Update app.py
Browse files
app.py
CHANGED
|
@@ -2,10 +2,10 @@ import os
|
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
|
|
|
| 5 |
|
| 6 |
# --- Constants ---
|
| 7 |
-
DEFAULT_API_URL = "https://agents-course-unit4-scoring
|
| 8 |
-
|
| 9 |
|
| 10 |
# --- Basic Agent Definition ---
|
| 11 |
# π You can customize this class with your own logic or tools
|
|
@@ -21,6 +21,23 @@ class BasicAgent:
|
|
| 21 |
return fixed_answer
|
| 22 |
|
| 23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 25 |
"""
|
| 26 |
Fetch all questions, run the agent, submit answers, and show results.
|
|
@@ -32,17 +49,33 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 32 |
print(f"π€ User logged in: {username}")
|
| 33 |
else:
|
| 34 |
print("β User not logged in.")
|
| 35 |
-
return "Please login to Hugging Face first.", None
|
| 36 |
|
| 37 |
api_url = DEFAULT_API_URL
|
| 38 |
questions_url = f"{api_url}/questions"
|
| 39 |
submit_url = f"{api_url}/submit"
|
| 40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
# 1οΈβ£ Create Agent
|
| 42 |
try:
|
| 43 |
agent = BasicAgent()
|
| 44 |
except Exception as e:
|
| 45 |
-
return f"Agent initialization failed: {e}", None
|
| 46 |
|
| 47 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "Local_Run"
|
| 48 |
print(f"π Agent code link: {agent_code}")
|
|
@@ -51,39 +84,72 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 51 |
try:
|
| 52 |
print("π‘ Fetching questions...")
|
| 53 |
response = requests.get(questions_url, timeout=15)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
response.raise_for_status()
|
| 55 |
questions_data = response.json()
|
|
|
|
| 56 |
if not questions_data:
|
| 57 |
-
return "Fetched question list is empty or invalid.", None
|
|
|
|
| 58 |
print(f"β
Retrieved {len(questions_data)} questions.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
except Exception as e:
|
| 60 |
-
return f"Error fetching questions: {e}", None
|
| 61 |
|
| 62 |
# 3οΈβ£ Run Agent
|
| 63 |
results_log = []
|
| 64 |
answers_payload = []
|
| 65 |
-
|
|
|
|
|
|
|
|
|
|
| 66 |
task_id = item.get("task_id")
|
| 67 |
question_text = item.get("question")
|
|
|
|
| 68 |
if not task_id or question_text is None:
|
|
|
|
| 69 |
continue
|
|
|
|
| 70 |
try:
|
|
|
|
| 71 |
submitted_answer = agent(question_text)
|
| 72 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 73 |
results_log.append({
|
| 74 |
"Task ID": task_id,
|
| 75 |
-
"Question": question_text,
|
| 76 |
-
"Submitted Answer": submitted_answer
|
| 77 |
})
|
| 78 |
except Exception as e:
|
|
|
|
|
|
|
| 79 |
results_log.append({
|
| 80 |
"Task ID": task_id,
|
| 81 |
-
"Question": question_text,
|
| 82 |
-
"Submitted Answer":
|
| 83 |
})
|
| 84 |
|
| 85 |
if not answers_payload:
|
| 86 |
-
return "No answers generated by the agent.", pd.DataFrame(results_log)
|
| 87 |
|
| 88 |
# 4οΈβ£ Submit Answers
|
| 89 |
submission_data = {
|
|
@@ -95,8 +161,19 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 95 |
try:
|
| 96 |
print("π€ Submitting answers...")
