Spaces:
Sleeping
Sleeping
Fix UI and evaluation functionality with simplified app.py
Browse files
app.py
CHANGED
|
@@ -1,30 +1,17 @@
|
|
| 1 |
-
"""
|
| 2 |
-
GAIA Agent Hugging Face Space Interface
|
| 3 |
-
|
| 4 |
-
A clean, straightforward interface for interacting with the GAIA Agent.
|
| 5 |
-
This implementation focuses on providing a smooth user experience with
|
| 6 |
-
proper error handling and clear messaging.
|
| 7 |
-
|
| 8 |
-
Environment variables:
|
| 9 |
-
- OPENAI_API_KEY: Required for the LLM
|
| 10 |
-
- SERPER_API_KEY: For web search capabilities
|
| 11 |
-
- SUPABASE_URL: Optional for memory storage
|
| 12 |
-
- SUPABASE_KEY: Optional for memory storage
|
| 13 |
-
"""
|
| 14 |
-
|
| 15 |
import os
|
| 16 |
import gradio as gr
|
|
|
|
|
|
|
| 17 |
import logging
|
| 18 |
import time
|
| 19 |
import traceback
|
| 20 |
-
import requests
|
| 21 |
-
import pandas as pd
|
| 22 |
from typing import Dict, Any, Optional, Tuple, List
|
| 23 |
|
| 24 |
-
# Constants
|
|
|
|
| 25 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 26 |
|
| 27 |
-
# Import our GAIA Agent
|
| 28 |
from gaia_agent import GAIAAgent
|
| 29 |
|
| 30 |
# Configure logging
|
|
@@ -32,423 +19,104 @@ logging.basicConfig(
|
|
| 32 |
level=logging.INFO,
|
| 33 |
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
| 34 |
)
|
| 35 |
-
logger = logging.getLogger("
|
| 36 |
-
|
| 37 |
-
# Load environment variables with error handling
|
| 38 |
-
def get_env_var(name, default="", required=False):
|
| 39 |
-
"""Helper function to get environment variables with logging"""
|
| 40 |
-
value = os.getenv(name, default)
|
| 41 |
-
if not value and required:
|
| 42 |
-
logger.error(f"Required environment variable {name} is not set")
|
| 43 |
-
elif not value:
|
| 44 |
-
logger.warning(f"Optional environment variable {name} is not set")
|
| 45 |
-
return value
|
| 46 |
-
|
| 47 |
-
# Check for required environment variables with default values
|
| 48 |
-
OPENAI_API_KEY = get_env_var("OPENAI_API_KEY", "", True) # Critical for agent operation
|
| 49 |
-
SERPER_API_KEY = get_env_var("SERPER_API_KEY", "") # Optional but recommended
|
| 50 |
-
PERPLEXITY_API_KEY = get_env_var("PERPLEXITY_API_KEY", "") # Optional alternative to SERPER
|
| 51 |
-
SUPABASE_URL = get_env_var("SUPABASE_URL", "") # Optional for memory persistence
|
| 52 |
-
SUPABASE_KEY = get_env_var("SUPABASE_KEY", "") # Optional for memory persistence
|
| 53 |
-
SPACE_HOST = get_env_var("SPACE_HOST", "") # For Hugging Face Space detection
|
| 54 |
-
SPACE_ID = get_env_var("SPACE_ID", "") # For Hugging Face Space identification
|
| 55 |
-
|
| 56 |
-
# Validate critical dependencies
|
| 57 |
-
try:
|
| 58 |
-
import gradio as gr
|
| 59 |
-
except ImportError as e:
|
| 60 |
-
logger.critical(f"Critical dependency 'gradio' not found: {e}")
|
| 61 |
-
raise ImportError("Gradio is required for the app interface. Please install it with 'pip install gradio'.")
|
| 62 |
-
|
| 63 |
-
# Log warnings and detailed diagnostics for missing environment variables
|
| 64 |
-
if not OPENAI_API_KEY:
|
| 65 |
-
logger.warning("OPENAI_API_KEY is not set. Agent will use fallback mode with limited capabilities.")
|
| 66 |
-
logger.info("To get an OpenAI API key, visit: https://platform.openai.com/api-keys")
|
| 67 |
-
|
| 68 |
-
# Check search capabilities
|
| 69 |
-
search_capability = "full"
|
| 70 |
-
if not SERPER_API_KEY and not PERPLEXITY_API_KEY:
|
| 71 |
-
search_capability = "limited"
|
| 72 |
-
logger.warning("Neither SERPER_API_KEY nor PERPLEXITY_API_KEY is set. Web search capabilities will be limited to DuckDuckGo.")
|
| 73 |
-
logger.info("To get a Serper API key, visit: https://serper.dev/")
|
| 74 |
-
elif not SERPER_API_KEY:
|
| 75 |
-
search_capability = "partial"
|
| 76 |
-
logger.warning("SERPER_API_KEY is not set. Will attempt to use Perplexity for search if available.")
|
| 77 |
-
elif not PERPLEXITY_API_KEY:
|
| 78 |
-
search_capability = "partial"
|
| 79 |
-
logger.warning("PERPLEXITY_API_KEY is not set. Will use Serper for search capabilities.")
|
| 80 |
|
| 81 |
-
#
|
| 82 |
-
|
| 83 |
-
logger.warning("SUPABASE_URL or SUPABASE_KEY is not set. Memory persistence will be limited to in-memory storage.")
|
| 84 |
-
if SUPABASE_URL and not SUPABASE_KEY:
|
| 85 |
-
logger.error("SUPABASE_URL is set but SUPABASE_KEY is missing. Memory persistence will not work correctly.")
|
| 86 |
-
elif SUPABASE_KEY and not SUPABASE_URL:
|
| 87 |
-
logger.error("SUPABASE_KEY is set but SUPABASE_URL is missing. Memory persistence will not work correctly.")
