
+                    # (temp_agent is discarded after this invocation, so no need to restore)
+                    if subagent_mcp_servers:
+                        temp_agent._toolsets = (
+                            temp_agent._toolsets + subagent_mcp_servers
+                        )
 
-                with SetWorkflowID(workflow_id):
+                    with SetWorkflowID(workflow_id):
+                        task = asyncio.create_task(
+                            temp_agent.run(
+                                prompt,
+                                message_history=message_history,
+                                usage_limits=UsageLimits(
+                                    request_limit=get_message_limit()
+                                ),
+                                event_stream_handler=stream_handler,
+                            )
+                        )
+                        _active_subagent_tasks.add(task)
+                else:
                     task = asyncio.create_task(
                         temp_agent.run(
                             prompt,
                             message_history=message_history,
                             usage_limits=UsageLimits(request_limit=get_message_limit()),
-                            event_stream_handler=event_stream_handler,
+                            event_stream_handler=stream_handler,
                         )
                     )
                     _active_subagent_tasks.add(task)
-            else:
-                task = asyncio.create_task(
-                    temp_agent.run(
-                        prompt,
-                        message_history=message_history,
-                        usage_limits=UsageLimits(request_limit=get_message_limit()),
-                        event_stream_handler=event_stream_handler,
-                    )
-                )
-                _active_subagent_tasks.add(task)
 
-            try:
-                result = await task
-            finally:
-                _active_subagent_tasks.discard(task)
-                if task.cancelled():
-                    if get_use_dbos() and workflow_id:
-                        DBOS.cancel_workflow(workflow_id)
+                try:
+                    result = await task
+                finally:
+                    _active_subagent_tasks.discard(task)
+                    if task.cancelled():
+                        if get_use_dbos() and workflow_id:
+                            DBOS.cancel_workflow(workflow_id)
 
             # Extract the response from the result
             response = result.output
@@ -603,13 +635,23 @@ def register_invoke_agent(agent):
                 )
             )
 
+            # Emit clean completion summary
+            emit_success(
+                f"✓ {agent_name} completed successfully", message_group=group_id
+            )
+
             return AgentInvokeOutput(
                 response=response, agent_name=agent_name, session_id=session_id
             )
 
-        except Exception:
+        except Exception as e:
+            # Emit clean failure summary
+            emit_error(f"✗ {agent_name} failed: {str(e)}", message_group=group_id)
+
+            # Full traceback for debugging
             error_msg = f"Error invoking agent '{agent_name}': {traceback.format_exc()}"
             emit_error(error_msg, message_group=group_id)
+
             return AgentInvokeOutput(
                 response=None,
                 agent_name=agent_name,
@@ -620,5 +662,13 @@ def register_invoke_agent(agent):
         finally:
             # Restore the previous session context
             set_session_context(previous_session_id)
+            # Reset terminal session context
+            _terminal_session_var.reset(terminal_session_token)
+            # Reset browser session context
+            from code_puppy.tools.browser.camoufox_manager import (
+                _browser_session_var,
+            )
+
+            _browser_session_var.reset(browser_session_token)
 
     return invoke_agent

code_puppy/tools/browser/__init__.py

@@ -0,0 +1,37 @@
+"""Browser tools for terminal automation.
+
+This module provides browser-based terminal automation tools.
+"""
+
+from code_puppy.config import get_banner_color
+
+from .camoufox_manager import (
+    cleanup_all_browsers,
+    get_browser_session,
+    get_session_browser_manager,
+    set_browser_session,
+)
+
+
+def format_terminal_banner(text: str) -> str:
+    """Format a terminal tool banner with the configured terminal_tool color.
+
+    Returns Rich markup string that can be used with Text.from_markup().
+
+    Args:
+        text: The banner text (e.g., "TERMINAL OPEN 🖥️ localhost:8765")
+
+    Returns:
+        Rich markup formatted string
+    """
+    color = get_banner_color("terminal_tool")
+    return f"[bold white on {color}] {text} [/bold white on {color}]"
+
+
+__all__ = [
+    "format_terminal_banner",
+    "cleanup_all_browsers",
+    "get_browser_session",
+    "get_session_browser_manager",
+    "set_browser_session",
+]

code_puppy/tools/browser/browser_control.py

@@ -7,7 +7,7 @@ from pydantic_ai import RunContext
 from code_puppy.messaging import emit_error, emit_info, emit_success, emit_warning
 from code_puppy.tools.common import generate_group_id
 
