Spaces:
Runtime error
Runtime error
Upload app.py
Browse files
app.py
CHANGED
|
@@ -1,206 +1,217 @@
|
|
| 1 |
import os
|
| 2 |
-
import inspect
|
| 3 |
import gradio as gr
|
| 4 |
import requests
|
| 5 |
import pandas as pd
|
| 6 |
from langchain_core.messages import HumanMessage
|
| 7 |
from agent import build_graph
|
| 8 |
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
# (Keep Constants as is)
|
| 12 |
# --- Constants ---
|
| 13 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 14 |
|
| 15 |
-
# --- Basic Agent Definition ---
|
| 16 |
-
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
|
| 17 |
-
|
| 18 |
-
|
| 19 |
class BasicAgent:
|
| 20 |
-
"""A langgraph agent."""
|
| 21 |
def __init__(self):
|
| 22 |
-
print("
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
def __call__(self, question: str) -> str:
|
| 26 |
-
print(f"
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
return "Please Login to Hugging Face with the button.", None
|
| 48 |
-
|
|
|
|
|
|
|
|
|
|
| 49 |
api_url = DEFAULT_API_URL
|
| 50 |
questions_url = f"{api_url}/questions"
|
| 51 |
submit_url = f"{api_url}/submit"
|
| 52 |
-
|
| 53 |
-
|
| 54 |
try:
|
| 55 |
agent = BasicAgent()
|
|
|
|
| 56 |
except Exception as e:
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
|
|
|
| 60 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
| 61 |
-
print(agent_code)
|
| 62 |
|
| 63 |
-
|
| 64 |
-
print(f"Fetching questions from: {questions_url}")
|
| 65 |
try:
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
print("Fetched questions list is empty.")
|
| 71 |
-
return "Fetched questions list is empty or invalid format.", None
|
| 72 |
-
print(f"Fetched {len(questions_data)} questions.")
|
| 73 |
-
except requests.exceptions.RequestException as e:
|
| 74 |
-
print(f"Error fetching questions: {e}")
|
| 75 |
-
return f"Error fetching questions: {e}", None
|
| 76 |
-
except requests.exceptions.JSONDecodeError as e:
|
| 77 |
-
print(f"Error decoding JSON response from questions endpoint: {e}")
|
| 78 |
-
print(f"Response text: {response.text[:500]}")
|
| 79 |
-
return f"Error decoding server response for questions: {e}", None
|
| 80 |
except Exception as e:
|
| 81 |
-
|
| 82 |
-
|
|
|
|
| 83 |
|
| 84 |
-
# 3. Run your Agent
|
| 85 |
results_log = []
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
|
|
|
| 89 |
task_id = item.get("task_id")
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
continue
|
|
|
|
| 94 |
try:
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 98 |
except Exception as e:
|
| 99 |
-
|
| 100 |
-
|
|
|
|
| 101 |
|
| 102 |
-
if not
|
| 103 |
-
print("Agent did not produce any answers to submit.")
|
| 104 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 105 |
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
print(status_update)
|
| 110 |
-
|
| 111 |
-
# 5. Submit
|
| 112 |
-
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
| 113 |
try:
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
|
|
|
| 118 |
f"Submission Successful!\n"
|
| 119 |
-
f"User: {
|
| 120 |
-
f"
|
| 121 |
-
f"({
|
| 122 |
-
f"
|
| 123 |
)
|
| 124 |
-
print("Submission successful
|
| 125 |
-
|
| 126 |
-
return
|
| 127 |
-
|
| 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 requests.exceptions.JSONDecodeError:
|
| 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(
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
-
# --- Build Gradio Interface using Blocks ---
|
| 156 |
with gr.Blocks() as demo:
|
| 157 |
-
gr.Markdown("#
|
| 158 |
-
gr.Markdown(
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
)
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
|
|
|
|
|
|
| 177 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
)
|
| 183 |
|
| 184 |
if __name__ == "__main__":
|
| 185 |
-
print("
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
| 192 |
-
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
| 193 |
else:
|
| 194 |
-
print("
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
print(f"✅ SPACE_ID found: {space_id_startup}")
|
| 198 |
-
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
| 199 |
-
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
| 200 |
-
else:
|
| 201 |
-
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
| 202 |
-
|
| 203 |
-
print("-"*(60 + len(" App Starting ")) + "\n")
|
| 204 |
-
|
| 205 |
-
print("Launching Gradio Interface for Basic Agent Evaluation...")
