Run benchmark with a CodeAgent that uses gpt 4 turbo and the following tools: PythonInterpreter, GoogleSearch, VisitWebpage and Final_Answer. Achieve 30%

.gitignore

@@ -0,0 +1,13 @@
+# Environment files
+.env
+
+# IDE folders
+.idea/
+.vscode/
+
+# Python cache
+__pycache__/
+*.pyc
+
+# System files
+.DS_Store

app.py

@@ -3,43 +3,37 @@ import openai
 import gradio as gr
 import requests
 import inspect
+import whisper
+from smolagents.tools import Tool, tool
 import pandas as pd
-from smolagents import CodeAgent, HfApiModel
+from dotenv import load_dotenv
+from smolagents import CodeAgent, TransformersModel, GoogleSearchTool, \
+    VisitWebpageTool, PythonInterpreterTool, WebSearchTool, FinalAnswerTool, OpenAIServerModel
 
 # (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+load_dotenv()
 OPEN_AI_KEY = os.getenv("OPEN_AI_KEY")
-
-
-class OpenAIModel:
-    def __init__(self):
-        openai.api_key = OPEN_AI_KEY
-    
-    def run(self, prompt: str) -> str:
-        response = openai.ChatCompletion.create(
-            model="gpt-3.5-turbo",  # or "gpt-4"
-            messages=[{"role": "user", "content": prompt}],
-            max_tokens=300
-        )
-        return response["choices"][0]["message"]["content"]
+SERPER_API_KEY = os.getenv("SERPER_API_KEY")
 
 
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
-class BasicAgent(CodeAgent): 
-    def __init__(self, tools, model, planning_interval):
-        super().__init__(tools=tools, model=model, planning_interval=planning_interval)
-        self.add_base_tools = True  # Add any additional base tools
-        self.planning_interval = planning_interval  # Enable planning every 3 steps
-        self.max_steps = 0
-        print("BasicAgent initialized.")
-    def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
-        answer =   self.run(question[:50], max_steps=5) # "This is a default answer."
-        print(f"Agent returning answer: {answer}")
-        return answer
+
+
+@tool
+def transcribe(mp3_filepath: str) -> str:
+    """
+    This tool returns the transcription of a mp3 file.
+
+    Args:
+        mp3_filepath (str): the filepath name to the mp3 file we want to transcribe
+        """
+    model = whisper.load_model("base")
+    return model.transcribe(mp3_filepath)
+
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
@@ -62,8 +56,21 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
 
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
-        model = OpenAIModel()
-        agent = BasicAgent(tools=[], model=model, planning_interval=3)
+        model = OpenAIServerModel(
+            model_id="gpt-4-turbo",  # or "gpt-4"
+            api_key=OPEN_AI_KEY,
+        )
+        python_tool = PythonInterpreterTool()
+        google_search = GoogleSearchTool(provider="serper")
+        visit_web_page = VisitWebpageTool()
+        final_answer_tool = FinalAnswerTool()
+        web_agent = CodeAgent(
+            tools=[python_tool, google_search, visit_web_page, transcribe, final_answer_tool],
+            model=model,
+            name="web_agent",
+            description="Browses the web to find information",
+            planning_interval=3,
+            add_base_tools=True)
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
@@ -103,7 +110,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = web_agent.run(question_text, max_steps=5)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:

requirements.txt

@@ -1,4 +1,8 @@
+accelerate
 gradio
+gradio[oauth]
 requests
 smolagents
-openai
+openai
+whisper
+ffmpeg-python