|
| 88 |
-
|
| 89 |
-
# Log overall capability assessment
|
| 90 |
-
logger.info(f"GAIA Agent capabilities: LLM: {'available' if OPENAI_API_KEY else 'unavailable'}, "
|
| 91 |
-
f"Search: {search_capability}, "
|
| 92 |
-
f"Memory: {'persistent' if (SUPABASE_URL and SUPABASE_KEY) else 'in-memory only'}")
|
| 93 |
-
|
| 94 |
-
# Custom CSS for improved appearance
|
| 95 |
-
custom_css = """
|
| 96 |
-
.message-bubble {
|
| 97 |
-
padding: 10px 15px;
|
| 98 |
-
border-radius: 15px;
|
| 99 |
-
margin-bottom: 10px;
|
| 100 |
-
}
|
| 101 |
-
.user-message {
|
| 102 |
-
background-color: #e6f7ff;
|
| 103 |
-
border: 1px solid #91d5ff;
|
| 104 |
-
}
|
| 105 |
-
.agent-message {
|
| 106 |
-
background-color: #f6ffed;
|
| 107 |
-
border: 1px solid #b7eb8f;
|
| 108 |
-
}
|
| 109 |
-
.banner {
|
| 110 |
-
padding: 10px;
|
| 111 |
-
background-color: #f9f9f9;
|
| 112 |
-
border-radius: 8px;
|
| 113 |
-
margin-bottom: 15px;
|
| 114 |
-
border: 1px solid #e0e0e0;
|
| 115 |
-
}
|
| 116 |
-
"""
|
| 117 |
-
|
| 118 |
-
def initialize_agent() -> Optional[GAIAAgent]:
|
| 119 |
-
"""
|
| 120 |
-
Initialize the GAIA Agent with proper error handling
|
| 121 |
-
|
| 122 |
-
Returns:
|
| 123 |
-
GAIAAgent instance or None if initialization fails
|
| 124 |
-
"""
|
| 125 |
try:
|
| 126 |
-
start_time = time.time()
|
| 127 |
agent = GAIAAgent()
|
| 128 |
-
|
| 129 |
-
logger.info(f"GAIA Agent initialized successfully in {init_time:.2f} seconds")
|
| 130 |
return agent
|
| 131 |
except Exception as e:
|
| 132 |
-
|
| 133 |
-
logger.error(error_msg)
|
| 134 |
logger.error(traceback.format_exc())
|
| 135 |
return None
|
| 136 |
|
| 137 |
-
def
|
| 138 |
-
"""
|
| 139 |
-
Check if all required environment variables are set
|
| 140 |
-
|
| 141 |
-
Returns:
|
| 142 |
-
Dictionary with status of each required environment variable
|
| 143 |
-
"""
|
| 144 |
-
return {
|
| 145 |
-
"OPENAI_API_KEY": bool(OPENAI_API_KEY),
|
| 146 |
-
"SERPER_API_KEY": bool(SERPER_API_KEY),
|
| 147 |
-
"PERPLEXITY_API_KEY": bool(PERPLEXITY_API_KEY),
|
| 148 |
-
"SUPABASE_URL": bool(SUPABASE_URL),
|
| 149 |
-
"SUPABASE_KEY": bool(SUPABASE_KEY),
|
| 150 |
-
}
|
| 151 |
-
|
| 152 |
-
def process_question(
|
| 153 |
-
message: str,
|
| 154 |
-
history: list,
|
| 155 |
-
agent_instance: Optional[GAIAAgent] = None,
|
| 156 |
-
progress=gr.Progress()
|
| 157 |
-
) -> tuple:
|
| 158 |
-
"""
|
| 159 |
-
Process a user question and get a response from the GAIA Agent
|
| 160 |
-
|
| 161 |
-
Args:
|
| 162 |
-
message: User's question
|
| 163 |
-
history: Conversation history
|
| 164 |
-
agent_instance: GAIA Agent instance
|
| 165 |
-
progress: Gradio progress indicator
|
| 166 |
-
|
| 167 |
-
Returns:
|
| 168 |
-
Tuple of (history, error_message)
|
| 169 |
-
"""
|
| 170 |
-
logger.info(f"Processing question request with message length: {len(message) if message else 0}")
|
| 171 |
-
|
| 172 |
-
# Input validation with detailed error messages
|
| 173 |
-
if not message:
|
| 174 |
-
logger.warning("Empty question received")
|
| 175 |
-
return history, "Please enter a question."
|
| 176 |
-
|
| 177 |
-
if not isinstance(message, str):
|
| 178 |
-
logger.error(f"Invalid message type: {type(message)}. Expected string.")
|
| 179 |
-
return history, "Invalid input type. Please enter a text question."
|
| 180 |
-
|
| 181 |
-
if not message.strip():
|
| 182 |
-
logger.warning("Blank question (only whitespace) received")
|
| 183 |
-
return history, "Please enter a non-empty question."
|
| 184 |
-
|
| 185 |
-
# Validate history parameter
|
| 186 |
-
if not isinstance(history, list):
|
| 187 |
-
logger.error(f"Invalid history type: {type(history)}. Expected list.")
|
| 188 |
-
history = [] # Reset to empty list for recovery
|
| 189 |
-
|
| 190 |
-
# Check if agent is available with enhanced diagnostics
|
| 191 |
-
if agent_instance is None:
|
| 192 |
-
logger.error("Agent instance is None when processing question")
|
| 193 |
-
# Generate a more specific error message based on environment variables
|
| 194 |
-
env_status = check_environment()
|
| 195 |
-
missing_vars = [key for key, value in env_status.items() if not value]
|
| 196 |
-
|
| 197 |
-
# Check if OPENAI_API_KEY is missing, as it's critical
|
| 198 |
-
critical_vars_missing = "OPENAI_API_KEY" in missing_vars
|
| 199 |
-
|
| 200 |
-
if missing_vars:
|
| 201 |
-
if critical_vars_missing:
|
| 202 |
-
error_detail = f"Critical environment variables missing: {', '.join(missing_vars)}"
|
| 203 |
-
logger.error(error_detail)
|
| 204 |
-
else:
|
| 205 |
-
error_detail = f"Some environment variables are missing: {', '.join(missing_vars)}"
|
| 206 |
-
logger.warning(error_detail)
|
| 207 |
-
else:
|
| 208 |
-
error_detail = "Unknown initialization error. Please check the application logs."
|
| 209 |
-
logger.error("Agent is None but no environment variables are missing")
|
| 210 |
-
|
| 211 |
-
user_message = "I'm unable to process your question because the AI agent isn't properly initialized."
|
| 212 |
-
if critical_vars_missing:
|
| 213 |
-
user_message += " This is likely due to missing API keys required for my operation."
|
| 214 |
-
|
| 215 |
-
return (
|
| 216 |
-
history + [(message, user_message)],
|
| 217 |
-
f"GAIA Agent initialization failed. {error_detail}"
|
| 218 |
-
)
|
| 219 |
-
|
| 220 |
-
# Initialize progress indicator with better error handling
|
| 221 |
-
try:
|
| 222 |
-
progress(0.1, desc="Processing question...")
|
| 223 |
-
except Exception as progress_err:
|
| 224 |
-
logger.warning(f"Non-critical error with progress indicator: {progress_err}")
|
| 225 |
-
# Continue execution - progress indicator issues shouldn't block processing
|
| 226 |
-
|
| 227 |
-
error_message = ""
|
| 228 |
-
|
| 229 |
-
try:
|
| 230 |
-
# Process the question with timing
|
| 231 |
-
start_time = time.time()
|
| 232 |
-
logger.info(f"Sending question to agent: '{message[:50]}{'...' if len(message) > 50 else ''}'")
|
| 233 |
-
|
| 234 |
-
try:
|
| 235 |
-
progress(0.4, desc="Thinking...")