|
| 206 |
-
demo.launch(debug=True, share=False)
|
|
|
|
| 1 |
import os
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
| 5 |
from langchain_core.messages import HumanMessage
|
| 6 |
from agent import build_graph
|
| 7 |
|
|
|
|
|
|
|
|
|
|
| 8 |
# --- Constants ---
|
| 9 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
class BasicAgent:
|
| 12 |
+
"""A langgraph agent using OpenAI."""
|
| 13 |
def __init__(self):
|
| 14 |
+
print("=== INITIALIZING OPENAI BASIC AGENT ===")
|
| 15 |
+
print(f"Current working directory: {os.getcwd()}")
|
| 16 |
+
print(f"Files in directory: {os.listdir('.')}")
|
| 17 |
+
|
| 18 |
+
# Check environment variables
|
| 19 |
+
print("=== ENVIRONMENT VARIABLES ===")
|
| 20 |
+
for key in sorted(os.environ.keys()):
|
| 21 |
+
if any(term in key.upper() for term in ['OPENAI', 'API_KEY', 'TOKEN', 'TAVILY']):
|
| 22 |
+
value = os.environ[key]
|
| 23 |
+
print(f"{key}: {value[:10] if value else 'None'}...")
|
| 24 |
+
|
| 25 |
+
# Check specifically for OpenAI API key
|
| 26 |
+
openai_key = os.getenv("OPENAI_API_KEY")
|
| 27 |
+
if openai_key:
|
| 28 |
+
print(f"✓ OpenAI API Key found: {openai_key[:15]}...")
|
| 29 |
+
else:
|
| 30 |
+
print("✗ OpenAI API Key not found!")
|
| 31 |
+
print("Please add OPENAI_API_KEY to your Hugging Face Space secrets")
|
| 32 |
+
|
| 33 |
+
try:
|
| 34 |
+
self.graph = build_graph()
|
| 35 |
+
print("✓ Graph built successfully")
|
| 36 |
+
except Exception as e:
|
| 37 |
+
print(f"✗ Error building graph: {e}")
|
| 38 |
+
raise e
|
| 39 |
|
| 40 |
def __call__(self, question: str) -> str:
|
| 41 |
+
print(f"=== AGENT CALL ===")
|
| 42 |
+
print(f"Question: {question[:100]}...")
|
| 43 |
+
|
| 44 |
+
try:
|
| 45 |
+
messages = [HumanMessage(content=question)]
|
| 46 |
+
print(f"Invoking graph with messages: {len(messages)}")
|
| 47 |
+
|
| 48 |
+
result = self.graph.invoke({"messages": messages})
|
| 49 |
+
print(f"Graph result keys: {result.keys() if isinstance(result, dict) else 'Not a dict'}")
|
| 50 |
+
|
| 51 |
+
if 'messages' in result and result['messages']:
|
| 52 |
+
answer = result['messages'][-1].content
|
| 53 |
+
print(f"Answer (first 100 chars): {answer[:100]}...")
|
| 54 |
+
return answer
|
| 55 |
+
else:
|
| 56 |
+
print("No messages in result")
|
| 57 |
+
return "I apologize, but I couldn't generate a response."
