add deep research agent

Author: vvincent1234

requirements.txt

@@ -5,4 +5,6 @@ json-repair
 langchain-mistralai==0.2.4
 MainContentExtractor==0.0.4
 langchain-ibm==0.3.10
-langchain_mcp_adapters==0.0.9
+langchain_mcp_adapters==0.0.9
+langgraph==0.3.34
+langchain-community==0.3.23

src/agent/deep_research/deep_research_agent.py

@@ -7,6 +7,7 @@
 from src.webui.webui_manager import WebuiManager
 from src.utils import config
 import logging
+from functools import partial
 
 logger = logging.getLogger(__name__)
 
@@ -23,10 +24,15 @@ def update_model_dropdown(llm_provider):
         return gr.Dropdown(choices=[], value="", interactive=True, allow_custom_value=True)
 
 
-def update_mcp_server(mcp_file: str):
+def update_mcp_server(mcp_file: str, webui_manager: WebuiManager):
     """
     Update the MCP server.
     """
+    if hasattr(webui_manager, "bu_controller") and webui_manager.bu_controller:
+        logger.warning("⚠️ Close controller because mcp file has changed!")
+        webui_manager.bu_controller.close_mcp_client()
+        webui_manager.bu_controller = None
+
     if not mcp_file or not os.path.exists(mcp_file) or not mcp_file.endswith('.json'):
         logger.warning(f"{mcp_file} is not a valid MCP file.")
         return None, gr.update(visible=False)
@@ -37,7 +43,7 @@ def update_mcp_server(mcp_file: str):
     return json.dumps(mcp_server, indent=2), gr.update(visible=True)
 
 
-def create_agent_settings_tab(webui_manager: WebuiManager) -> dict[str, Component]:
+def create_agent_settings_tab(webui_manager: WebuiManager):
     """
     Creates an agent settings tab.
     """
@@ -252,7 +258,7 @@ def create_agent_settings_tab(webui_manager: WebuiManager) -> dict[str, Componen
     )
 
     mcp_json_file.change(
-        update_mcp_server,
+        partial(update_mcp_server, webui_manager=webui_manager),
         inputs=[mcp_json_file],
         outputs=[mcp_server_config, mcp_server_config]
     )

src/webui/components/agent_settings_tab.py

@@ -14,13 +14,16 @@ async def close_browser(webui_manager: WebuiManager):
     if webui_manager.bu_current_task and not webui_manager.bu_current_task.done():
         webui_manager.bu_current_task.cancel()
         webui_manager.bu_current_task = None
-    if webui_manager.bu_browser:
-        await webui_manager.bu_browser.close()
-        webui_manager.bu_browser = None
+
     if webui_manager.bu_browser_context:
+        logger.info("⚠️ Closing browser context when changing browser config.")
         await webui_manager.bu_browser_context.close()
         webui_manager.bu_browser_context = None
 
+    if webui_manager.bu_browser:
+        logger.info("⚠️ Closing browser when changing browser config.")
+        await webui_manager.bu_browser.close()
+        webui_manager.bu_browser = None
 
 def create_browser_settings_tab(webui_manager: WebuiManager):
     """
@@ -43,6 +46,7 @@ def create_browser_settings_tab(webui_manager: WebuiManager):
                 interactive=True,
                 placeholder="Leave it empty if you use your default user data",
             )
+    with gr.Group():
         with gr.Row():
             use_own_browser = gr.Checkbox(
                 label="Use Own Browser",
@@ -64,11 +68,12 @@ def create_browser_settings_tab(webui_manager: WebuiManager):
             )
             disable_security = gr.Checkbox(
                 label="Disable Security",
-                value=True,
-                info="Disable browser security features",
+                value=False,
+                info="Disable browser security",
                 interactive=True
             )
 
+    with gr.Group():
         with gr.Row():
             window_w = gr.Number(
                 label="Window Width",
@@ -82,7 +87,7 @@ def create_browser_settings_tab(webui_manager: WebuiManager):
                 info="Browser window height",
                 interactive=True
             )
-
+    with gr.Group():
         with gr.Row():
             cdp_url = gr.Textbox(
                 label="CDP URL",
@@ -94,7 +99,7 @@ def create_browser_settings_tab(webui_manager: WebuiManager):
                 info="WSS URL for browser remote debugging",
                 interactive=True,
             )
-
+    with gr.Group():
         with gr.Row():
             save_recording_path = gr.Textbox(
                 label="Recording Path",

src/webui/components/browser_settings_tab.py

@@ -1,3 +1,5 @@
+import pdb
+
 import gradio as gr
 from gradio.components import Component
 import asyncio
@@ -388,7 +390,6 @@ async def ask_callback_wrapper(query: str, browser_context: BrowserContext) -> D
                     extra_args += [f"--user-data-dir={chrome_user_data}"]
             else:
                 browser_binary_path = None
-
             webui_manager.bu_browser = CustomBrowser(
                 config=BrowserConfig(
                     headless=headless,
@@ -432,7 +433,6 @@ def done_callback_wrapper(history: AgentHistoryList):
             logger.info(f"Initializing new agent for task: {task}")
             if not webui_manager.bu_browser or not webui_manager.bu_browser_context:
                 raise ValueError("Browser or Context not initialized, cannot create agent.")
-
             webui_manager.bu_agent = BrowserUseAgent(
                 task=task,
                 llm=main_llm,
@@ -456,6 +456,9 @@ def done_callback_wrapper(history: AgentHistoryList):
             webui_manager.bu_agent.state.agent_id = webui_manager.bu_agent_task_id
             webui_manager.bu_agent.add_new_task(task)
             webui_manager.bu_agent.settings.generate_gif = gif_path
+            webui_manager.bu_agent.browser = webui_manager.bu_browser
+            webui_manager.bu_agent.browser_context = webui_manager.bu_browser_context
+            webui_manager.bu_agent.controller = webui_manager.bu_controller
 
         # --- 6. Run Agent Task and Stream Updates ---
         agent_run_coro = webui_manager.bu_agent.run(max_steps=max_steps)
@@ -832,15 +835,13 @@ def create_browser_use_agent_tab(webui_manager: WebuiManager):
 
     async def submit_wrapper(components_dict: Dict[Component, Any]) -> AsyncGenerator[Dict[Component, Any], None]:
         """Wrapper for handle_submit that yields its results."""
-        # handle_submit is an async generator, iterate and yield
         async for update in handle_submit(webui_manager, components_dict):
             yield update
 
     async def stop_wrapper() -> AsyncGenerator[Dict[Component, Any], None]:
         """Wrapper for handle_stop."""
-        # handle_stop is async def but returns a single dict. We yield it once.
         update_dict = await handle_stop(webui_manager)
-        yield update_dict  # Yield the final dictionary
+        yield update_dict
 
     async def pause_resume_wrapper() -> AsyncGenerator[Dict[Component, Any], None]:
         """Wrapper for handle_pause_resume."""

src/webui/components/browser_use_agent_tab.py

@@ -5,7 +5,7 @@
 from src.utils import config
 
 
-def create_deep_research_agent_tab(webui_manager: WebuiManager) -> dict[str, Component]:
+def create_deep_research_agent_tab(webui_manager: WebuiManager):
     """
     Creates a deep research agent tab
     """

src/webui/components/deep_research_agent_tab.py

@@ -5,15 +5,15 @@
 from src.utils import config
 
 
-def create_load_save_config_tab(webui_manager: WebuiManager) -> dict[str, Component]:
+def create_load_save_config_tab(webui_manager: WebuiManager):
     """
     Creates a load and save config tab.
     """
     input_components = set(webui_manager.get_components())
     tab_components = {}
 
     config_file = gr.File(
-        label="Load UI Settings from Config File",
+        label="Load UI Settings from json",
         file_types=[".json"],
         interactive=True
     )

src/webui/components/load_save_config_tab.py

@@ -194,7 +194,6 @@ async def test_browser_use_parallel():
     #     api_key=os.getenv("OPENAI_API_KEY", ""),
     # )
 
-
     # llm = utils.get_llm_model(
     #     provider="google",
     #     model_name="gemini-2.0-flash",
@@ -335,6 +334,70 @@ async def test_browser_use_parallel():
             await browser.close()
 
 
+async def test_deep_research_agent():
+    from src.agent.deep_research.deep_research_agent import DeepSearchAgent
+    from src.utils import llm_provider
+
+    llm = llm_provider.get_llm_model(
+        provider="azure_openai",
+        model_name="gpt-4o",
+        temperature=0.5,
+        base_url=os.getenv("AZURE_OPENAI_ENDPOINT", ""),
+        api_key=os.getenv("AZURE_OPENAI_API_KEY", ""),
+    )
+
+    mcp_server_config = {
+        "mcpServers": {
+            "desktop-commander": {
+                "command": "npx",
+                "args": [
+                    "-y",
+                    "@wonderwhy-er/desktop-commander"
+                ]
+            },
+        }
+    }
+
+    browser_config = {"headless": False, "window_width": 1280, "window_height": 1100, "use_own_browser": False}
+    agent = DeepSearchAgent(llm=llm, browser_config=browser_config, mcp_server_config=mcp_server_config)
+
+    research_topic = "Impact of Microplastics on Marine Ecosystems"
+    task_id_to_resume = None  # Set this to resume a previous task ID
+
+    print(f"Starting research on: {research_topic}")
+
+    try:
+        # Call run and wait for the final result dictionary
+        result = await agent.run(research_topic, task_id=task_id_to_resume)
+
+        print("\n--- Research Process Ended ---")
+        print(f"Status: {result.get('status')}")
+        print(f"Message: {result.get('message')}")
+        print(f"Task ID: {result.get('task_id')}")
+
+        # Check the final state for the report
+        final_state = result.get('final_state', {})
+        if final_state:
+            print("\n--- Final State Summary ---")
+            print(
+                f"  Plan Steps Completed: {sum(1 for item in final_state.get('research_plan', []) if item.get('status') == 'completed')}")
+            print(f"  Total Search Results Logged: {len(final_state.get('search_results', []))}")
+            if final_state.get("final_report"):
+                print("  Final Report: Generated (content omitted). You can find it in the output directory.")
+                # print("\n--- Final Report ---") # Optionally print report
+                # print(final_state["final_report"])
+            else:
+                print("  Final Report: Not generated.")
+        else:
+            print("Final state information not available.")
+
+
+    except Exception as e:
+        print(f"\n--- An unhandled error occurred outside the agent run ---")
+        print(e)
+
+
 if __name__ == "__main__":
     # asyncio.run(test_browser_use_agent())
-    asyncio.run(test_browser_use_parallel())
+    # asyncio.run(test_browser_use_parallel())
+    asyncio.run(test_deep_research_agent())

tests/test_agents.py

@@ -32,6 +32,7 @@ async def test_mcp_client():
     }
 
     mcp_tools, mcp_client = await setup_mcp_client_and_tools(test_server_config)
+
     for tool in mcp_tools:
         tool_param_model = create_tool_param_model(tool)
         print(tool.name)