Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,196 +1,134 @@
|
|
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("
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
21 |
|
22 |
-
def
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
print("User not logged in.")
|
35 |
return "Please Login to Hugging Face with the button.", None
|
|
|
36 |
|
37 |
-
|
38 |
-
|
39 |
-
submit_url = f"{api_url}/submit"
|
40 |
|
41 |
-
# 1. Instantiate Agent ( modify this part to create your agent)
|
42 |
try:
|
43 |
-
agent =
|
44 |
except Exception as e:
|
45 |
-
print(f"Error instantiating agent: {e}")
|
46 |
return f"Error initializing agent: {e}", None
|
47 |
-
# In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
|
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 |
-
|
55 |
-
|
56 |
-
|
57 |
-
if not questions_data:
|
58 |
-
print("Fetched questions list is empty.")
|
59 |
-
return "Fetched questions list is empty or invalid format.", None
|
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 |
-
except requests.exceptions.JSONDecodeError as e:
|
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 |
-
|
70 |
-
return f"An unexpected error occurred fetching questions: {e}", None
|
71 |
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
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 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
# 4. Prepare Submission
|
95 |
-
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
|
96 |
-
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
97 |
-
print(status_update)
|
98 |
-
|
99 |
-
# 5. Submit
|
100 |
-
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
101 |
try:
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
f"Submission Successful
|
107 |
-
f"
|
108 |
-
f"Overall Score: {result_data.get('score', 'N/A')}% "
|
109 |
-
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
|
110 |
-
f"Message: {result_data.get('message', 'No message received.')}"
|
111 |
)
|
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 |
-
|
138 |
-
|
139 |
-
results_df = pd.DataFrame(results_log)
|
140 |
-
return status_message, results_df
|
141 |
-
|
142 |
|
143 |
-
# ---
|
144 |
with gr.Blocks() as demo:
|
145 |
-
gr.Markdown("
|
146 |
-
gr.Markdown(
|
147 |
-
"""
|
148 |
-
**Instructions:**
|
149 |
-
|
150 |
-
1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
|
151 |
-
2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
|
152 |
-
3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
|
153 |
-
|
154 |
-
---
|
155 |
-
**Disclaimers:**
|
156 |
-
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).
|
157 |
-
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.
|
158 |
-
"""
|
159 |
-
)
|
160 |
-
|
161 |
gr.LoginButton()
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
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 |
-
)
|
173 |
|
174 |
if __name__ == "__main__":
|
175 |
-
|
176 |
-
# Check for SPACE_HOST and SPACE_ID at startup for information
|
177 |
-
space_host_startup = os.getenv("SPACE_HOST")
|
178 |
-
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
179 |
-
|
180 |
-
if space_host_startup:
|
181 |
-
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
182 |
-
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
183 |
-
else:
|
184 |
-
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
185 |
-
|
186 |
-
if space_id_startup: # Print repo URLs if SPACE_ID is found
|
187 |
-
print(f"✅ SPACE_ID found: {space_id_startup}")
|
188 |
-
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
189 |
-
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
190 |
-
else:
|
191 |
-
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
192 |
-
|
193 |
-
print("-"*(60 + len(" App Starting ")) + "\n")
|
194 |
-
|
195 |
-
print("Launching Gradio Interface for Basic Agent Evaluation...")
|
196 |
-
demo.launch(debug=True, share=False)
|
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
import requests
|
|
|
4 |
import pandas as pd
|
5 |
+
from transformers import pipeline
|
6 |
+
from duckduckgo_search import ddg
|
7 |
|
|
|
8 |
# --- Constants ---
|
9 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
10 |
|
11 |
+
# --- Tool Definitions ---
|
12 |
+
class WebSearchTool:
|
13 |
+
"""A simple open-source web search tool using DuckDuckGo."""
|
14 |
+
def search(self, query: str) -> list[str]:
|
15 |
+
# Perform a DuckDuckGo search and return top 3 titles
|
16 |
+
try:
|
17 |
+
results = ddg(query, max_results=3)
|
18 |
+
if not results:
|
19 |
+
return []
|
20 |
+
return [item.get("title", "").strip() for item in results]
|
21 |
+
except Exception as e:
|
22 |
+
print(f"DuckDuckGo search error: {e}")
|
23 |
+
return []
|
24 |
+
|
25 |
+
class CalculatorTool:
|
26 |
+
"""Evaluates simple arithmetic expressions."""
|
27 |
+
def calculate(self, expression: str) -> float:
|
28 |
+
# WARNING: using eval; in production, use a safe parser
|
29 |
+
return eval(expression, {}, {})
|
30 |
+
|
31 |
+
# --- Agent Definition ---
|
32 |
+
class ToolsAgent:
|
33 |
def __init__(self):
|
34 |
+
print("ToolsAgent initialized.")
