Update multiagents.py

multiagents.py

@@ -1,34 +1,73 @@
+
+# a multi agent proposal to solve HF agent course final assignment
 import os
-from smolagents import CodeAgent, OpenAIServerModel, PythonInterpreterTool
+import dotenv
+from smolagents import CodeAgent
+from smolagents import OpenAIServerModel
 from tools.fetch import fetch_webpage, search_web
+from smolagents import PythonInterpreterTool
 from tools.yttranscript import get_youtube_transcript, get_youtube_title_description
 from tools.stt import get_text_transcript_from_audio_file
 from tools.image import analyze_image
 from common.mylogger import mylog
-from myprompts import output_format
-import dotenv
+import myprompts
 
 dotenv.load_dotenv()
 
-# --- Models ---
-openai_model = OpenAIServerModel(
-    model_id="gpt-3.5-turbo",
+gemini_model = OpenAIServerModel(
+    model_id="gemini-2.0-flash",
+    api_key=os.environ["GEMINI_API_KEY"],
+    # Google Gemini OpenAI-compatible API base URL
+    api_base="https://generativelanguage.googleapis.com/v1beta/openai/",
+)
+
+vllm_model = OpenAIServerModel(
+    model_id="Qwen/Qwen2.5-1.5B-Instruct",
+    api_base="http://192.168.1.39:18000/v1",  
+    api_key="token-abc123",  
+)
+
+openai_41nano_model = OpenAIServerModel(
+    model_id="gpt-4.1-nano",
+    api_base="https://api.openai.com/v1",
     api_key=os.environ["OPENAI_API_KEY"],
+)
+
+openai_41mini_model = OpenAIServerModel(
+    model_id="gpt-3.5-turbo",
     api_base="https://api.openai.com/v1",
+    api_key=os.environ["OPENAI_API_KEY"],
 )
 
-# --- Tool Agents ---
+
+def check_final_answer(final_answer, agent_memory)  -> bool:
+    """
+    Check if the final answer is correct.
+    basic check on the length of the answer.
+    """
+    mylog("check_final_answer", final_answer)
+    # if return answer is more than 200 characters, we will assume it is not correct    
+    if len(str(final_answer)) > 200:
+        return False
+    else:
+        return True
+
+
 web_agent = CodeAgent(
-    model=openai_model,
-    tools=[search_web, fetch_webpage],
+    model=openai_41nano_model,
+    tools=[
+        search_web,
+        fetch_webpage,                
+    ],
     name="web_agent",
-    description="Searches the web and fetches content from webpages.",
-    max_steps=6,
-    verbosity_level=1,
+    description="Use search engine to find webpages related to a subject and get the page content",
+    additional_authorized_imports=["pandas", "numpy","bs4"],
+    verbosity_level=1,    
+    max_steps=7,
 )
 
 audiovideo_agent = CodeAgent(
-    model=openai_model,
+    model=openai_41nano_model,
     tools=[
         get_youtube_transcript,
         get_youtube_title_description,
@@ -36,57 +75,63 @@ audiovideo_agent = CodeAgent(
         analyze_image
     ],
     name="audiovideo_agent",
-    description="Processes audio, video, and image content.",
-    max_steps=6,
+    description="Extracts information from image, video or audio files from the web",
+    additional_authorized_imports=["pandas", "numpy","bs4", "requests"],
     verbosity_level=1,
+    max_steps=7,
 )
 
-# --- Answer Checker ---
-def check_final_answer(answer: str, _) -> bool:
-    return isinstance(answer, str) and len(answer.strip()) <= 200
 
-# --- Manager Agent ---
+
 manager_agent = CodeAgent(
-    model=openai_model,
-    tools=[PythonInterpreterTool()],
-    managed_agents=[web_agent, audiovideo_agent],
-    name="manager_agent",
-    description="Manages sub-agents to answer complex queries.",
-    final_answer_checks=[check_final_answer],
+    model=openai_41mini_model,
+    tools=[ PythonInterpreterTool()],
+    managed_agents=[web_agent, audiovideo_agent],    
+    additional_authorized_imports=["pandas", "numpy","bs4"],
+    planning_interval=5,
     verbosity_level=2,
-    max_steps=12,
+    final_answer_checks=[check_final_answer],
+    max_steps=15,
+    name="manager_agent",
+    description="A manager agent that coordinates the work of other agents to answer questions.",
 )
 
-# --- Entry Class ---
 class MultiAgent:
     def __init__(self):
-        print("✅ MultiAgent initialized")
+        print("BasicAgent initialized.")
 
     def __call__(self, question: str) -> str:
-        mylog("MultiAgent", question)
+        mylog(self.__class__.__name__, question)        
+
         try:
-            system_instruction = (
-                "You are the lead agent in a multi-agent AI system. Use the sub-agents wisely."
-                " Answer using reasoning, planning, and tools where needed."
-                " Stay concise and always format the answer exactly as required."
-            )
+            prefix = """You are the top agent of a multi-agent system that can answer questions by coordinating the work of other agents.
+            You will receive a question and you will decide which agent to use to answer it.
+            You can use the web_agent to search the web for information and for fetching the content of a web page, or the audiovideo_agent to extract information from video or audio files.
+            You can also use your own knowledge to answer the question.
+            You need to respect the output format that is given to you.
+            Finding the correct answer to the question need reasoning and plannig, read the question carrefully, think step by step and do not skip any steps.
+            """
 
-            prompt = f"""
-{system_instruction}
+            question = prefix + "\nTHE QUESTION:\n" + question + '\n' + myprompts.output_format
 
-THE QUESTION:
-{question.strip()}
+            fixed_answer = ""
 
-{output_format.strip()}
-"""
-            return str(manager_agent.run(prompt))
+            fixed_answer = manager_agent.run(question)
+            
+            return fixed_answer
         except Exception as e:
-            error_msg = f"An error occurred while processing the question: {e}"
-            print(error_msg)
-            return error_msg
+            error = f"An error occurred while processing the question: {e}"
+            print(error)
+            return error
 
 
 if __name__ == "__main__":
-    example_question = "What was the actual enrollment of the Malko competition in 2023?"
+    # Example usage
+   
+    question = """
+What was the actual enrollment of the Malko competition in 2023?
+"""
     agent = MultiAgent()
-    print(agent(example_question))
+    answer = agent(question)
+    print(f"Answer: {answer}")
+