Update app.py

app.py

@@ -1,80 +1,36 @@
 
 import os
-import gradio as gr
 import requests
 import pandas as pd
+import gradio as gr
+from smolagents import ToolCallingAgent
+from smolagents.models import OpenAIServerModel
 
-from smolagents import ToolCallingAgent, OpenAIServerModel
 from audio_transcriber import AudioTranscriptionTool
 from image_analyzer import ImageAnalysisTool
 from wikipedia_searcher import WikipediaSearcher
 
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-SYSTEM_PROMPT = (
-    "You are an agent solving the GAIA benchmark and must provide exact answers.\n"
-    "Rules:\n"
-    "1. Return only the exact requested answer: no explanation.\n"
-    "2. For yes/no, return 'Yes' or 'No'.\n"
-    "3. For dates, use the exact requested format.\n"
-    "4. For numbers, use only the number.\n"
-    "5. For names, use the exact name from sources.\n"
-    "6. If the question has a file, download it using the task ID.\n"
-    "Examples:\n"
-    "- '42'\n"
-    "- 'Arturo Nunez'\n"
-    "- 'Yes'\n"
-    "- 'October 5, 2001'\n"
-    "- 'Buenos Aires'\n"
-    "Never say 'the answer is...'. Only return the answer.\n"
-)
+DEFAULT_API_URL = "https://gaia-benchmark.com/api"
 
 class GaiaAgent:
     def __init__(self):
-        print("Gaia Agent Initialized")
+        api_key = os.getenv("OPENAI_API_KEY")
+        if not api_key:
+            raise EnvironmentError("OPENAI_API_KEY not found in environment variables.")
 
-        # Initialize the OpenAI GPT-3.5-turbo model via smolagents OpenAIServerModel
-        self.model = OpenAIServerModel(
-            model_name="gpt-3.5-turbo",
-            api_key=os.getenv("OPENAI_API_KEY")  # Make sure you set this in your environment
+        model = OpenAIServerModel(
+            model_id="gpt-3.5-turbo",
+            api_key=api_key
         )
-
-        # Initialize the tools
-        self.tools = [
+        tools = [
             AudioTranscriptionTool(),
             ImageAnalysisTool(),
             WikipediaSearcher()
         ]
+        self.agent = ToolCallingAgent(model=model, tools=tools)
 
-        # Create the agent with tools and model
-        self.agent = ToolCallingAgent(
-            tools=self.tools,
-            model=self.model
-        )
-
-    def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
-
-        full_prompt = f"{SYSTEM_PROMPT}\nQUESTION:\n{question}"
-
-        try:
-            result = self.agent.run(full_prompt)
-            print(f"Raw result from agent: {result}")
-
-            if isinstance(result, dict) and "answer" in result:
-                return str(result["answer"]).strip()
-            elif isinstance(result, str):
-                return result.strip()
-            elif isinstance(result, list):
-                for item in reversed(result):
-                    if isinstance(item, dict) and item.get("role") == "assistant" and "content" in item:
-                        return item["content"].strip()
-                return "ERROR: Unexpected list format"
-            else:
-                return "ERROR: Unexpected result type"
-        except Exception as e:
-            print(f"Exception during agent run: {e}")
-            return f"AGENT ERROR: {e}"
+    def __call__(self, prompt: str) -> str:
+        return self.agent.run([{"role": "user", "content": prompt}])
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     space_id = os.getenv("SPACE_ID")
@@ -119,7 +75,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
         question_text = item.get("question", "")
 
-        # Download associated file if any (mp3 or jpeg) according to GAIA benchmark task
         file_url = item.get("file_url")
         local_file_path = None
         if file_url:
@@ -133,8 +88,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
                             for chunk in r.iter_content(chunk_size=8192):
                                 f.write(chunk)
                     print(f"Downloaded file for task {task_id} to {local_file_path}")
-
-                    # Append info about the file path to the question so the agent knows to use it
                     question_text += f"\n\nFile path: {local_file_path}"
             except Exception as e:
                 print(f"Failed to download file for task {task_id}: {e}")
@@ -155,7 +108,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
                 "Submitted Answer": error_msg
             })
 
-        # Cleanup downloaded file
         if local_file_path:
             try:
                 os.remove(local_file_path)
@@ -196,7 +148,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     except Exception as e:
         return f"An unexpected error occurred during submission: {e}", pd.DataFrame(results_log)
 
-# Gradio UI
+
 with gr.Blocks() as demo:
     gr.Markdown("# Basic Agent Evaluation Runner")
     gr.Markdown("""
@@ -220,16 +172,16 @@ if __name__ == "__main__":
     space_id = os.getenv("SPACE_ID")
 
     if space_host:
-        print(f"✅ SPACE_HOST found: {space_host}")
+        print(f"\u2705 SPACE_HOST found: {space_host}")
         print(f"   Runtime URL should be: https://{space_host}.hf.space")
     else:
-        print("ℹ️  SPACE_HOST not found.")
+        print("\u2139\ufe0f  SPACE_HOST not found.")
 
     if space_id:
-        print(f"✅ SPACE_ID found: {space_id}")
+        print(f"\u2705 SPACE_ID found: {space_id}")
         print(f"   Repo URL: https://huggingface.co/spaces/{space_id}")
     else:
-        print("ℹ️  SPACE_ID not found.")
+        print("\u2139\ufe0f  SPACE_ID not found.")
 
     print("-"*(60 + len(" App Starting ")) + "\n")
     demo.launch(debug=True, share=False)