|
| 236 |
-
except:
|
| 237 |
-
pass # Non-critical, continue even if progress update fails
|
| 238 |
-
|
| 239 |
-
response = agent_instance.process_question(message)
|
| 240 |
-
|
| 241 |
-
try:
|
| 242 |
-
progress(1.0, desc="Done!")
|
| 243 |
-
except:
|
| 244 |
-
pass # Non-critical
|
| 245 |
-
|
| 246 |
-
processing_time = time.time() - start_time
|
| 247 |
-
logger.info(f"Question processed successfully in {processing_time:.2f} seconds")
|
| 248 |
-
|
| 249 |
-
# Validate response
|
| 250 |
-
if not response or not isinstance(response, str):
|
| 251 |
-
logger.warning(f"Agent returned invalid response type: {type(response)}")
|
| 252 |
-
response = "I generated a response but it was in an invalid format. Please try asking again."
|
| 253 |
-
|
| 254 |
-
# Return the updated history and empty error message
|
| 255 |
-
return history + [(message, response)], ""
|
| 256 |
-
|
| 257 |
-
except Exception as e:
|
| 258 |
-
error_details = traceback.format_exc()
|
| 259 |
-
error_type = type(e).__name__
|
| 260 |
-
logger.error(f"Error processing question: {error_type}: {e}")
|
| 261 |
-
logger.error(error_details)
|
| 262 |
-
|
| 263 |
-
# Create user-friendly error message based on exception type
|
| 264 |
-
user_message = "I encountered an error while processing your question."
|
| 265 |
-
|
| 266 |
-
if "ConnectionError" in error_type or "Timeout" in error_type:
|
| 267 |
-
user_message += " It seems I'm having trouble connecting to my knowledge services. This might be a temporary network issue."
|
| 268 |
-
elif "JSONDecodeError" in error_type:
|
| 269 |
-
user_message += " I had trouble processing some data. This might be due to an unexpected format."
|
| 270 |
-
elif "AuthenticationError" in error_type or "api key" in str(e).lower():
|
| 271 |
-
user_message += " There seems to be an issue with my authentication to external services."
|
| 272 |
-
elif "MemoryError" in error_type or "OutOfMemoryError" in error_type:
|
| 273 |
-
user_message += " The question might be too complex for me to process with current resources."
|
| 274 |
-
else:
|
| 275 |
-
user_message += " Please try again or rephrase your question."
|
| 276 |
-
|
| 277 |
-
error_message = f"Error processing question: {error_type}: {str(e)}"
|
| 278 |
-
return history + [(message, user_message)], error_message
|
| 279 |
-
|
| 280 |
-
def clear_conversation(agent_instance: Optional[GAIAAgent] = None) -> tuple:
|
| 281 |
-
"""
|
| 282 |
-
Clear the conversation history and agent memory
|
| 283 |
-
|
| 284 |
-
Args:
|
| 285 |
-
agent_instance: GAIA Agent instance
|
| 286 |
-
|
| 287 |
-
Returns:
|
| 288 |
-
Empty history and status message
|
| 289 |
-
"""
|
| 290 |
-
logger.info("Clear conversation request received")
|
| 291 |
-
|
| 292 |
-
# Track completion status to provide feedback
|
| 293 |
-
agent_memory_cleared = False
|
| 294 |
-
|
| 295 |
-
if agent_instance:
|
| 296 |
-
try:
|
| 297 |
-
# Validate agent instance has clear_memory method
|
| 298 |
-
if not hasattr(agent_instance, 'clear_memory') or not callable(getattr(agent_instance, 'clear_memory')):
|
| 299 |
-
logger.error("Agent instance does not have clear_memory method or it's not callable")
|
| 300 |
-
return [], "Failed to clear conversation: Invalid agent configuration."
|
| 301 |
-
|
| 302 |
-
# Clear the agent memory with detailed logging
|
| 303 |
-
start_time = time.time()
|
| 304 |
-
agent_instance.clear_memory()
|
| 305 |
-
clear_time = time.time() - start_time
|
| 306 |
-
|
| 307 |
-
logger.info(f"Agent memory cleared successfully in {clear_time:.2f} seconds")
|
| 308 |
-
agent_memory_cleared = True
|
| 309 |
-
|
| 310 |
-
except Exception as e:
|
| 311 |
-
error_type = type(e).__name__
|
| 312 |
-
error_details = traceback.format_exc()
|
| 313 |
-
logger.error(f"Error clearing agent memory: {error_type}: {e}")
|
| 314 |
-
logger.error(error_details)
|
| 315 |
-
|
| 316 |
-
# Return more specific error message based on exception type
|
| 317 |
-
if "AttributeError" in error_type:
|
| 318 |
-
return [], "Failed to clear conversation: Agent memory interface error."
|
| 319 |
-
elif "ConnectionError" in error_type:
|
| 320 |
-
return [], "Failed to clear persistent memory: Connection issue."
|
| 321 |
-
else:
|
| 322 |
-
return [], f"Partially cleared conversation: {error_type}."
|
| 323 |
-
else:
|
| 324 |
-
logger.warning("Clear conversation called with None agent instance")
|
| 325 |
-
|
| 326 |
-
# Provide different message based on what was cleared
|
| 327 |
-
if agent_memory_cleared:
|
| 328 |
-
return [], "Conversation and agent memory cleared successfully."
|
| 329 |
-
else:
|
| 330 |
-
return [], "Conversation history cleared. Note: Agent memory may not have been cleared."
|
| 331 |
-
|
| 332 |
-
def run_and_submit_all(
|
| 333 |
-
profile: Optional[gr.OAuthProfile] = None,
|
| 334 |
-
submission_username: str = "",
|
| 335 |
-
agent_code: str = "",
|
| 336 |
-
agent_instance: Optional[GAIAAgent] = None,
|
| 337 |
-
progress=gr.Progress()
|
| 338 |
-
) -> Tuple[str, Optional[pd.DataFrame]]:
|
| 339 |
"""
|
| 340 |
Fetches all questions, runs the GAIA Agent on them, submits all answers,
|
| 341 |
and displays the results.