|
| 58 |
+
|
| 59 |
+
except Exception as e:
|
| 60 |
+
print(f"Error in agent call: {e}")
|
| 61 |
+
return f"Error: {str(e)}"
|
| 62 |
+
|
| 63 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 64 |
+
print("=== STARTING RUN AND SUBMIT ===")
|
| 65 |
+
space_id = os.getenv("SPACE_ID")
|
| 66 |
+
print(f"Space ID: {space_id}")
|
| 67 |
+
|
| 68 |
+
if not profile:
|
| 69 |
return "Please Login to Hugging Face with the button.", None
|
| 70 |
+
|
| 71 |
+
username = profile.username
|
| 72 |
+
print(f"Username: {username}")
|
| 73 |
+
|
| 74 |
api_url = DEFAULT_API_URL
|
| 75 |
questions_url = f"{api_url}/questions"
|
| 76 |
submit_url = f"{api_url}/submit"
|
| 77 |
+
|
| 78 |
+
print("=== INITIALIZING AGENT ===")
|
| 79 |
try:
|
| 80 |
agent = BasicAgent()
|
| 81 |
+
print("✓ Agent initialized successfully")
|
| 82 |
except Exception as e:
|
| 83 |
+
error_msg = f"Error initializing agent: {e}"
|
| 84 |
+
print(error_msg)
|
| 85 |
+
return error_msg, None
|
| 86 |
+
|
| 87 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
| 88 |
+
print(f"Agent code URL: {agent_code}")
|
| 89 |
|
| 90 |
+
print("=== FETCHING QUESTIONS ===")
|
|
|
|
| 91 |
try:
|
| 92 |
+
resp_q = requests.get(questions_url, timeout=15)
|
| 93 |
+
resp_q.raise_for_status()
|
| 94 |
+
questions = resp_q.json()
|
| 95 |
+
print(f"✓ Fetched {len(questions)} questions")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
except Exception as e:
|
| 97 |
+
error_msg = f"Error fetching questions: {e}"
|
| 98 |
+
print(error_msg)
|
| 99 |
+
return error_msg, None
|
| 100 |
|
|
|
|
| 101 |
results_log = []
|
| 102 |
+
answers = []
|
| 103 |
+
|
| 104 |
+
print("=== PROCESSING QUESTIONS ===")
|
| 105 |
+
for i, item in enumerate(questions):
|
| 106 |
task_id = item.get("task_id")
|
| 107 |
+
q = item.get("question")
|
| 108 |
+
|
| 109 |
+
print(f"\n--- Question {i+1}/{len(questions)} ---")
|
| 110 |
+
print(f"Task ID: {task_id}")
|
| 111 |
+
print(f"Question: {q[:100]}...")
|
| 112 |
+
|
| 113 |
+
if not task_id or q is None:
|
| 114 |
+
print("Skipping - missing task_id or question")
|
| 115 |
continue
|
| 116 |
+
|
| 117 |
try:
|
| 118 |
+
print("Calling agent...")
|
| 119 |
+
ans = agent(q)
|
| 120 |
+
print(f"Answer: {ans[:100]}...")
|
| 121 |
+
|
| 122 |
+
answers.append({"task_id": task_id, "submitted_answer": ans})
|
| 123 |
+
results_log.append({"Task ID": task_id, "Question": q, "Submitted Answer": ans})
|
| 124 |
+
print("✓ Question processed successfully")
|
| 125 |
+
|
| 126 |
except Exception as e:
|
| 127 |
+
error_msg = f"ERROR: {e}"
|
| 128 |
+
print(f"✗ Error processing question: {error_msg}")
|
| 129 |
+
results_log.append({"Task ID": task_id, "Question": q, "Submitted Answer": error_msg})
|
| 130 |
|
| 131 |
+
if not answers:
|
|
|
|
| 132 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 133 |
|
| 134 |
+
print(f"=== SUBMITTING {len(answers)} ANSWERS ===")
|
| 135 |
+
payload = {"username": username.strip(), "agent_code": agent_code, "answers": answers}
|
| 136 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
try:
|
| 138 |
+
resp_s = requests.post(submit_url, json=payload, timeout=60)
|
| 139 |
+
resp_s.raise_for_status()
|
| 140 |
+
data = resp_s.json()
|
| 141 |
+
|
| 142 |
+
status = (
|
| 143 |
f"Submission Successful!\n"
|
| 144 |
+
f"User: {data.get('username')}\n"
|
| 145 |
+
f"Score: {data.get('score', 'N/A')}% "
|
| 146 |
+
f"({data.get('correct_count', '?')}/{data.get('total_attempted', '?')})\n"
|
| 147 |
+
f"{data.get('message', '')}"
|
| 148 |
)
|
| 149 |
+
print("✓ Submission successful")
|
| 150 |
+
print(status)
|
| 151 |
+
return status, pd.DataFrame(results_log)
|
| 152 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
except Exception as e:
|
| 154 |
+
error_msg = f"Submission Failed: {e}"
|
| 155 |
+
print(error_msg)
|
| 156 |
+
return error_msg, pd.DataFrame(results_log)
|
| 157 |
+
|
| 158 |
+
# Simple test function for debugging
|
| 159 |
+
def test_agent():
|
| 160 |
+
"""Test function to verify agent works"""
|
| 161 |
+
print("=== TESTING AGENT ===")
|
| 162 |
+
try:
|
| 163 |
+
agent = BasicAgent()
|
| 164 |
+
test_questions = [
|
| 165 |
+
"What is 2 + 3?",
|
| 166 |
+
"What is 10 * 5?",
|
| 167 |
+
"Search for information about Python programming"
|
| 168 |
+
]
|
| 169 |
+
|
| 170 |
+
for q in test_questions:
|
| 171 |
+
print(f"\nTesting: {q}")
|
| 172 |
+
answer = agent(q)
|
| 173 |
+
print(f"Answer: {answer}")
|
| 174 |
+
|
| 175 |
+
except Exception as e:
|
| 176 |
+
print(f"Test failed: {e}")
|
| 177 |
|
|
|
|
| 178 |
with gr.Blocks() as demo:
|
| 179 |
+
gr.Markdown("# OpenAI-Powered Agent Evaluation Runner")
|
| 180 |
+
gr.Markdown("""
|
| 181 |
+
This agent uses OpenAI's GPT models instead of Hugging Face.
|
| 182 |
+
|
| 183 |
+
## Setup Instructions:
|
| 184 |
+
1. Get an OpenAI API key from https://platform.openai.com/api-keys
|
| 185 |
+
2. Add it as `OPENAI_API_KEY` in your Hugging Face Space secrets
|
| 186 |
+
3. (Optional) Add `TAVILY_API_KEY` for web search functionality
|
| 187 |
+
4. Log in with the button below
|
| 188 |
+
5. Click **Run Evaluation & Submit All Answers**
|
| 189 |
+
|
| 190 |
+
## Current Configuration:
|
| 191 |
+
- Model: GPT-3.5-turbo (change to GPT-4 in agent.py if you have access)
|
| 192 |
+
- Tools: Math operations, Wikipedia search, Arxiv search, Web search (if Tavily configured)
|
| 193 |
+
""")
|
| 194 |
+
|
| 195 |
+
with gr.Row():
|
| 196 |
+
gr.LoginButton()
|
| 197 |
+
test_btn = gr.Button("Test Agent", variant="secondary")
|
| 198 |
+
|
| 199 |
+
run_btn = gr.Button("Run Evaluation & Submit All Answers", variant="primary")
|
| 200 |
+
status_out = gr.Textbox(label="Run Status / Submission Result", lines=5)
|
| 201 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
| 202 |
+
|
| 203 |
+
# Button actions
|
| 204 |
+
run_btn.click(fn=run_and_submit_all, outputs=[status_out, results_table])
|
| 205 |
+
test_btn.click(fn=test_agent, outputs=[])
|
|
|
|
| 206 |
|
| 207 |
if __name__ == "__main__":
|
| 208 |
+
print("=== STARTING OPENAI GRADIO APP ===")
|
| 209 |
+
|
| 210 |
+
# Quick environment check
|
| 211 |
+
openai_key = os.getenv("OPENAI_API_KEY")
|
| 212 |
+
if openai_key:
|
| 213 |
+
print(f"✓ OpenAI API Key configured: {openai_key[:15]}...")
|
|
|
|
|
|
|
| 214 |
else:
|
| 215 |
+
print("⚠️ OpenAI API Key not found - please add OPENAI_API_KEY to your Space secrets")
|
| 216 |
+
|
| 217 |
+
demo.launch(debug=True, share=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|