anshullm commited on
Commit
7ede86d
·
verified ·
1 Parent(s): 81917a3

Update GAIA agent

Browse files

Improve the agent logic and add web search capabilities for GAIA evaluation.

Files changed (1) hide show
  1. app.py +40 -5
app.py CHANGED
@@ -3,6 +3,7 @@ import gradio as gr
3
  import requests
4
  import inspect
5
  import pandas as pd
 
6
 
7
  # (Keep Constants as is)
8
  # --- Constants ---
@@ -12,12 +13,46 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
12
  # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
13
  class BasicAgent:
14
  def __init__(self):
15
- print("BasicAgent initialized.")
 
 
 
 
 
 
 
 
 
 
 
16
  def __call__(self, question: str) -> str:
17
- print(f"Agent received question (first 50 chars): {question[:50]}...")
18
- fixed_answer = "This is a default answer."
19
- print(f"Agent returning fixed answer: {fixed_answer}")
20
- return fixed_answer
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
 
22
  def run_and_submit_all( profile: gr.OAuthProfile | None):
23
  """
 
3
  import requests
4
  import inspect
5
  import pandas as pd
6
+ from smolagents import CodeAgent, HfApiModel, DuckDuckGoSearchTool
7
 
8
  # (Keep Constants as is)
9
  # --- Constants ---
 
13
  # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
14
  class BasicAgent:
15
  def __init__(self):
16
+ print("Initializing GAIA Agent...")
17
+
18
+ self.model = HfApiModel()
19
+
20
+ self.agent = CodeAgent(
21
+ tools=[
22
+ DuckDuckGoSearchTool(max_results=8),
23
+ ],
24
+ model=self.model,
25
+ max_steps=10,
26
+ )
27
+
28
  def __call__(self, question: str) -> str:
29
+ print(f"Agent received question: {question[:100]}...")
30
+
31
+ prompt = f"""
32
+ You are solving a GAIA benchmark question.
33
+
34
+ Solve the question carefully and completely.
35
+ Use web search whenever factual information or research is needed.
36
+
37
+ IMPORTANT:
38
+ - Return ONLY the final answer.
39
+ - Do not write "FINAL ANSWER".
40
+ - Do not add explanations.
41
+ - Follow the exact format requested by the question.
42
+
43
+ Question:
44
+ {question}
45
+ """
46
+
47
+ try:
48
+ result = self.agent.run(prompt)
49
+ answer = str(result).strip()
50
+ print(f"Agent returned: {answer}")
51
+ return answer
52
+
53
+ except Exception as e:
54
+ print(f"Agent error: {e}")
55
+ return f"ERROR: {e}"
56
 
57
  def run_and_submit_all( profile: gr.OAuthProfile | None):
58
  """