|
| 342 |
-
|
| 343 |
-
Args:
|
| 344 |
-
profile: User OAuth profile (if using OAuth)
|
| 345 |
-
submission_username: Username for submission (if not using OAuth)
|
| 346 |
-
agent_code: URL to the agent code repository
|
| 347 |
-
agent_instance: GAIA Agent instance
|
| 348 |
-
progress: Gradio progress tracker
|
| 349 |
-
|
| 350 |
-
Returns:
|
| 351 |
-
Tuple of (status_message, results_dataframe)
|
| 352 |
"""
|
| 353 |
-
# Configure logging
|
| 354 |
-
logger = logging.getLogger("gaia_evaluation")
|
| 355 |
-
|
| 356 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
| 357 |
space_id = os.getenv("SPACE_ID", "")
|
| 358 |
-
|
| 359 |
-
# Set username based on profile or input
|
| 360 |
if profile:
|
| 361 |
username = profile.username
|
| 362 |
logger.info(f"User logged in: {username}")
|
| 363 |
-
elif submission_username:
|
| 364 |
-
username = submission_username
|
| 365 |
-
logger.info(f"Using provided username: {username}")
|
| 366 |
else:
|
| 367 |
-
logger.warning("
|
| 368 |
-
return "Please
|
| 369 |
-
|
| 370 |
-
# Set API endpoints
|
| 371 |
api_url = DEFAULT_API_URL
|
| 372 |
questions_url = f"{api_url}/questions"
|
| 373 |
submit_url = f"{api_url}/submit"
|
| 374 |
-
|
| 375 |
-
#
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
-
|
|
|
|
| 383 |
logger.error(error_msg)
|
| 384 |
return error_msg, None
|
| 385 |
-
|
| 386 |
-
#
|
| 387 |
-
|
| 388 |
-
|
| 389 |
-
|
| 390 |
-
|
| 391 |
logger.info(f"Fetching questions from: {questions_url}")
|
| 392 |
try:
|
| 393 |
response = requests.get(questions_url, timeout=15)
|
| 394 |
response.raise_for_status()
|
| 395 |
questions_data = response.json()
|
| 396 |
if not questions_data:
|
| 397 |
-
|
| 398 |
-
|
| 399 |
-
|
| 400 |
-
|
| 401 |
-
question_count = len(questions_data)
|
| 402 |
-
logger.info(f"Fetched {question_count} questions.")
|
| 403 |
-
progress(0.2, desc=f"Fetched {question_count} questions")
|
| 404 |
except requests.exceptions.RequestException as e:
|
| 405 |
error_msg = f"Error fetching questions: {e}"
|
| 406 |
logger.error(error_msg)
|
| 407 |
return error_msg, None
|
| 408 |
except requests.exceptions.JSONDecodeError as e:
|
| 409 |
-
|
| 410 |
-
|
| 411 |
-
|
| 412 |
except Exception as e:
|
| 413 |
error_msg = f"An unexpected error occurred fetching questions: {e}"
|
| 414 |
logger.error(error_msg)
|
| 415 |
logger.error(traceback.format_exc())
|
| 416 |
return error_msg, None
|
| 417 |
-
|
| 418 |
-
#
|
| 419 |
results_log = []
|
| 420 |
answers_payload = []
|
| 421 |
-
|
| 422 |
-
|
| 423 |
-
logger.info(f"Running agent on {total_questions} questions...")
|
| 424 |
-
|
| 425 |
-
question_progress_increment = 0.6 / total_questions # 60% of progress bar for processing questions
|
| 426 |
|
| 427 |
for i, item in enumerate(questions_data):
|
| 428 |
-
# Extract question information
|
| 429 |
task_id = item.get("task_id", item.get("id", f"q{i+1}"))
|
| 430 |
question_text = item.get("question")
|
| 431 |
|
| 432 |
if not task_id or question_text is None:
|
| 433 |
logger.warning(f"Skipping item with missing task_id or question: {item}")
|
| 434 |
continue
|
| 435 |
-
|
| 436 |
-
progress_value = 0.2 + (i * question_progress_increment)
|
| 437 |
-
progress(progress_value, desc=f"Processing question {i+1}/{total_questions}...")
|
| 438 |
|
| 439 |
-
logger.info(f"Processing question {i+1}/{
|
| 440 |
-
|
| 441 |
|
| 442 |
try:
|
| 443 |
-
#
|
| 444 |
-
submitted_answer =
|
| 445 |
-
|
| 446 |
-
processing_time = time.time() - question_start_time
|
| 447 |
|
| 448 |
-
#
|
| 449 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 450 |
-
|
| 451 |
-
# Store
|
| 452 |
results_log.append({
|
| 453 |
"Task ID": task_id,
|
| 454 |
"Question": question_text,
|
|
@@ -457,7 +125,7 @@ def run_and_submit_all(
|
|
| 457 |
"Status": "Success"
|
| 458 |
})
|
| 459 |
|
| 460 |
-
logger.info(f"Question {i+1} processed in {processing_time:.2f}s
|
| 461 |
|
| 462 |
except Exception as e:
|
| 463 |
error_msg = f"Error running agent on task {task_id}: {e}"
|
|
@@ -468,72 +136,46 @@ def run_and_submit_all(
|
|
| 468 |
"Task ID": task_id,
|
| 469 |
"Question": question_text,
|
| 470 |
"Submitted Answer": f"AGENT ERROR: {e}",
|
| 471 |
-
"Processing Time": f"{time.time() -
|
| 472 |
"Status": "Error"
|
| 473 |
})
|
| 474 |
-
|
| 475 |
# Check if we have answers to submit
|
| 476 |
if not answers_payload:
|
| 477 |
-
|
| 478 |
-
|
| 479 |
-
|
| 480 |
-
|
| 481 |
-
# Prepare submission
|
| 482 |
-
progress(0.8, desc="Preparing submission...")
|
| 483 |
-
|
| 484 |
-
# Format submission payload
|
| 485 |
submission_data = {
|
| 486 |
-
"username": username.strip(),
|
| 487 |
-
"agent_code": agent_code,
|
| 488 |
"answers": answers_payload
|
| 489 |
}
|
| 490 |
-
|
| 491 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
| 492 |
logger.info(status_update)
|
| 493 |
-
|
| 494 |
-
#
|
| 495 |
-
|
| 496 |
-
logger.info(f"Submitting answers to: {submit_url}")
|
| 497 |
-
|
| 498 |
try:
|
| 499 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
| 500 |
response.raise_for_status()
|
| 501 |
result_data = response.json()
|
| 502 |
|
| 503 |
-
# Extract results
|
| 504 |
-
correct_count = result_data.get(
|
| 505 |
-
total_attempted = result_data.get(
|
| 506 |
-
score = result_data.get(
|
| 507 |
|
| 508 |
final_status = (
|
| 509 |
-
f"
|
| 510 |
f"User: {result_data.get('username', username)}\n"
|
| 511 |
f"Overall Score: {score}% "
|
| 512 |
f"({correct_count}/{total_attempted} correct)\n"
|
| 513 |
f"Message: {result_data.get('message', 'No message received.')}"
|
| 514 |
)
|
| 515 |
logger.info("Submission successful.")
|
| 516 |
-
|
| 517 |
-
# Create results DataFrame
|
| 518 |
results_df = pd.DataFrame(results_log)
|
| 519 |
-
|
| 520 |
-
# Calculate statistics if we have results
|
| 521 |
-
if not results_df.empty:
|
| 522 |
-
success_count = sum(1 for status in results_df.get("Status", []) if status == "Success")
|
| 523 |
-
error_count = sum(1 for status in results_df.get("Status", []) if status == "Error")
|
| 524 |
-
|
| 525 |
-
try:
|
| 526 |
-
avg_time = sum(float(time.strip('s')) for time in results_df["Processing Time"]) / len(results_df)
|
| 527 |
-
avg_time_str = f"{avg_time:.2f}s"
|
| 528 |
-
except:
|
| 529 |
-
avg_time_str = "N/A"
|
| 530 |
-
|
| 531 |
-
logger.info(f"Processed {len(results_df)} questions: {success_count} successes, {error_count} errors, avg time: {avg_time_str}")
|
| 532 |
-
|
| 533 |
-
# Complete progress and return results
|
| 534 |
-
progress(1.0, desc="Evaluation complete!")
|
| 535 |
return final_status, results_df
|
| 536 |
-
|
| 537 |
except requests.exceptions.HTTPError as e:
|
| 538 |
error_detail = f"Server responded with status {e.response.status_code}."
|
| 539 |
try:
|
|
@@ -543,322 +185,88 @@ def run_and_submit_all(
|
|
| 543 |
error_detail += f" Response: {e.response.text[:500]}"
|
| 544 |
status_message = f"Submission Failed: {error_detail}"
|
| 545 |
logger.error(status_message)
|
| 546 |
-
results_df = pd.DataFrame(results_log)
|
| 547 |
return status_message, results_df
|
| 548 |
-
|
| 549 |
except requests.exceptions.Timeout:
|
| 550 |
status_message = "Submission Failed: The request timed out."
|
| 551 |
logger.error(status_message)
|
| 552 |
-
results_df = pd.DataFrame(results_log)
|
| 553 |
return status_message, results_df
|
| 554 |
-
|
| 555 |
except requests.exceptions.RequestException as e:
|
| 556 |
status_message = f"Submission Failed: Network error - {e}"
|
| 557 |
logger.error(status_message)
|
| 558 |
-
results_df = pd.DataFrame(results_log)
|
| 559 |
return status_message, results_df
|
| 560 |
-
|
| 561 |
except Exception as e:
|
| 562 |
status_message = f"An unexpected error occurred during submission: {e}"
|
| 563 |
logger.error(status_message)
|
| 564 |
logger.error(traceback.format_exc())
|
| 565 |
-
results_df = pd.DataFrame(results_log)
|
| 566 |
return status_message, results_df
|
| 567 |
|
| 568 |
-
def create_evaluation_tab(demo, agent_instance):
|
| 569 |
-
"""
|
| 570 |
-
Creates and configures the evaluation tab for the Gradio interface.
|
| 571 |
-
|
| 572 |
-
Args:
|
| 573 |
-
demo: The Gradio Blocks instance to add the tab to
|
| 574 |
-
agent_instance: The GAIA Agent instance to use for evaluation
|
| 575 |
-
|
| 576 |
-
Returns:
|
| 577 |
-
None - modifies the demo in place
|
| 578 |
-
"""
|
| 579 |
-
with gr.Tab("Run Evaluation"):
|
| 580 |
-
gr.Markdown("## GAIA Agent Evaluation")
|
| 581 |
-
|
| 582 |
-
with gr.Row():
|
| 583 |
-
with gr.Column(scale=2):
|
| 584 |
-
gr.Markdown(
|
| 585 |
-
"""
|
| 586 |
-
This tab allows you to evaluate the GAIA Agent against a set of benchmark questions.
|
| 587 |
-
|
| 588 |
-
The evaluation process:
|
| 589 |
-
1. Initializes the GAIA Agent
|
| 590 |
-
2. Fetches questions from the evaluation server
|
| 591 |
-
3. Runs the agent on each question
|
| 592 |
-
4. Submits the answers and retrieves scores
|
| 593 |
-
5. Displays detailed results
|
| 594 |
-
|
| 595 |
-
You can either log in with your Hugging Face account or enter a username for the submission.
|
| 596 |
-
"""
|
| 597 |
-
)
|
| 598 |
-
|
| 599 |
-
with gr.Column(scale=1):
|
| 600 |
-
# OAuth login option
|
| 601 |
-
gr.LoginButton(visible=True)
|
| 602 |
-
|
| 603 |
-
# Alternative username field
|
| 604 |
-
username_input = gr.Textbox(
|
| 605 |
-
label="Username for Submission",
|
| 606 |
-
placeholder="Enter your username here if not using HF login",
|
| 607 |
-
lines=1
|
| 608 |
-
)
|
| 609 |
-
|
| 610 |
-
# Optional code URL field
|
| 611 |
-
code_url_input = gr.Textbox(
|
| 612 |
-
label="Agent Code URL (optional)",
|
| 613 |
-
placeholder="URL to your code repository",
|
| 614 |
-
lines=1,
|
| 615 |
-
value=""
|
| 616 |
-
)
|
| 617 |
-
|
| 618 |
-
# Run button
|
| 619 |
-
run_button = gr.Button("Run Evaluation & Submit All Answers", variant="primary")
|
| 620 |
-
|
| 621 |
-
# Output components
|
| 622 |
-
status_output = gr.Textbox(
|
| 623 |
-
label="Evaluation Status / Submission Result",
|
| 624 |
-
lines=5,
|
| 625 |
-
interactive=False
|
| 626 |
-
)
|
| 627 |
-
|
| 628 |
-
results_table = gr.DataFrame(
|
| 629 |
-
label="Questions and Agent Answers",
|
| 630 |
-
wrap=True,
|
| 631 |
-
visible=True
|
| 632 |
-
)
|
| 633 |
-
|
| 634 |
-
# Set up event handler
|
| 635 |
-
run_button.click(
|
| 636 |
-
fn=run_and_submit_all,
|
| 637 |
-
inputs=[
|
| 638 |
-
username_input,
|
| 639 |
-
code_url_input,
|
| 640 |
-
agent_instance
|
| 641 |
-
],
|
| 642 |
-
outputs=[status_output, results_table]
|
| 643 |
-
)
|
| 644 |
|
| 645 |
-
|
| 646 |
-
|
| 647 |
-
|
| 648 |
-
|
| 649 |
-
|
| 650 |
-
|
| 651 |
-
|
| 652 |
-
|
| 653 |
-
|
| 654 |
-
|
| 655 |
-
|
| 656 |
-
|
| 657 |
-
|
| 658 |
-
|
| 659 |
-
|
| 660 |
-
|
| 661 |
-
|
| 662 |
-
|
| 663 |
-
|
| 664 |
-
|
| 665 |
-
|
| 666 |
-
|
| 667 |
-
|
| 668 |
-
|
| 669 |
-
|
| 670 |
-
|
| 671 |
-
gr.Markdown(f"### Environment Status\n{env_message}")
|
| 672 |
-
if agent_instance is None:
|
| 673 |
-
gr.Markdown("❌ **Agent initialization failed.** Some features may not work correctly.")
