Spaces:
Sleeping
Sleeping
Added CodeAgent with web search and scraping tools, added GAIA prompt.
Browse files
app.py
CHANGED
@@ -1,23 +1,23 @@
|
|
|
|
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
-
import inspect
|
5 |
import pandas as pd
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
# (Keep Constants as is)
|
8 |
# --- Constants ---
|
9 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
10 |
|
11 |
-
|
12 |
-
|
13 |
-
|
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 |
"""
|
@@ -30,6 +30,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
30 |
if profile:
|
31 |
username= f"{profile.username}"
|
32 |
print(f"User logged in: {username}")
|
|
|
33 |
else:
|
34 |
print("User not logged in.")
|
35 |
return "Please Login to Hugging Face with the button.", None
|
@@ -40,31 +41,44 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
40 |
|
41 |
# 1. Instantiate Agent ( modify this part to create your agent)
|
42 |
try:
|
43 |
-
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
45 |
print(f"Error instantiating agent: {e}")
|
46 |
return f"Error initializing agent: {e}", None
|
47 |
-
|
|
|
|
|
48 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
49 |
print(agent_code)
|
50 |
|
51 |
# 2. Fetch Questions
|
52 |
print(f"Fetching questions from: {questions_url}")
|
|
|
53 |
try:
|
54 |
response = requests.get(questions_url, timeout=15)
|
55 |
response.raise_for_status()
|
56 |
questions_data = response.json()
|
|
|
57 |
if not questions_data:
|
58 |
-
|
59 |
-
|
|
|
60 |
print(f"Fetched {len(questions_data)} questions.")
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
except requests.exceptions.RequestException as e:
|
62 |
print(f"Error fetching questions: {e}")
|
63 |
return f"Error fetching questions: {e}", None
|
64 |
-
|
65 |
-
print(f"Error decoding JSON response from questions endpoint: {e}")
|
66 |
-
print(f"Response text: {response.text[:500]}")
|
67 |
-
return f"Error decoding server response for questions: {e}", None
|
68 |
except Exception as e:
|
69 |
print(f"An unexpected error occurred fetching questions: {e}")
|
70 |
return f"An unexpected error occurred fetching questions: {e}", None
|
@@ -72,20 +86,37 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
72 |
# 3. Run your Agent
|
73 |
results_log = []
|
74 |
answers_payload = []
|
|
|
75 |
print(f"Running agent on {len(questions_data)} questions...")
|
|
|
76 |
for item in questions_data:
|
77 |
task_id = item.get("task_id")
|
78 |
question_text = item.get("question")
|
|
|
79 |
if not task_id or question_text is None:
|
80 |
print(f"Skipping item with missing task_id or question: {item}")
|
81 |
continue
|
|
|
82 |
try:
|
83 |
-
submitted_answer = agent(
|
|
|
|
|
|
|
|
|
84 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
85 |
-
results_log.append({
|
|
|
|
|
|
|
|
|
|
|
86 |
except Exception as e:
|
87 |
print(f"Error running agent on task {task_id}: {e}")
|
88 |
-
results_log.append({
|
|
|
|
|
|
|
|
|
89 |
|
90 |
if not answers_payload:
|
91 |
print("Agent did not produce any answers to submit.")
|
@@ -112,27 +143,34 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
112 |
print("Submission successful.")
|
113 |
results_df = pd.DataFrame(results_log)
|
114 |
return final_status, results_df
|
|
|
115 |
except requests.exceptions.HTTPError as e:
|
116 |
error_detail = f"Server responded with status {e.response.status_code}."
|
|
|
117 |
try:
|
118 |
error_json = e.response.json()
|
119 |
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
|
|
120 |
except requests.exceptions.JSONDecodeError:
|
121 |
error_detail += f" Response: {e.response.text[:500]}"
|
|
|
122 |
status_message = f"Submission Failed: {error_detail}"
|
123 |
print(status_message)
|
124 |
results_df = pd.DataFrame(results_log)
|
125 |
return status_message, results_df
|
|
|
126 |
except requests.exceptions.Timeout:
|
127 |
status_message = "Submission Failed: The request timed out."
|
128 |
print(status_message)
|
129 |
results_df = pd.DataFrame(results_log)
|
130 |
return status_message, results_df
|
|
|
131 |
except requests.exceptions.RequestException as e:
|
132 |
status_message = f"Submission Failed: Network error - {e}"
|
133 |
print(status_message)
|
134 |
results_df = pd.DataFrame(results_log)
|
135 |
return status_message, results_df
|
|
|
136 |
except Exception as e:
|
137 |
status_message = f"An unexpected error occurred during submission: {e}"
|
138 |
print(status_message)
|
@@ -166,7 +204,7 @@ with gr.Blocks() as demo:
|
|
166 |
# Removed max_rows=10 from DataFrame constructor
|
167 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
168 |
|
169 |
-
run_button.click(
|
170 |
fn=run_and_submit_all,
|
171 |
outputs=[status_output, results_table]
|
172 |
)
|
|
|
1 |
+
'''HuggingFace Agents course final project GAIA agent benchmark'''
|
2 |
+
|
3 |
import os
|
4 |
import gradio as gr
|
5 |
import requests
|
|
|
6 |
import pandas as pd
|
7 |
+
from smolagents import (
|
8 |
+
CodeAgent,
|
9 |
+
DuckDuckGoSearchTool,
|
10 |
+
InferenceClientModel,
|
11 |
+
VisitWebpageTool
|
12 |
+
)
|
13 |
|
14 |
# (Keep Constants as is)
|
15 |
# --- Constants ---
|
16 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
17 |
|
18 |
+
INSTRUCTIONS = '''
|
19 |
+
YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.'''
|
20 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
def run_and_submit_all( profile: gr.OAuthProfile | None):
|
23 |
"""
|
|
|
30 |
if profile:
|
31 |
username= f"{profile.username}"
|
32 |
print(f"User logged in: {username}")
|
33 |
+
|
34 |
else:
|
35 |
print("User not logged in.")
|
36 |
return "Please Login to Hugging Face with the button.", None
|
|
|
41 |
|
42 |
# 1. Instantiate Agent ( modify this part to create your agent)
|
43 |
try:
|
44 |
+
model = InferenceClientModel("Qwen/Qwen2.5-Coder-32B-Instruct")
|
45 |
+
agent = CodeAgent(
|
46 |
+
tools=[DuckDuckGoSearchTool(), VisitWebpageTool()],
|
47 |
+
model=model
|
48 |
+
)
|
49 |
+
|
50 |
+
except Exception as e: # pyline: disable=W0703
|
51 |
print(f"Error instantiating agent: {e}")
|
52 |
return f"Error initializing agent: {e}", None
|
53 |
+
|
54 |
+
# In the case of an app running as a hugging Face space, this link points toward your
|
55 |
+
# codebase ( useful for others so please keep it public)
|
56 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
57 |
print(agent_code)
|
58 |
|
59 |
# 2. Fetch Questions
|
60 |
print(f"Fetching questions from: {questions_url}")
|
61 |
+
|
62 |
try:
|
63 |
response = requests.get(questions_url, timeout=15)
|
64 |
response.raise_for_status()
|
65 |
questions_data = response.json()
|
66 |
+
|
67 |
if not questions_data:
|
68 |
+
print("Fetched questions list is empty.")
|
69 |
+
return "Fetched questions list is empty or invalid format.", None
|
70 |
+
|
71 |
print(f"Fetched {len(questions_data)} questions.")
|
72 |
+
|
73 |
+
except requests.exceptions.JSONDecodeError as e:
|
74 |
+
print(f"Error decoding JSON response from questions endpoint: {e}")
|
75 |
+
print(f"Response text: {response.text[:500]}")
|
76 |
+
return f"Error decoding server response for questions: {e}", None
|
77 |
+
|
78 |
except requests.exceptions.RequestException as e:
|
79 |
print(f"Error fetching questions: {e}")
|
80 |
return f"Error fetching questions: {e}", None
|
81 |
+
|
|
|
|
|
|
|
82 |
except Exception as e:
|
83 |
print(f"An unexpected error occurred fetching questions: {e}")
|
84 |
return f"An unexpected error occurred fetching questions: {e}", None
|
|
|
86 |
# 3. Run your Agent
|
87 |
results_log = []
|
88 |
answers_payload = []
|
89 |
+
|
90 |
print(f"Running agent on {len(questions_data)} questions...")
|
91 |
+
|
92 |
for item in questions_data:
|
93 |
task_id = item.get("task_id")
|
94 |
question_text = item.get("question")
|
95 |
+
|
96 |
if not task_id or question_text is None:
|
97 |
print(f"Skipping item with missing task_id or question: {item}")
|
98 |
continue
|
99 |
+
|
100 |
try:
|
101 |
+
submitted_answer = agent.run(
|
102 |
+
INSTRUCTIONS,
|
103 |
+
additional_args={'user_prompt': question_text}
|
104 |
+
)
|
105 |
+
|
106 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
107 |
+
results_log.append({
|
108 |
+
"Task ID": task_id,
|
109 |
+
"Question": question_text,
|
110 |
+
"Submitted Answer": submitted_answer
|
111 |
+
})
|
112 |
+
|
113 |
except Exception as e:
|
114 |
print(f"Error running agent on task {task_id}: {e}")
|
115 |
+
results_log.append({
|
116 |
+
"Task ID": task_id,
|
117 |
+
"Question": question_text,
|
118 |
+
"Submitted Answer": f"AGENT ERROR: {e}"
|
119 |
+
})
|
120 |
|
121 |
if not answers_payload:
|
122 |
print("Agent did not produce any answers to submit.")
|
|
|
143 |
print("Submission successful.")
|
144 |
results_df = pd.DataFrame(results_log)
|
145 |
return final_status, results_df
|
146 |
+
|
147 |
except requests.exceptions.HTTPError as e:
|
148 |
error_detail = f"Server responded with status {e.response.status_code}."
|
149 |
+
|
150 |
try:
|
151 |
error_json = e.response.json()
|
152 |
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
153 |
+
|
154 |
except requests.exceptions.JSONDecodeError:
|
155 |
error_detail += f" Response: {e.response.text[:500]}"
|
156 |
+
|
157 |
status_message = f"Submission Failed: {error_detail}"
|
158 |
print(status_message)
|
159 |
results_df = pd.DataFrame(results_log)
|
160 |
return status_message, results_df
|
161 |
+
|
162 |
except requests.exceptions.Timeout:
|
163 |
status_message = "Submission Failed: The request timed out."
|
164 |
print(status_message)
|
165 |
results_df = pd.DataFrame(results_log)
|
166 |
return status_message, results_df
|
167 |
+
|
168 |
except requests.exceptions.RequestException as e:
|
169 |
status_message = f"Submission Failed: Network error - {e}"
|
170 |
print(status_message)
|
171 |
results_df = pd.DataFrame(results_log)
|
172 |
return status_message, results_df
|
173 |
+
|
174 |
except Exception as e:
|
175 |
status_message = f"An unexpected error occurred during submission: {e}"
|
176 |
print(status_message)
|
|
|
204 |
# Removed max_rows=10 from DataFrame constructor
|
205 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
206 |
|
207 |
+
run_button.click( # pylint: disable=E1101
|
208 |
fn=run_and_submit_all,
|
209 |
outputs=[status_output, results_table]
|
210 |
)
|