Update app.py

app.py

@@ -1,55 +1,44 @@
 import torch
 import os
-import gradio as gr
 import requests
-import inspect
+import gradio as gr
 import pandas as pd
+from io import StringIO
 from transformers import pipeline
 
-
-# (Keep Constants as is)
-# --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class HybridAgent:
     def __init__(self):
-        # Load a small language model for answering text questions
         self.llm = pipeline("text-generation", model="google/flan-t5-small")
 
     def __call__(self, question: str, task_id: str = None) -> str:
-        """
-        Decide how to answer:
-        - If there is a file, fetch it and try to compute the answer with pandas
-        - Otherwise, use the language model
-        """
-        # --- Step 1: Handle file-based questions ---
+        print("=== Running on task:", task_id)
+        print("Question:", question)
+
         if task_id:
             resp = requests.get(f"{DEFAULT_API_URL}/files/{task_id}")
             if resp.status_code == 200:
                 file_content = resp.text
+                print("File content preview:", file_content[:200])
                 try:
-                    # Try reading it as CSV
                     df = pd.read_csv(StringIO(file_content))
-
-                    # Very simple heuristics: if question asks for "average"
-                    if "average" in question.lower() or "mean" in question.lower():
+                    q_lower = question.lower()
+                    if "average" in q_lower or "mean" in q_lower:
                         return str(df.iloc[:, 0].mean())
-                    if "sum" in question.lower() or "total" in question.lower():
+                    if "sum" in q_lower or "total" in q_lower:
                         return str(df.iloc[:, 0].sum())
-                    
-                    # Otherwise just return first few rows (fallback)
                     return df.head().to_string(index=False)
-                except Exception:
-                    # If not CSV or fails, just give the raw file text
+                except Exception as e:
+                    print("Error reading CSV, returning raw:", e)
                     return file_content.strip()
 
-        # --- Step 2: Handle text-only questions with LLM ---
         prompt = f"Answer this question briefly and only with the final answer:\n{question}"
         result = self.llm(prompt, max_length=50)[0]["generated_text"]
-        return result.strip()
-
+        cleaned = result.replace("Final Answer:", "").replace("Answer:", "").strip()
+        print("Raw LLM response:", result)
+        print("Cleaned answer:", cleaned)
+        return cleaned
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """