Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,202 +2,111 @@ import os
|
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
import pandas as pd
|
5 |
-
import
|
6 |
-
from smolagents import
|
7 |
-
from smolagents import DuckDuckGoSearchTool
|
8 |
|
9 |
-
|
10 |
-
# --- Constants ---
|
11 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
12 |
|
13 |
-
#
|
14 |
-
class
|
15 |
-
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
17 |
FINAL ANSWER: [YOUR FINAL ANSWER].
|
18 |
YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list
|
19 |
-
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,
|
20 |
-
|
21 |
-
"""
|
22 |
-
|
23 |
-
def __init__(self, api_key=None, model="gpt-4o"):
|
24 |
-
if api_key is None:
|
25 |
-
api_key = os.getenv("OPENAI_API_KEY")
|
26 |
-
if not api_key:
|
27 |
-
raise RuntimeError("OPENAI_API_KEY not set in environment.")
|
28 |
-
openai.api_key = api_key
|
29 |
-
self.model = model
|
30 |
-
|
31 |
-
# Add DuckDuckGoSearchTool to tools
|
32 |
-
self.tools = [DuckDuckGoSearchTool()]
|
33 |
-
self.agent = CodeAgent(model=self.model, tools=self.tools)
|
34 |
|
35 |
def __call__(self, question: str) -> str:
|
36 |
-
|
37 |
-
try:
|
38 |
-
full_prompt = self.SYSTEM_PROMPT.strip() + "\n\nQuestion: " + question
|
39 |
-
answer = self.agent.run(full_prompt)
|
40 |
-
print(f"SmolCodeAgent answer (first 100 chars): {answer[:100]}...")
|
41 |
-
return answer
|
42 |
-
except Exception as e:
|
43 |
-
error_msg = f"AGENT ERROR: {e}"
|
44 |
-
print(error_msg)
|
45 |
-
return error_msg
|
46 |
|
47 |
-
#
|
48 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
49 |
space_id = os.getenv("SPACE_ID")
|
50 |
-
|
51 |
if profile:
|
52 |
username = profile.username
|
53 |
print(f"User logged in: {username}")
|
54 |
else:
|
55 |
-
print("User not logged in.")
|
56 |
return "Please Login to Hugging Face with the button.", None
|
57 |
|
58 |
-
|
59 |
-
questions_url = f"{
|
60 |
-
submit_url = f"{
|
61 |
|
62 |
-
# Instantiate your SmolCodeAgent here
|
63 |
try:
|
64 |
-
agent =
|
65 |
except Exception as e:
|
66 |
-
print(f"Error instantiating agent: {e}")
|
67 |
return f"Error initializing agent: {e}", None
|
68 |
|
69 |
-
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
70 |
-
print(f"Agent code URL: {agent_code}")
|
71 |
-
|
72 |
# Fetch Questions
|
73 |
-
print(f"Fetching questions from: {questions_url}")
|
74 |
try:
|
75 |
-
|
76 |
-
|
77 |
-
questions_data =
|
78 |
-
if not questions_data:
|
79 |
-
print("Fetched questions list is empty.")
|
80 |
-
return "Fetched questions list is empty or invalid format.", None
|
81 |
-
print(f"Fetched {len(questions_data)} questions.")
|
82 |
except Exception as e:
|
83 |
-
|
84 |
-
return f"Error fetching questions: {e}", None
|
85 |
|
86 |
-
# Run the agent on each question
|
87 |
results_log = []
|
88 |
answers_payload = []
|
89 |
-
|
90 |
for item in questions_data:
|
91 |
task_id = item.get("task_id")
|
92 |
-
|
93 |
-
if not task_id or
|
94 |
-
print(f"Skipping item with missing task_id or question: {item}")
|
95 |
continue
|
96 |
try:
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
except Exception as e:
|
101 |
-
|
102 |
-
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
103 |
|
104 |
if not answers_payload:
|
105 |
-
|
106 |
-
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
107 |
|
108 |
-
|
109 |
-
|
110 |
-
|
|
|
|
|
111 |
|
112 |
-
# Submit answers
|
113 |
try:
|
114 |
-
|
115 |
-
|
116 |
-
result_data =
|
117 |
-
|
118 |
f"Submission Successful!\n"
|
119 |
f"User: {result_data.get('username')}\n"
|
120 |
-
f"
|
121 |
-
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')}
|
122 |
-
f"Message: {result_data.get('message', '
|
123 |
)
|
124 |
-
|
125 |
-
results_df = pd.DataFrame(results_log)
|
126 |
-
return final_status, results_df
|
127 |
-
except requests.exceptions.HTTPError as e:
|
128 |
-
error_detail = f"Server responded with status {e.response.status_code}."