|
| 97 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 98 |
response.raise_for_status()
|
| 99 |
result_data = response.json()
|
|
|
|
| 100 |
final_status = (
|
| 101 |
f"β
Submission Successful!\n"
|
| 102 |
f"π€ User: {result_data.get('username')}\n"
|
|
@@ -106,12 +183,17 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 106 |
)
|
| 107 |
results_df = pd.DataFrame(results_log)
|
| 108 |
return final_status, results_df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 109 |
except Exception as e:
|
| 110 |
-
return f"Submission failed: {e}", pd.DataFrame(results_log)
|
| 111 |
|
| 112 |
|
| 113 |
# --- Gradio Interface ---
|
| 114 |
-
with gr.Blocks() as demo:
|
| 115 |
gr.Markdown("# π€ Basic Agent Evaluation Runner")
|
| 116 |
gr.Markdown(
|
| 117 |
"""
|
|
@@ -122,13 +204,16 @@ with gr.Blocks() as demo:
|
|
| 122 |
---
|
| 123 |
The process may take time (the agent answers all questions).
|
| 124 |
You can customize the agent with reasoning, search tools, or memory.
|
|
|
|
|
|
|
|
|
|
| 125 |
"""
|
| 126 |
)
|
| 127 |
|
| 128 |
gr.LoginButton()
|
| 129 |
-
run_button = gr.Button("π Run Evaluation & Submit All Answers")
|
| 130 |
|
| 131 |
-
status_output = gr.Textbox(label="Run Status / Submission Result", lines=
|
| 132 |
results_table = gr.DataFrame(label="Questions and Agent Answers")
|
| 133 |
|
| 134 |
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
|
@@ -136,4 +221,4 @@ with gr.Blocks() as demo:
|
|
| 136 |
|
| 137 |
if __name__ == "__main__":
|
| 138 |
print("π Launching Gradio Interface...")
|
| 139 |
-
demo.launch(debug=True, share=False)
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
| 5 |
+
import time
|
| 6 |
|
| 7 |
# --- Constants ---
|
| 8 |
+
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
|
|
|
| 9 |
|
| 10 |
# --- Basic Agent Definition ---
|
| 11 |
# π You can customize this class with your own logic or tools
|
|
|
|
| 21 |
return fixed_answer
|
| 22 |
|
| 23 |
|
| 24 |
+
def check_api_status(api_url: str) -> tuple[bool, str]:
|
| 25 |
+
"""Check if the API is accessible"""
|
| 26 |
+
try:
|
| 27 |
+
# Try to access the base URL or a health check endpoint
|
| 28 |
+
response = requests.get(api_url, timeout=10)
|
| 29 |
+
if response.status_code == 200:
|
| 30 |
+
return True, "API is accessible"
|
| 31 |
+
else:
|
| 32 |
+
return False, f"API returned status code: {response.status_code}"
|
| 33 |
+
except requests.exceptions.Timeout:
|
| 34 |
+
return False, "API request timed out"
|
| 35 |
+
except requests.exceptions.ConnectionError:
|
| 36 |
+
return False, "Cannot connect to API (connection error)"
|
| 37 |
+
except Exception as e:
|
| 38 |
+
return False, f"API check failed: {str(e)}"
|
| 39 |
+
|
| 40 |
+
|
| 41 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 42 |
"""
|
| 43 |
Fetch all questions, run the agent, submit answers, and show results.
|
|
|
|
| 49 |
print(f"π€ User logged in: {username}")
|
| 50 |
else:
|
| 51 |
print("β User not logged in.")
|
| 52 |
+
return "β Please login to Hugging Face first.", None
|
| 53 |
|
| 54 |
api_url = DEFAULT_API_URL
|
| 55 |
questions_url = f"{api_url}/questions"
|
| 56 |
submit_url = f"{api_url}/submit"
|
| 57 |
|
| 58 |
+
# Check API status first
|
| 59 |
+
print("π Checking API status...")
|
| 60 |
+
api_ok, api_msg = check_api_status(api_url)
|
| 61 |
+
if not api_ok:
|
| 62 |
+
error_msg = (
|
| 63 |
+
f"β οΈ API Status Check Failed\n\n"
|
| 64 |
+
f"Issue: {api_msg}\n\n"
|
| 65 |
+
f"Possible solutions:\n"
|
| 66 |
+
f"1. The scoring API at {api_url} may be down or moved\n"
|
| 67 |
+
f"2. Check if the API URL is correct\n"
|
| 68 |
+
f"3. The Hugging Face Space might need to be restarted\n"
|
| 69 |
+
f"4. Try again in a few minutes\n\n"
|
| 70 |
+
f"You can verify the API status by visiting: {api_url}"
|
| 71 |
+
)
|
| 72 |
+
return error_msg, None
|
| 73 |
+
|
| 74 |
# 1οΈβ£ Create Agent
|
| 75 |
try:
|
| 76 |
agent = BasicAgent()
|
| 77 |
except Exception as e:
|
| 78 |
+
return f"β Agent initialization failed: {e}", None
|
| 79 |
|
| 80 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "Local_Run"
|
| 81 |
print(f"π Agent code link: {agent_code}")
|
|
|
|
| 84 |
try:
|
| 85 |
print("π‘ Fetching questions...")
|
| 86 |
response = requests.get(questions_url, timeout=15)
|
| 87 |
+
|
| 88 |
+
if response.status_code == 404:
|
| 89 |
+
error_msg = (
|
| 90 |
+
f"β οΈ Questions endpoint not found (404)\n\n"
|
| 91 |
+
f"The endpoint {questions_url} is not available.\n\n"
|
| 92 |
+
f"Possible issues:\n"
|
| 93 |
+
f"1. The API structure may have changed\n"
|
| 94 |
+
f"2. The scoring service might be under maintenance\n"
|
| 95 |
+
f"3. You may need to update the API URL\n\n"
|
| 96 |
+
f"Please check the course materials or contact the instructor for the correct API endpoint."
|
| 97 |
+
)
|
| 98 |
+
return error_msg, None
|
| 99 |
+
|
| 100 |
response.raise_for_status()
|
| 101 |
questions_data = response.json()
|
| 102 |
+
|
| 103 |
if not questions_data:
|
| 104 |
+
return "β οΈ Fetched question list is empty or invalid.", None
|
| 105 |
+
|
| 106 |
print(f"β
Retrieved {len(questions_data)} questions.")
|
| 107 |
+
|
| 108 |
+
except requests.exceptions.Timeout:
|
| 109 |
+
return f"β±οΈ Request timed out while fetching questions from {questions_url}", None
|
| 110 |
+
except requests.exceptions.ConnectionError:
|
| 111 |
+
return f"π Connection error: Cannot reach {questions_url}", None
|
| 112 |
+
except requests.exceptions.HTTPError as e:
|
| 113 |
+
return f"β HTTP Error fetching questions: {e}", None
|
| 114 |
+
except ValueError as e:
|
| 115 |
+
return f"β Invalid JSON response from questions endpoint: {e}", None
|
| 116 |
except Exception as e:
|
| 117 |
+
return f"β Error fetching questions: {e}", None
|
| 118 |
|
| 119 |
# 3οΈβ£ Run Agent
|
| 120 |
results_log = []
|
| 121 |
answers_payload = []
|
| 122 |
+
|
| 123 |
+
print(f"π€ Running agent on {len(questions_data)} questions...")
|
| 124 |
+
|
| 125 |
+
for i, item in enumerate(questions_data):
|
| 126 |
task_id = item.get("task_id")
|
| 127 |
question_text = item.get("question")
|
| 128 |
+
|
| 129 |
if not task_id or question_text is None:
|
| 130 |
+
print(f"β οΈ Skipping invalid question item: {item}")
|
| 131 |
continue
|
| 132 |
+
|
| 133 |
try:
|
| 134 |
+
print(f"Processing question {i+1}/{len(questions_data)}: {task_id}")
|
| 135 |
submitted_answer = agent(question_text)
|
| 136 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 137 |
results_log.append({
|
| 138 |
"Task ID": task_id,
|
| 139 |
+
"Question": question_text[:100] + "..." if len(question_text) > 100 else question_text,
|
| 140 |
+
"Submitted Answer": submitted_answer[:100] + "..." if len(str(submitted_answer)) > 100 else submitted_answer
|
| 141 |
})
|
| 142 |
except Exception as e:
|
| 143 |
+
error_msg = f"AGENT ERROR: {e}"
|
| 144 |
+
print(f"β {error_msg} for task {task_id}")
|
| 145 |
results_log.append({
|
| 146 |
"Task ID": task_id,
|
| 147 |
+
"Question": question_text[:100] + "..." if len(question_text) > 100 else question_text,
|
| 148 |
+
"Submitted Answer": error_msg
|
| 149 |
})
|
| 150 |
|
| 151 |
if not answers_payload:
|
| 152 |
+
return "β οΈ No answers generated by the agent.", pd.DataFrame(results_log)
|
| 153 |
|
| 154 |
# 4οΈβ£ Submit Answers
|
| 155 |
submission_data = {
|
|
|
|
| 161 |
try:
|
| 162 |
print("π€ Submitting answers...")
|
| 163 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
| 164 |
+
|
| 165 |
+
if response.status_code == 404:
|
| 166 |
+
error_msg = (
|
| 167 |
+
f"β οΈ Submit endpoint not found (404)\n\n"
|
| 168 |
+
f"Generated {len(answers_payload)} answers but cannot submit them.\n"
|
| 169 |
+
f"The endpoint {submit_url} is not available.\n\n"
|
| 170 |
+
f"Your answers are shown in the table below for reference."
|
| 171 |
+
)
|
| 172 |
+
return error_msg, pd.DataFrame(results_log)
|
| 173 |
+
|
| 174 |
response.raise_for_status()
|
| 175 |
result_data = response.json()
|
| 176 |
+
|
| 177 |
final_status = (
|
| 178 |
f"β
Submission Successful!\n"
|
| 179 |
f"π€ User: {result_data.get('username')}\n"
|
|
|
|
| 183 |
)
|
| 184 |
results_df = pd.DataFrame(results_log)
|
| 185 |
return final_status, results_df
|
| 186 |
+
|
| 187 |
+
except requests.exceptions.Timeout:
|
| 188 |
+
return f"β±οΈ Submission timed out. Your answers:\n{len(answers_payload)} answers generated but not submitted.", pd.DataFrame(results_log)
|
| 189 |
+
except requests.exceptions.HTTPError as e:
|
| 190 |
+
return f"β HTTP Error during submission: {e}\nYour answers are shown below.", pd.DataFrame(results_log)
|
| 191 |
except Exception as e:
|
| 192 |
+
return f"β Submission failed: {e}\nYour answers are shown below.", pd.DataFrame(results_log)
|
| 193 |
|
| 194 |
|
| 195 |
# --- Gradio Interface ---
|
| 196 |
+
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
| 197 |
gr.Markdown("# π€ Basic Agent Evaluation Runner")
|
| 198 |
gr.Markdown(
|
| 199 |
"""
|
|
|
|
| 204 |
---
|
| 205 |
The process may take time (the agent answers all questions).
|
| 206 |
You can customize the agent with reasoning, search tools, or memory.
|
| 207 |
+
|
| 208 |
+
β οΈ **Note**: If you see a 404 error, the scoring API may be temporarily unavailable.
|
| 209 |
+
Check with your instructor for the correct API endpoint.
|
| 210 |
"""
|
| 211 |
)
|
| 212 |
|
| 213 |
gr.LoginButton()
|
| 214 |
+
run_button = gr.Button("π Run Evaluation & Submit All Answers", variant="primary")
|
| 215 |
|
| 216 |
+
status_output = gr.Textbox(label="Run Status / Submission Result", lines=8, interactive=False)
|
| 217 |
results_table = gr.DataFrame(label="Questions and Agent Answers")
|
| 218 |
|
| 219 |
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
|
|
|
| 221 |
|
| 222 |
if __name__ == "__main__":
|
| 223 |
print("π Launching Gradio Interface...")
|
| 224 |
+
demo.launch(debug=True, share=False)
|