Merge branch 'main' into refactor/cleanup-duplicates-in-agent

Author: yufei311

pyproject.toml

@@ -37,7 +37,7 @@ pyautogui="^0.9.54"
 json5="^0.12.0"
 tenacity="^9.1.2"
 crawl4ai="^0.6.3"
-camelot="^12.06.29"
+camelot-py="^1.0.0"
 xlrd="^0.7.1"
 toml = "^0.10.2"
 googlesearch-python = "^1.3.0"

requirements.txt

@@ -26,7 +26,7 @@ pyautogui>=0.9.54
 json5>=0.12.0
 tenacity>=9.1.2
 crawl4ai>=0.6.3
-camelot>=12.06.29
+camelot-py>=1.0.0
 toml>=0.10.2
 xlrd>=0.7.1
 googlesearch-python>=1.3.0

src/agent/agent.py

@@ -51,14 +51,13 @@ def create_agent():
                 raise ValueError(f"Tool ID '{tool_id}' is not registered.")
             tools.append(REGISTED_TOOLS[tool_id]())
 
-        #tools = tools + sub_agent_tools
+        tools = tools + sub_agent_tools
 
         agent = REGISTED_AGENTS["planning_agent"](
             config=planning_agent_config,
             model=model_manager.registed_models[planning_agent_config.model_id],
             tools=tools,
             max_steps=planning_agent_config.max_steps,
-            managed_agents=sub_agents,
             description=planning_agent_config.description,
             name=planning_agent_config.name,
             provide_run_summary=True,

src/config/cfg.py

@@ -21,14 +21,14 @@ class SearcherToolConfig(BaseModel):
     max_length: int = Field(default=50000, description="Maximum character length for the content to be fetched")
 
 class DeepResearcherToolConfig(BaseModel):
-    model_id: str = Field(default="claude-4-sonnet", description="Model ID for the LLM to use")
+    model_id: str = Field(default="gpt-4.1", description="Model ID for the LLM to use")
     max_depth: int = Field(default=2, description="Maximum depth for the search")
     max_insights: int = Field(default=20, description="Maximum number of insights to extract")
     time_limit_seconds: int = Field(default=60, description="Time limit for the search in seconds")
     max_follow_ups: int = Field(default=3, description="Maximum number of follow-up questions to ask")
 
 class BrowserToolConfig(BaseModel):
-    model_id: str = Field(default="langchain-gpt-4.1", description="Model ID for the LLM to use")
+    model_id: str = Field(default="gpt-4.1", description="Model ID for the LLM to use")
     headless: bool = Field(False, description="Whether to run browser in headless mode")
     disable_security: bool = Field(True, description="Disable browser security features")
     extra_chromium_args: List[str] = Field(default_factory=list, description="Extra arguments to pass to the browser")
@@ -48,7 +48,7 @@ class DeepAnalyzerToolConfig(BaseModel):
     summarizer_model_id: str = Field(default="gemini-2.5-pro", description="Model ID for the LLM to use")
 
 class AgentConfig(BaseModel):
-    model_id: str = Field(default="claude-4-sonnet",
+    model_id: str = Field(default="gpt-4.1",
                           description="Model ID for the LLM to use")
     name: str = Field(default="agent", 
                       description="Name of the agent")
@@ -64,10 +64,10 @@ class AgentConfig(BaseModel):
                                       description="List of agents the agent can manage")
 
 class HierarchicalAgentConfig(BaseModel):
-    name: str = Field(default="agentscope", description="Name of the hierarchical agent")
+    name: str = Field(default="dra", description="Name of the hierarchical agent")
     use_hierarchical_agent: bool = Field(default=True, description="Whether to use hierarchical agent")
     planning_agent_config: AgentConfig = Field(default_factory=lambda: AgentConfig(
-        model_id="claude-4-sonnet",
+        model_id="gpt-4.1",
         name="planning_agent",
         description="A planning agent that can plan the steps to complete the task.",
         max_steps=20,
@@ -76,7 +76,7 @@ class HierarchicalAgentConfig(BaseModel):
         managed_agents=["deep_analyzer_agent", "browser_use_agent", "deep_researcher_agent"],
     ))
     deep_analyzer_agent_config: AgentConfig = Field(default_factory=lambda: AgentConfig(
-        model_id="claude-4-sonnet",
+        model_id="gpt-4.1",
         name="deep_analyzer_agent",
         description="A team member that that performs systematic, step-by-step analysis of a given task, optionally leveraging information from external resources such as attached file or uri to provide comprehensive reasoning and answers. For any tasks that require in-depth analysis, particularly those involving attached file or uri, game, chess, computational tasks, or any other complex tasks. Please ask him for the reasoning and the final answer.",
         max_steps=3,
@@ -92,7 +92,7 @@ class HierarchicalAgentConfig(BaseModel):
         tools=["auto_browser_use", "python_interpreter"],
     ))
     deep_researcher_agent_config: AgentConfig = Field(default_factory=lambda: AgentConfig(
-        model_id="claude-4-sonnet",
+        model_id="gpt-4.1",
         name="deep_researcher_agent",
         description="A team member capable of conducting extensive web searches to complete tasks, primarily focused on retrieving broad and preliminary information for quickly understanding a topic or obtaining rough answers. For tasks that require precise, structured, or interactive page-level information retrieval, please use the `browser_use_agent`.",
         max_steps=3,