-from .camoufox_manager import get_camoufox_manager
+from .camoufox_manager import get_session_browser_manager
 
 
 async def initialize_browser(
@@ -22,7 +22,7 @@ async def initialize_browser(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
 
         # Configure browser settings
         browser_manager.headless = headless
@@ -75,7 +75,7 @@ async def close_browser() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         await browser_manager.close()
 
         emit_warning("Browser closed successfully", message_group=group_id)
@@ -94,7 +94,7 @@ async def get_browser_status() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
 
         if not browser_manager._initialized:
             return {
@@ -139,7 +139,7 @@ async def create_new_page(url: Optional[str] = None) -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
 
         if not browser_manager._initialized:
             return {
@@ -168,7 +168,7 @@ async def list_pages() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
 
         if not browser_manager._initialized:
             return {"success": False, "error": "Browser not initialized"}

code_puppy/tools/browser/browser_interactions.py

@@ -7,7 +7,7 @@ from pydantic_ai import RunContext
 from code_puppy.messaging import emit_error, emit_info, emit_success
 from code_puppy.tools.common import generate_group_id
 
-from .camoufox_manager import get_camoufox_manager
+from .camoufox_manager import get_session_browser_manager
 
 
 async def click_element(
@@ -24,14 +24,15 @@ async def click_element(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        # Find element
-        element = page.locator(selector)
+        # Find element - use .first to handle cases where selector matches multiple elements
+        # This avoids Playwright's strict mode violation errors
+        element = page.locator(selector).first
 
         # Wait for element to be visible and enabled
         await element.wait_for(state="visible", timeout=timeout)
@@ -69,13 +70,13 @@ async def double_click_element(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
         await element.dblclick(force=force, timeout=timeout)
 
@@ -99,13 +100,13 @@ async def hover_element(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
         await element.hover(force=force, timeout=timeout)
 
@@ -130,13 +131,13 @@ async def set_element_text(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
 
         if clear_first:
@@ -169,13 +170,13 @@ async def get_element_text(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
 
         text = await element.text_content()
@@ -197,13 +198,13 @@ async def get_element_value(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
 
         value = await element.input_value()
@@ -231,13 +232,13 @@ async def select_option(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
 
         if value is not None:
@@ -278,13 +279,13 @@ async def check_element(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
         await element.check(timeout=timeout)
 
@@ -307,13 +308,13 @@ async def uncheck_element(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
             return {"success": False, "error": "No active browser page available"}
 
-        element = page.locator(selector)
+        element = page.locator(selector).first
         await element.wait_for(state="visible", timeout=timeout)
         await element.uncheck(timeout=timeout)
 

code_puppy/tools/browser/browser_locators.py

@@ -7,7 +7,7 @@ from pydantic_ai import RunContext
 from code_puppy.messaging import emit_info, emit_success
 from code_puppy.tools.common import generate_group_id
 
-from .camoufox_manager import get_camoufox_manager
+from .camoufox_manager import get_session_browser_manager
 
 
 async def find_by_role(
@@ -23,7 +23,7 @@ async def find_by_role(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -75,7 +75,7 @@ async def find_by_text(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -127,7 +127,7 @@ async def find_by_label(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -190,7 +190,7 @@ async def find_by_placeholder(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -248,7 +248,7 @@ async def find_by_test_id(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -304,7 +304,7 @@ async def run_xpath_query(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -359,7 +359,7 @@ async def find_buttons(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -410,7 +410,7 @@ async def find_links(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:

code_puppy/tools/browser/browser_navigation.py

@@ -7,7 +7,7 @@ from pydantic_ai import RunContext
 from code_puppy.messaging import emit_error, emit_info, emit_success
 from code_puppy.tools.common import generate_group_id
 
-from .camoufox_manager import get_camoufox_manager
+from .camoufox_manager import get_session_browser_manager
 
 
 async def navigate_to_url(url: str) -> Dict[str, Any]:
@@ -18,7 +18,7 @@ async def navigate_to_url(url: str) -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -48,7 +48,7 @@ async def get_page_info() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -71,7 +71,7 @@ async def go_back() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -93,7 +93,7 @@ async def go_forward() -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -115,7 +115,7 @@ async def reload_page(wait_until: str = "domcontentloaded") -> Dict[str, Any]:
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page:
@@ -139,7 +139,7 @@ async def wait_for_load_state(
         message_group=group_id,
     )
     try:
-        browser_manager = get_camoufox_manager()
+        browser_manager = get_session_browser_manager()
         page = await browser_manager.get_current_page()
 
         if not page: