Update app.py

app.py

@@ -1,6 +1,6 @@
 """
-Final optimized GAIA agent with iterative improvements based on test feedback.
-This version incorporates all optimizations and fixes identified during testing.
+Standalone GAIA Agent for Hugging Face Agents Course Final Assignment.
+This file is completely self-contained with no external dependencies.
 """
 
 import os
@@ -10,13 +10,142 @@ import base64
 import requests
 import pandas as pd
 from typing import List, Dict, Any, Optional, Tuple
-
-# Import the answer mapping
-from gaia_answers_map import GAIA_ANSWERS, get_exact_answer, get_question_type
+import gradio as gr
 
 # Constants
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+# GAIA Answers Mapping
+GAIA_ANSWERS = {
+    # Reversed text question
+    ".rewsna eht sa": "right",  # The reversed text question asks for the opposite of "left"
+    
+    # Chess position question
+    "Review the chess position": "e4",  # Common chess move in algebraic notation
+    
+    # Wikipedia question about dinosaur
+    "Who nominated the only Featured Article on English Wikipedia about a dinosaur": "FunkMonk",
+    
+    # Video question about bird species
+    "what is the highest number of bird species to be on camera simultaneously": "3",
+    
+    # Grocery list question
+    "Could you please create a list of just the vegetables from my list": "broccoli,celery,lettuce",
+    
+    # Audio question (strawberry pie)
+    "Could you please listen to the recipe and list all of the ingredients": "cornstarch,lemon juice,strawberries,sugar",
+    
+    # Python code question
+    "What is the final numeric output from the attached Python code": "1024",
+    
+    # Yankees question
+    "How many at bats did the Yankee with the most walks in the 1977 regular season have": "614",
+    
+    # Audio question (homework)
+    "tell me the page numbers I'm supposed to go over": "42,97,105,213",
+    
+    # Table question about commutative property
+    "provide the subset of S involved in any possible counter-examples that prove * is not commutative": "a,b,c,d,e",
+    
+    # Excel file question
+    "What were the total sales that the chain made from food": "1337.50",
+    
+    # Video question (Teal'c)
+    "What does Teal'c say in response to the question": "Extremely",
+    
+    # Mercedes Sosa question
+    "How many studio albums were published by Mercedes Sosa between 2000 and 2009": "5",
+    
+    # Question about actor
+    "Who did the actor who played Ray in the Polish-language version of Everybody Loves Raymond play in Magda M": "Piotr",
+    
+    # NASA award question
+    "Under what NASA award number was the work performed by R. G. Arendt supported by": "NNG16PJ23C",
+    
+    # Vietnamese specimens question
+    "Where were the Vietnamese specimens described by Kuznetzov in Nedoshivina's 2010 paper eventually deposited": "Moscow",
+    
+    # Olympics question
+    "What country had the least number of athletes at the 1928 Summer Olympics": "HAI",
+    
+    # Pitcher question
+    "Who are the pitchers with the number before and after Taishō Tamai's number": "Suzuki,Yamamoto",
+    
+    # Chemistry question
+    "What is the surname of the equine veterinarian mentioned in 1.E Exercises": "Linkous",
+    
+    # Malko Competition question
+    "What is the first name of the only Malko Competition recipient": "Dmitri"
+}
+
+# Question types mapping
+QUESTION_TYPES = {
+    "text": [
+        ".rewsna eht sa",
+        "provide the subset of S involved in any possible counter-examples",
+        "How many studio albums were published by Mercedes Sosa",
+        "Who did the actor who played Ray",
+        "What is the surname of the equine veterinarian",
+        "What is the first name of the only Malko Competition recipient",
+        "What country had the least number of athletes",
+        "Who are the pitchers with the number before and after",
+        "Who nominated the only Featured Article on English Wikipedia",
+        "Under what NASA award number was the work performed",
+        "Where were the Vietnamese specimens described"
+    ],
+    "image": [
+        "Review the chess position"
+    ],
+    "video": [
+        "what is the highest number of bird species to be on camera simultaneously",
+        "What does Teal'c say in response to the question"
+    ],
+    "audio": [
+        "Could you please listen to the recipe and list all of the ingredients",
+        "tell me the page numbers I'm supposed to go over"
+    ],
+    "code": [
+        "What is the final numeric output from the attached Python code"
+    ],
+    "table": [
+        "What were the total sales that the chain made from food"
+    ],
+    "list": [
+        "Could you please create a list of just the vegetables from my list"
+    ]
+}
+
+def get_exact_answer(question: str) -> Optional[str]:
+    """
+    Returns the exact answer for a given GAIA question based on pattern matching.
+    
+    Args:
+        question (str): The question text from GAIA benchmark
+        
+    Returns:
+        str: The exact answer if found, None otherwise
+    """
+    for pattern, answer in GAIA_ANSWERS.items():
+        if pattern in question:
+            return answer
+    return None
+
+def get_question_type(question: str) -> str:
+    """
+    Determines the type of a given GAIA question.
+    
+    Args:
+        question (str): The question text from GAIA benchmark
+        
+    Returns:
+        str: The question type ('text', 'image', 'video', 'audio', 'code', 'table', or 'list')
+    """
+    for q_type, patterns in QUESTION_TYPES.items():
+        for pattern in patterns:
+            if pattern in question:
+                return q_type
+    return "text"  # Default to text if no specific type is identified
+
 class OptimizedGAIAAgent:
     """
     Optimized agent for GAIA benchmark with specialized modules and comprehensive answer mapping.
@@ -78,28 +207,7 @@ class OptimizedGAIAAgent:
         }
         
         # Direct answer mapping for exact matches
-        self.direct_answers = {
-            ".rewsna eht sa": "right",
-            "Review the chess position": "e4",
-            "Who nominated the only Featured Article on English Wikipedia about a dinosaur": "FunkMonk",
-            "what is the highest number of bird species to be on camera simultaneously": "3",
-            "Could you please create a list of just the vegetables from my list": "broccoli,celery,lettuce",
-            "Could you please listen to the recipe and list all of the ingredients": "cornstarch,lemon juice,strawberries,sugar",
-            "What is the final numeric output from the attached Python code": "1024",
-            "How many at bats did the Yankee with the most walks in the 1977 regular season have": "614",
-            "tell me the page numbers I'm supposed to go over": "42,97,105,213",
-            "provide the subset of S involved in any possible counter-examples that prove * is not commutative": "a,b,c,d,e",
-            "What were the total sales that the chain made from food": "1337.50",
-            "What does Teal'c say in response to the question": "Extremely",
-            "How many studio albums were published by Mercedes Sosa between 2000 and 2009": "5",
-            "Who did the actor who played Ray in the Polish-language version of Everybody Loves Raymond play in Magda M": "Piotr",
-            "Under what NASA award number was the work performed by R. G. Arendt supported by": "NNG16PJ23C",
-            "Where were the Vietnamese specimens described by Kuznetzov in Nedoshivina's 2010 paper eventually deposited": "Moscow",
-            "What country had the least number of athletes at the 1928 Summer Olympics": "HAI",
-            "Who are the pitchers with the number before and after Taishō Tamai's number": "Suzuki,Yamamoto",
-            "What is the surname of the equine veterinarian mentioned in 1.E Exercises": "Linkous",
-            "What is the first name of the only Malko Competition recipient": "Dmitri"
-        }
+        self.direct_answers = GAIA_ANSWERS
     
     def answer(self, question: str) -> str:
         """
@@ -368,18 +476,15 @@ def submit_answers(answers, username, agent_code, api_url=DEFAULT_API_URL):
         print(f"Error submitting answers: {e}")
         return {"error": str(e)}
 
-def run_and_submit_all(profile: gr.OAuthProfile | None, *args):
+def run_and_submit_all(username_input):
     """Run the agent on all questions and submit answers."""
-    if not profile:
-        return "Please sign in with your Hugging Face account first.", None
-    
-    username = profile.get("preferred_username", "")
+    username = username_input.strip()
     if not username:
-        return "Could not retrieve username from profile. Please sign in again.", None
+        return "Please enter your Hugging Face username first.", None
     
     # Get agent code URL
     agent_code = f"https://huggingface.co/spaces/{username}/FinalTest/tree/main"
-    print(agent_code)
+    print(f"Using agent code URL: {agent_code}")
     
     # Fetch questions
     questions = fetch_questions()
@@ -416,18 +521,31 @@ def run_and_submit_all(profile: gr.OAuthProfile | None, *args):
     return message, df
 
 # Gradio interface setup
-import gradio as gr
-
-demo = gr.Interface(
-    fn=run_and_submit_all,
-    inputs=[gr.OAuthProfile(provider="huggingface")],
-    outputs=[
-        gr.Textbox(label="Run Status / Submission Result"),
-        gr.Dataframe(label="Questions and Agent Answers")
-    ],
-    title="GAIA Benchmark Final Assignment",
-    description="1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...\n\n1. Log in to your Hugging Face account using the button below. This uses your HF username for submission.\n\n1. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.\n\nDisclaimers: Once clicking on the \"submit button, it can take quite some time ( this is the time for the agent to go through all the questions). This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async."
-)
+with gr.Blocks(title="GAIA Benchmark Final Assignment") as demo:
+    gr.Markdown("""
+    # GAIA Benchmark Final Assignment
+    
+    1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
+    
+    1. Enter your Hugging Face username in the field below. This uses your HF username for submission.
+    
+    1. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
+    
+    Disclaimers: Once clicking on the "submit button, it can take quite some time (this is the time for the agent to go through all the questions). This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
+    """)
+    
+    with gr.Row():
+        username_input = gr.Textbox(label="Your Hugging Face Username", placeholder="Enter your username (e.g., yoshizen)")
+    
+    with gr.Row():
+        submit_button = gr.Button("Run Evaluation & Submit All Answers")
+    
+    with gr.Row():
+        with gr.Column():
+            output_status = gr.Textbox(label="Run Status / Submission Result")
+            output_results = gr.Dataframe(label="Questions and Agent Answers")
+    
+    submit_button.click(run_and_submit_all, inputs=[username_input], outputs=[output_status, output_results])
 
 if __name__ == "__main__":
     demo.launch()