src/tools/auto_browser.py

@@ -33,7 +33,6 @@ class AutoBrowserUseTool(AsyncTool):
     output_type = "any"
 
     def __init__(self):
-        super(AutoBrowserUseTool, self).__init__()
 
         self.browser_tool_config = config.browser_tool
 
@@ -43,6 +42,8 @@ def __init__(self):
 
         self._init_pdf_server()
 
+        super(AutoBrowserUseTool, self).__init__()
+
     def _init_pdf_server(self):
 
         server_proc = subprocess.Popen(
@@ -70,8 +71,8 @@ async def _browser_task(self, task):
 
         assert model_id in ['gpt-4.1'], f"Model should be in [gpt-4.1, ], but got {model_id}. Please check your config file."
 
-        if "lanchain" not in model_id:
-            model_id = f"lanchain-{model_id}"
+        if "langchain" not in model_id:
+            model_id = f"langchain-{model_id}"
 
         model = model_manager.registed_models[model_id]
 

src/tools/deep_analyzer.py

@@ -55,11 +55,10 @@ class DeepAnalyzerTool(AsyncTool):
         "additionalProperties": False,
     }
     output_type = "any"
-    
-    analyzer_config = config.deep_analyzer_tool
 
     def __init__(self):
-        super().__init__()
+
+        self.analyzer_config = config.deep_analyzer_tool
 
         self.analyzer_models = {
             model_id: model_manager.registed_models[model_id]
@@ -73,6 +72,8 @@ def __init__(self):
             timeout=30,
         )
 
+        super(DeepAnalyzerTool, self).__init__()
+
     async def _analyze(self,
                  model,
                  task: Optional[str] = None,

src/tools/deep_researcher.py

@@ -272,7 +272,7 @@ def __init__(self):
         self.model = model_manager.registed_models[self.deep_researcher_tool_config.model_id]
         self.web_searcher = WebSearcherTool()
         self.web_searcher.fetch_content = True # Enable content fetching
-        super().__init__()
+        super(DeepResearcherTool, self).__init__()
 
     async def forward(
         self,
@@ -566,7 +566,8 @@ async def _analyze_content(
         return insights
 
     async def _summary(self, query: str, reference_materials: str) -> str:
-        model = model_manager.registed_models["Qwen"]
+
+        model = model_manager.registed_models['gpt-4o-search-preview']
 
         messages = [
             {"role": "user", "content": query}

src/tools/search/google_search.py

@@ -146,6 +146,9 @@ def search(params):
     """
 
     base_url = os.getenv("SKYWORK_GOOGLE_SEARCH_API", None)
+
+    query = params.get("q", "")
+    filter_year = params.get("filter_year", None)
 
     # Use local google search api
     if base_url is not None:

tests/test_local_python_executor.py

@@ -3,7 +3,12 @@
 
 # It's good practice to define a small, fixed list for default authorized_imports in tests
 # unless a test specifically needs to modify it.
-TEST_DEFAULT_AUTHORIZED_IMPORTS = ["math"] # Example, can be empty if preferred for stricter tests
+TEST_DEFAULT_AUTHORIZED_IMPORTS = [
+    "math",
+    "subprocess",
+    "os",
+    "os.path",
+] # Example, can be empty if preferred for stricter tests
 
 class TestPythonInterpreterSandbox(unittest.TestCase):
 

tests/test_mdconverter.py

@@ -16,5 +16,5 @@
     model_manager.init_models(use_local_proxy=True)
 
     mdconvert = MarkitdownConverter()
-    md = mdconvert.convert(os.path.join(root, "data/GAIA/2023/validation/1f975693-876d-457b-a649-393859e79bf3.mp3"))
+    md = mdconvert.convert(os.path.join(root, "data/GAIA/2023/validation/366e2f2b-8632-4ef2-81eb-bc3877489217.pdf"))
     print(md)