|
129 |
-
try:
|
130 |
-
error_json = e.response.json()
|
131 |
-
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
132 |
-
except Exception:
|
133 |
-
error_detail += f" Response: {e.response.text[:500]}"
|
134 |
-
status_message = f"Submission Failed: {error_detail}"
|
135 |
-
print(status_message)
|
136 |
-
results_df = pd.DataFrame(results_log)
|
137 |
-
return status_message, results_df
|
138 |
-
except requests.exceptions.Timeout:
|
139 |
-
status_message = "Submission Failed: The request timed out."
|
140 |
-
print(status_message)
|
141 |
-
results_df = pd.DataFrame(results_log)
|
142 |
-
return status_message, results_df
|
143 |
-
except requests.exceptions.RequestException as e:
|
144 |
-
status_message = f"Submission Failed: Network error - {e}"
|
145 |
-
print(status_message)
|
146 |
-
results_df = pd.DataFrame(results_log)
|
147 |
-
return status_message, results_df
|
148 |
except Exception as e:
|
149 |
-
|
150 |
-
print(status_message)
|
151 |
-
results_df = pd.DataFrame(results_log)
|
152 |
-
return status_message, results_df
|
153 |
-
|
154 |
|
155 |
-
#
|
156 |
with gr.Blocks() as demo:
|
157 |
-
gr.Markdown("#
|
158 |
-
gr.Markdown(
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
Submissions may take time while the agent processes questions.
|
167 |
-
"""
|
168 |
-
)
|
169 |
-
|
170 |
-
login_btn = gr.LoginButton()
|
171 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
172 |
-
status_output = gr.Textbox(label="
|
173 |
-
results_table = gr.DataFrame(label="
|
174 |
|
175 |
-
|
176 |
-
run_button.click(
|
177 |
-
fn=run_and_submit_all,
|
178 |
-
inputs=[login_btn],
|
179 |
-
outputs=[status_output, results_table]
|
180 |
-
)
|
181 |
|
182 |
if __name__ == "__main__":
|
183 |
-
print("
|
184 |
-
|
185 |
-
space_id_startup = os.getenv("SPACE_ID")
|
186 |
-
|
187 |
-
if space_host_startup:
|
188 |
-
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
189 |
-
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
190 |
-
else:
|
191 |
-
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
192 |
-
|
193 |
-
if space_id_startup:
|
194 |
-
print(f"✅ SPACE_ID found: {space_id_startup}")
|
195 |
-
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
196 |
-
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
197 |
-
else:
|
198 |
-
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
199 |
-
|
200 |
-
print("-"*(60 + len(" App Starting ")) + "\n")
|
201 |
-
|
202 |
-
print("Launching Gradio Interface for Basic Agent Evaluation...")
|
203 |
-
demo.launch(debug=True, share=False)
|
|
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
import pandas as pd
|
5 |
+
from smolagents import CodeAgent, DuckDuckGoSearchTool
|
6 |
+
from smolagents.models import OpenAIServerModel
|
|
|
7 |
|
8 |
+
# Constants
|
|
|
9 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
10 |
|
11 |
+
# === Define the Smol Agent ===
|
12 |
+
class MyAgent:
|
13 |
+
def __init__(self):
|
14 |
+
self.model = OpenAIServerModel() # Assumes OpenAI-compatible API via env vars
|
15 |
+
self.agent = CodeAgent(
|
16 |
+
tools=[DuckDuckGoSearchTool()],
|
17 |
+
model=self.model,
|
18 |
+
system_message="""You are a general AI assistant. I will ask you a question.
|
19 |
+
Report your thoughts, and finish your answer with the following template:
|
20 |
FINAL ANSWER: [YOUR FINAL ANSWER].