|
| 674 |
-
else:
|
| 675 |
-
gr.Markdown("✅ **Agent initialized successfully.**")
|
| 676 |
-
|
| 677 |
-
# Create tabs for chat and evaluation
|
| 678 |
-
with gr.Tabs():
|
| 679 |
-
# Chat tab
|
| 680 |
-
with gr.Tab("Chat"):
|
| 681 |
-
# Chat interface
|
| 682 |
-
with gr.Group():
|
| 683 |
-
chatbot = gr.Chatbot(
|
| 684 |
-
label="Conversation",
|
| 685 |
-
elem_id="gaia_chatbot",
|
| 686 |
-
height=450,
|
| 687 |
-
show_copy_button=True
|
| 688 |
-
)
|
| 689 |
-
|
| 690 |
-
with gr.Row():
|
| 691 |
-
msg = gr.Textbox(
|
| 692 |
-
label="Your question",
|
| 693 |
-
placeholder="Ask me anything...",
|
| 694 |
-
lines=2,
|
| 695 |
-
max_lines=10,
|
| 696 |
-
show_label=False,
|
| 697 |
-
container=False
|
| 698 |
-
)
|
| 699 |
-
submit_btn = gr.Button("Submit", variant="primary")
|
| 700 |
-
|
| 701 |
-
with gr.Row():
|
| 702 |
-
clear_btn = gr.Button("Clear Conversation")
|
| 703 |
-
error_box = gr.Textbox(
|
| 704 |
-
label="Status",
|
| 705 |
-
visible=True,
|
| 706 |
-
interactive=False,
|
| 707 |
-
show_label=False
|
| 708 |
-
)
|
| 709 |
-
|
| 710 |
-
# Add evaluation tab
|
| 711 |
-
create_evaluation_tab(demo, agent_instance)
|
| 712 |
-
|
| 713 |
-
# About section
|
| 714 |
-
with gr.Accordion("About GAIA Agent", open=False):
|
| 715 |
-
gr.Markdown(
|
| 716 |
-
"""
|
| 717 |
-
## About GAIA Agent
|
| 718 |
-
|
| 719 |
-
The GAIA (Grounded AI Assistant) Agent is designed to provide helpful and accurate responses by
|
| 720 |
-
leveraging web search tools and robust reasoning capabilities.
|
| 721 |
-
|
| 722 |
-
### Key Features:
|
| 723 |
-
|
| 724 |
-
- **Web Search:** Retrieves real-time information from the web
|
| 725 |
-
- **Content Extraction:** Can extract and analyze content from web pages
|
| 726 |
-
- **Structured Approach:** Uses a systematic process for analyzing questions and formulating answers
|
| 727 |
-
- **Error Handling:** Includes robust error handling and fallback mechanisms
|
| 728 |
-
|
| 729 |
-
### Resources:
|
| 730 |
-
|
| 731 |
-
- The agent is built with LangChain and LangGraph
|
| 732 |
-
- Uses OpenAI models for reasoning and processing
|
| 733 |
-
|
| 734 |
-
### Attribution:
|
| 735 |
-
|
| 736 |
-
GAIA Agent is developed as part of the Hugging Face Agents course.
|
| 737 |
-
"""
|
| 738 |
-
)
|
| 739 |
-
|
| 740 |
-
# Set up event handlers with error handling wrapper
|
| 741 |
-
def safe_handler(fn):
|
| 742 |
-
"""Wrapper to provide additional error handling for event handlers"""
|
| 743 |
-
def wrapped_fn(*args, **kwargs):
|
| 744 |
-
try:
|
| 745 |
-
return fn(*args, **kwargs)
|
| 746 |
-
except Exception as e:
|
| 747 |
-
error_type = type(e).__name__
|
| 748 |
-
logger.error(f"Unhandled error in UI event handler: {error_type}: {str(e)}")
|
| 749 |
-
logger.error(traceback.format_exc())
|
| 750 |
-
|
| 751 |
-
# Determine what function was called for better error messages
|
| 752 |
-
if fn.__name__ == "process_question":
|
| 753 |
-
# For process_question failures, return original history + error message
|
| 754 |
-
history = args[1] if len(args) > 1 and isinstance(args[1], list) else []
|
| 755 |
-
user_msg = args[0] if args and isinstance(args[0], str) else "Unknown question"
|
| 756 |
-
return history + [(user_msg, f"The system encountered an unexpected error. Please try again later.")], f"UI handler error: {error_type}"
|
| 757 |
-
else:
|
| 758 |
-
# For other function failures (e.g., clear_conversation)
|
| 759 |
-
return [], f"Operation failed due to a system error: {error_type}"
|
| 760 |
-
# Preserve original function metadata
|
| 761 |
-
wrapped_fn.__name__ = fn.__name__
|
| 762 |
-
wrapped_fn.__doc__ = fn.__doc__
|
| 763 |
-
return wrapped_fn
|
| 764 |
-
|
| 765 |
-
# Use the safe handler wrapper for all event handlers
|
| 766 |
-
try:
|
| 767 |
-
# Set up submit button handler
|
| 768 |
-
submit_btn.click(
|
| 769 |
-
fn=safe_handler(process_question),
|
| 770 |
-
inputs=[msg, chatbot, agent_instance],
|
| 771 |
-
outputs=[chatbot, error_box],
|
| 772 |
-
api_name="submit"
|
| 773 |
-
)
|
| 774 |
-
logger.info("Submit button handler registered successfully")
|
| 775 |
-
|
| 776 |
-
# Set up message input submission handler
|
| 777 |
-
msg.submit(
|
| 778 |
-
fn=safe_handler(process_question),
|
| 779 |
-
inputs=[msg, chatbot, agent_instance],
|
| 780 |
-
outputs=[chatbot, error_box],
|
| 781 |
-
api_name="ask"
|
| 782 |
-
)
|
| 783 |
-
logger.info("Message submission handler registered successfully")
|
| 784 |
-
|
| 785 |
-
# Set up clear button handler
|
| 786 |
-
clear_btn.click(
|
| 787 |
-
fn=safe_handler(clear_conversation),
|
| 788 |
-
inputs=[agent_instance],
|
| 789 |
-
outputs=[chatbot, error_box],
|
| 790 |
-
api_name="clear"
|
| 791 |
-
)
|
| 792 |
-
logger.info("Clear button handler registered successfully")
|
| 793 |
-
except Exception as handler_err:
|
| 794 |
-
logger.error(f"Failed to register event handlers: {str(handler_err)}")
|
| 795 |
-
logger.error(traceback.format_exc())
|
| 796 |
-
# Add error message to the interface
|
| 797 |
-
error_box.value = "Warning: Some interface features may not be working correctly due to initialization errors."