|
35 |
+
# Initialize LLM for general reasoning
|
36 |
+
self.llm = pipeline(
|
37 |
+
"text-generation", model="gpt2", tokenizer="gpt2", return_full_text=False
|
38 |
+
)
|
39 |
+
# Initialize tools
|
40 |
+
self.tools = {
|
41 |
+
"search": WebSearchTool(),
|
42 |
+
"calc": CalculatorTool(),
|
43 |
+
}
|
44 |
|
45 |
+
def __call__(self, question: str) -> str:
|
46 |
+
print(f"Agent received question: {question}")
|
47 |
+
q_lower = question.lower().strip()
|
48 |
+
# 1) If calculation question
|
49 |
+
if any(op in q_lower for op in ["+", "-", "*", "/"]):
|
50 |
+
try:
|
51 |
+
result = self.tools["calc"].calculate(q_lower)
|
52 |
+
return str(result)
|
53 |
+
except Exception as e:
|
54 |
+
print(f"Calc error: {e}")
|
55 |
+
# 2) If search instruction
|
56 |
+
if q_lower.startswith("search") or "find" in q_lower:
|
57 |
+
try:
|
58 |
+
titles = self.tools["search"].search(question)
|
59 |
+
# Return only the first word of the top title
|
60 |
+
return titles[0].split()[0] if titles else ""
|
61 |
+
except Exception as e:
|
62 |
+
print(f"Search error: {e}")
|
63 |
+
# 3) Fallback to LLM
|
64 |
+
try:
|
65 |
+
out = self.llm(question, max_length=50, num_return_sequences=1)[0]["generated_text"].strip()
|
66 |
+
# Return only the first token (single word/number)
|
67 |
+
return out.split()[0]
|
68 |
+
except Exception as e:
|
69 |
+
print(f"LLM error: {e}")
|
70 |
+
return ""
|
71 |
|
72 |
+
# --- Evaluation and Submission Logic ---
|
73 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
74 |
+
space_id = os.getenv("SPACE_ID")
|
75 |
+
if not profile:
|
|
|
76 |
return "Please Login to Hugging Face with the button.", None
|
77 |
+
username = profile.username.strip()
|
78 |
|
79 |
+
questions_url = f"{DEFAULT_API_URL}/questions"
|
80 |
+
submit_url = f"{DEFAULT_API_URL}/submit"
|
|
|
81 |
|
|
|
82 |
try:
|
83 |
+
agent = ToolsAgent()
|
84 |
except Exception as e:
|
|
|
85 |
return f"Error initializing agent: {e}", None
|
|
|
|
|
|
|
86 |
|
|
|
|
|
87 |
try:
|
88 |
+
resp = requests.get(questions_url, timeout=15)
|
89 |
+
resp.raise_for_status()
|
90 |
+
qs = resp.json()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
except Exception as e:
|
92 |
+
return f"Error fetching questions: {e}", None
|
|
|
93 |
|
94 |
+
answers, log = [], []
|
95 |
+
for item in qs:
|
96 |
+
tid = item.get("task_id")
|
97 |
+
q = item.get("question")
|
98 |
+
if not tid or q is None:
|
|
|
|
|
|
|
|
|
99 |
continue
|
100 |
+
ans = agent(q)
|
101 |
+
answers.append({"task_id": tid, "submitted_answer": ans})
|
102 |
+
log.append({"Task ID": tid, "Question": q, "Submitted Answer": ans})
|
103 |
+
|
104 |
+
if not answers:
|
105 |
+
return "No answers generated.", pd.DataFrame(log)
|
106 |
+
|
107 |
+
payload = {
|
108 |
+
"username": username,
|
109 |
+
"agent_code": f"https://huggingface.co/spaces/{space_id}/tree/main",
|
110 |
+
"answers": answers
|
111 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
try:
|
113 |
+
res = requests.post(submit_url, json=payload, timeout=60)
|
114 |
+
res.raise_for_status()
|
115 |
+
data = res.json()
|
116 |
+
status = (
|
117 |
+
f"Submission Successful! User: {data.get('username')} "
|
118 |
+
f"Score: {data.get('score', 'N/A')}%"
|
|
|
|
|
|
|
119 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
except Exception as e:
|
121 |
+
status = f"Submission Failed: {e}"
|
122 |
+
return status, pd.DataFrame(log)
|
|
|
|
|
|
|
123 |
|
124 |
+
# --- Gradio Interface ---
|
125 |
with gr.Blocks() as demo:
|
126 |
+
gr.Markdown("## Agents Course: DuckDuckGo Search + Calculator Agent")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
gr.LoginButton()
|
128 |
+
run_btn = gr.Button("Run Evaluation & Submit All Answers")
|
129 |
+
out_txt = gr.Textbox(label="Result", lines=3)
|
130 |
+
out_df = gr.DataFrame(label="Log", wrap=True)
|
131 |
+
run_btn.click(fn=run_and_submit_all, outputs=[out_txt, out_df])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
|
133 |
if __name__ == "__main__":
|
134 |
+
demo.launch(debug=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|