|
21 |
YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list
|
22 |
+
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."""
|
23 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
def __call__(self, question: str) -> str:
|
26 |
+
return self.agent.run(question)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
|
28 |
+
# === Submission Logic ===
|
29 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
30 |
space_id = os.getenv("SPACE_ID")
|
|
|
31 |
if profile:
|
32 |
username = profile.username
|
33 |
print(f"User logged in: {username}")
|
34 |
else:
|
|
|
35 |
return "Please Login to Hugging Face with the button.", None
|
36 |
|
37 |
+
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
38 |
+
questions_url = f"{DEFAULT_API_URL}/questions"
|
39 |
+
submit_url = f"{DEFAULT_API_URL}/submit"
|
40 |
|
|
|
41 |
try:
|
42 |
+
agent = MyAgent()
|
43 |
except Exception as e:
|
|
|
44 |
return f"Error initializing agent: {e}", None
|
45 |
|
|
|
|
|
|
|
46 |
# Fetch Questions
|
|
|
47 |
try:
|
48 |
+
res = requests.get(questions_url, timeout=15)
|
49 |
+
res.raise_for_status()
|
50 |
+
questions_data = res.json()
|
|
|
|
|
|
|
|
|
51 |
except Exception as e:
|
52 |
+
return f"Failed to fetch questions: {e}", None
|
|
|
53 |
|
|
|
54 |
results_log = []
|
55 |
answers_payload = []
|
56 |
+
|
57 |
for item in questions_data:
|
58 |
task_id = item.get("task_id")
|
59 |
+
question = item.get("question")
|
60 |
+
if not task_id or question is None:
|
|
|
61 |
continue
|
62 |
try:
|
63 |
+
answer = agent(question)
|
64 |
+
results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": answer})
|
65 |
+
answers_payload.append({"task_id": task_id, "submitted_answer": answer})
|
66 |
except Exception as e:
|
67 |
+
results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": f"ERROR: {e}"})
|
|
|
68 |
|
69 |
if not answers_payload:
|
70 |
+
return "No answers generated.", pd.DataFrame(results_log)
|
|
|
71 |
|
72 |
+
submission_data = {
|
73 |
+
"username": username,
|
74 |
+
"agent_code": agent_code,
|
75 |
+
"answers": answers_payload
|
76 |
+
}
|
77 |
|
|
|
78 |
try:
|
79 |
+
res = requests.post(submit_url, json=submission_data, timeout=60)
|
80 |
+
res.raise_for_status()
|
81 |
+
result_data = res.json()
|
82 |
+
summary = (
|
83 |
f"Submission Successful!\n"
|
84 |
f"User: {result_data.get('username')}\n"
|
85 |
+
f"Score: {result_data.get('score', '?')}% "
|
86 |
+
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')})\n"
|
87 |
+
f"Message: {result_data.get('message', '')}"
|
88 |
)
|
89 |
+
return summary, pd.DataFrame(results_log)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
except Exception as e:
|
91 |
+
return f"Submission failed: {e}", pd.DataFrame(results_log)
|
|
|
|
|
|
|
|
|
92 |
|
93 |
+
# === Gradio UI ===
|
94 |
with gr.Blocks() as demo:
|
95 |
+
gr.Markdown("# Agent Evaluation Runner (SmolAgents)")
|
96 |
+
gr.Markdown("""
|
97 |
+
**Instructions:**
|
98 |
+
1. Clone this space and customize your agent.
|
99 |
+
2. Log in with Hugging Face.
|
100 |
+
3. Click 'Run Evaluation' to answer and submit.
|
101 |
+
""")
|
102 |
+
|
103 |
+
gr.LoginButton()
|
|
|
|
|
|
|
|
|
|
|
104 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
105 |
+
status_output = gr.Textbox(label="Status", lines=4, interactive=False)
|
106 |
+
results_table = gr.DataFrame(label="Results", wrap=True)
|
107 |
|
108 |
+
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
|
|
|
|
|
|
|
|
|
|
109 |
|
110 |
if __name__ == "__main__":
|
111 |
+
print("Launching...")
|
112 |
+
demo.launch(debug=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|