|
| 798 |
-
|
| 799 |
-
return demo
|
| 800 |
|
| 801 |
if __name__ == "__main__":
|
| 802 |
-
|
| 803 |
-
|
| 804 |
-
|
| 805 |
-
|
| 806 |
-
|
| 807 |
-
|
| 808 |
-
|
| 809 |
-
|
| 810 |
-
|
| 811 |
-
|
| 812 |
-
|
| 813 |
-
|
| 814 |
-
|
| 815 |
-
|
| 816 |
-
|
| 817 |
-
|
| 818 |
-
|
| 819 |
-
|
| 820 |
-
|
| 821 |
-
|
| 822 |
-
|
| 823 |
-
|
| 824 |
-
|
| 825 |
-
|
| 826 |
-
|
| 827 |
-
|
| 828 |
-
|
| 829 |
-
|
| 830 |
-
|
| 831 |
-
|
| 832 |
-
|
| 833 |
-
|
| 834 |
-
|
| 835 |
-
|
| 836 |
-
|
| 837 |
-
|
| 838 |
-
|
| 839 |
-
|
| 840 |
-
except Exception as e:
|
| 841 |
-
error_type = type(e).__name__
|
| 842 |
-
logger.critical(f"Fatal error launching application: {error_type}: {str(e)}")
|
| 843 |
-
logger.critical(traceback.format_exc())
|
| 844 |
-
|
| 845 |
-
# Create a minimal emergency interface to show the error
|
| 846 |
-
try:
|
| 847 |
-
with gr.Blocks(title="GAIA Agent - Error") as emergency_ui:
|
| 848 |
-
gr.Markdown("# ❌ GAIA Agent - Critical Error")
|
| 849 |
-
gr.Markdown(f"**Error Type**: {error_type}")
|
| 850 |
-
gr.Markdown(f"**Error Message**: {str(e)}")
|
| 851 |
-
gr.Markdown("Please check the application logs for details.")
|
| 852 |
-
gr.Markdown("### Common Solutions:")
|
| 853 |
-
gr.Markdown("- Check that required environment variables are set")
|
| 854 |
-
gr.Markdown("- Verify network connectivity for API access")
|
| 855 |
-
gr.Markdown("- Ensure all dependencies are correctly installed")
|
| 856 |
-
|
| 857 |
-
emergency_ui.launch(debug=True, share=False)
|
| 858 |
-
except:
|
| 859 |
-
# If even the emergency UI fails, print to console as last resort
|
| 860 |
-
print("=" * 80)
|
| 861 |
-
print("CRITICAL ERROR: Could not launch GAIA Agent application")
|
| 862 |
-
print(f"Error: {error_type}: {str(e)}")
|
| 863 |
-
print("Check log files for details.")
|
| 864 |
-
print("=" * 80)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
import gradio as gr
|
| 3 |
+
import requests
|
| 4 |
+
import pandas as pd
|
| 5 |
import logging
|
| 6 |
import time
|
| 7 |
import traceback
|
|
|
|
|
|
|
| 8 |
from typing import Dict, Any, Optional, Tuple, List
|
| 9 |
|
| 10 |
+
# (Keep Constants as is)
|
| 11 |
+
# --- Constants ---
|
| 12 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 13 |
|
| 14 |
+
# --- Import our GAIA Agent ---
|
| 15 |
from gaia_agent import GAIAAgent
|
| 16 |
|
| 17 |
# Configure logging
|
|
|
|
| 19 |
level=logging.INFO,
|
| 20 |
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
| 21 |
)
|
| 22 |
+
logger = logging.getLogger("gaia_evaluation")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
|
| 24 |
+
# Initialize the GAIA Agent
|
| 25 |
+
def initialize_agent():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
try:
|
|
|
|
| 27 |
agent = GAIAAgent()
|
| 28 |
+
logger.info("GAIA Agent initialized successfully")
|
|
|
|
| 29 |
return agent
|
| 30 |
except Exception as e:
|
| 31 |
+
logger.error(f"Error initializing GAIA Agent: {e}")
|
|
|
|
| 32 |
logger.error(traceback.format_exc())
|
| 33 |
return None
|
| 34 |
|
| 35 |
+
def run_and_submit_all(profile: Optional[gr.OAuthProfile] = None):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
"""
|
| 37 |
Fetches all questions, runs the GAIA Agent on them, submits all answers,
|
| 38 |
and displays the results.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
"""
|
|
|
|
|
|
|
|
|
|
| 40 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
| 41 |
space_id = os.getenv("SPACE_ID", "")
|
| 42 |
+
|
|
|
|
| 43 |
if profile:
|
| 44 |
username = profile.username
|
| 45 |
logger.info(f"User logged in: {username}")
|
|
|
|
|
|
|
|
|
|
| 46 |
else:
|
| 47 |
+
logger.warning("User not logged in.")
|
| 48 |
+
return "Please Login to Hugging Face with the button.", None
|
| 49 |
+
|
|
|
|
| 50 |
api_url = DEFAULT_API_URL
|
| 51 |
questions_url = f"{api_url}/questions"
|
| 52 |
submit_url = f"{api_url}/submit"
|
| 53 |
+
|
| 54 |
+
# 1. Instantiate Agent
|
| 55 |
+
try:
|
| 56 |
+
agent = initialize_agent()
|
| 57 |
+
if agent is None:
|
| 58 |
+
error_msg = "Error initializing GAIA Agent. Check logs for details."
|
| 59 |
+
logger.error(error_msg)
|
| 60 |
+
return error_msg, None
|
| 61 |
+
except Exception as e:
|
| 62 |
+
error_msg = f"Error instantiating agent: {e}"
|
| 63 |
logger.error(error_msg)
|
| 64 |
return error_msg, None
|
| 65 |
+
|
| 66 |
+
# Link to the code repository in Hugging Face Space
|
| 67 |
+
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
| 68 |
+
logger.info(f"Using agent code URL: {agent_code}")
|
| 69 |
+
|
| 70 |
+
# 2. Fetch Questions
|
| 71 |
logger.info(f"Fetching questions from: {questions_url}")
|
| 72 |
try:
|
| 73 |
response = requests.get(questions_url, timeout=15)
|
| 74 |
response.raise_for_status()
|
| 75 |
questions_data = response.json()
|
| 76 |
if not questions_data:
|
| 77 |
+
logger.warning("Fetched questions list is empty.")
|
| 78 |
+
return "Fetched questions list is empty or invalid format.", None
|
| 79 |
+
logger.info(f"Fetched {len(questions_data)} questions.")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
except requests.exceptions.RequestException as e:
|
| 81 |
error_msg = f"Error fetching questions: {e}"
|
| 82 |
logger.error(error_msg)
|
| 83 |
return error_msg, None
|
| 84 |
except requests.exceptions.JSONDecodeError as e:
|
| 85 |
+
error_msg = f"Error decoding JSON response from questions endpoint: {e}"
|
| 86 |
+
logger.error(f"{error_msg}\nResponse text: {response.text[:500]}")
|
| 87 |
+
return error_msg, None
|
| 88 |
except Exception as e:
|
| 89 |
error_msg = f"An unexpected error occurred fetching questions: {e}"
|
| 90 |
logger.error(error_msg)
|
| 91 |
logger.error(traceback.format_exc())
|
| 92 |
return error_msg, None
|
| 93 |
+
|
| 94 |
+
# 3. Run the GAIA Agent on all questions
|
| 95 |
results_log = []
|
| 96 |
answers_payload = []
|
| 97 |
+
logger.info(f"Running agent on {len(questions_data)} questions...")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 98 |
|
| 99 |
for i, item in enumerate(questions_data):
|
| 100 |
+
# Extract question information
|
| 101 |
task_id = item.get("task_id", item.get("id", f"q{i+1}"))
|
| 102 |
question_text = item.get("question")
|
| 103 |
|
| 104 |
if not task_id or question_text is None:
|
| 105 |
logger.warning(f"Skipping item with missing task_id or question: {item}")
|
| 106 |
continue
|
|
|
|
|
|
|
|
|
|
| 107 |
|
| 108 |
+
logger.info(f"Processing question {i+1}/{len(questions_data)}: {question_text[:50]}...")
|
| 109 |
+
start_time = time.time()
|
| 110 |
|
| 111 |
try:
|
| 112 |
+
# Process the question with the GAIA Agent
|
| 113 |
+
submitted_answer = agent.process_question(question_text)
|
| 114 |
+
processing_time = time.time() - start_time
|
|
|
|
| 115 |
|
| 116 |
+
# Add to submission payload
|
| 117 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 118 |
+
|
| 119 |
+
# Store for results display
|
| 120 |
results_log.append({
|
| 121 |
"Task ID": task_id,
|
| 122 |
"Question": question_text,
|
|
|
|
| 125 |
"Status": "Success"
|
| 126 |
})
|
| 127 |
|
| 128 |
+
logger.info(f"Question {i+1} processed successfully in {processing_time:.2f}s")
|
| 129 |
|
| 130 |
except Exception as e:
|
| 131 |
error_msg = f"Error running agent on task {task_id}: {e}"
|
|
|
|
| 136 |
"Task ID": task_id,
|
| 137 |
"Question": question_text,
|
| 138 |
"Submitted Answer": f"AGENT ERROR: {e}",
|
| 139 |
+
"Processing Time": f"{time.time() - start_time:.2f}s",
|
| 140 |
"Status": "Error"
|
| 141 |
})
|
| 142 |
+
|
| 143 |
# Check if we have answers to submit
|
| 144 |
if not answers_payload:
|
| 145 |
+
logger.warning("Agent did not produce any answers to submit.")
|
| 146 |
+
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 147 |
+
|
| 148 |
+
# 4. Prepare Submission
|
|
|
|
|
|
|
|
|
|
|
|
|
| 149 |
submission_data = {
|
| 150 |
+
"username": username.strip(),
|
| 151 |
+
"agent_code": agent_code,
|
| 152 |
"answers": answers_payload
|
| 153 |
}
|
|
|
|
| 154 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
| 155 |
logger.info(status_update)
|
| 156 |
+
|
| 157 |
+
# 5. Submit
|
| 158 |
+
logger.info(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
|
|
|
|
|
|
| 159 |
try:
|
| 160 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
| 161 |
response.raise_for_status()
|
| 162 |
result_data = response.json()
|
| 163 |
|
| 164 |
+
# Extract results data
|
| 165 |
+
correct_count = result_data.get("correct_count", 0)
|
| 166 |
+
total_attempted = result_data.get("total_attempted", 0)
|
| 167 |
+
score = result_data.get("score", "N/A")
|
| 168 |
|
| 169 |
final_status = (
|
| 170 |
+
f"Submission Successful!\n"
|
| 171 |
f"User: {result_data.get('username', username)}\n"
|
| 172 |
f"Overall Score: {score}% "
|
| 173 |
f"({correct_count}/{total_attempted} correct)\n"
|
| 174 |
f"Message: {result_data.get('message', 'No message received.')}"
|
| 175 |
)
|
| 176 |
logger.info("Submission successful.")
|
|
|
|
|
|
|
| 177 |
results_df = pd.DataFrame(results_log)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 178 |
return final_status, results_df
|
|
|
|
| 179 |
except requests.exceptions.HTTPError as e:
|
| 180 |
error_detail = f"Server responded with status {e.response.status_code}."
|
| 181 |
try:
|
|
|
|
| 185 |
error_detail += f" Response: {e.response.text[:500]}"
|
| 186 |
status_message = f"Submission Failed: {error_detail}"
|
| 187 |
logger.error(status_message)
|
| 188 |
+
results_df = pd.DataFrame(results_log)
|
| 189 |
return status_message, results_df
|
|
|
|
| 190 |
except requests.exceptions.Timeout:
|
| 191 |
status_message = "Submission Failed: The request timed out."
|
| 192 |
logger.error(status_message)
|
| 193 |
+
results_df = pd.DataFrame(results_log)
|
| 194 |
return status_message, results_df
|
|
|
|
| 195 |
except requests.exceptions.RequestException as e:
|
| 196 |
status_message = f"Submission Failed: Network error - {e}"
|
| 197 |
logger.error(status_message)
|
| 198 |
+
results_df = pd.DataFrame(results_log)
|
| 199 |
return status_message, results_df
|
|
|
|
| 200 |
except Exception as e:
|
| 201 |
status_message = f"An unexpected error occurred during submission: {e}"
|
| 202 |
logger.error(status_message)
|
| 203 |
logger.error(traceback.format_exc())
|
| 204 |
+
results_df = pd.DataFrame(results_log)
|
| 205 |
return status_message, results_df
|
| 206 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 207 |
|
| 208 |
+
# --- Build Gradio Interface using Blocks ---
|
| 209 |
+
with gr.Blocks() as demo:
|
| 210 |
+
gr.Markdown("# GAIA Agent Evaluation Runner")
|
| 211 |
+
gr.Markdown(
|
| 212 |
+
"""
|
| 213 |
+
**Instructions:**
|
| 214 |
+
|
| 215 |
+
1. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
|
| 216 |
+
2. Click 'Run Evaluation & Submit All Answers' to fetch questions, run the GAIA agent, submit answers, and see the score.
|
| 217 |
+
|
| 218 |
+
---
|
| 219 |
+
**Note:** Running the evaluation may take some time as the agent processes all questions. Please be patient.
|
| 220 |
+
"""
|
| 221 |
+
)
|
| 222 |
+
|
| 223 |
+
gr.LoginButton()
|
| 224 |
+
|
| 225 |
+
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
| 226 |
+
|
| 227 |
+
status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
|
| 228 |
+
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
| 229 |
+
|
| 230 |
+
run_button.click(
|
| 231 |
+
fn=run_and_submit_all,
|
| 232 |
+
outputs=[status_output, results_table]
|
| 233 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 234 |
|
| 235 |
if __name__ == "__main__":
|
| 236 |
+
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
| 237 |
+
# Check for SPACE_HOST and SPACE_ID at startup for information
|
| 238 |
+
space_host_startup = os.getenv("SPACE_HOST")
|
| 239 |
+
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
| 240 |
+
|
| 241 |
+
if space_host_startup:
|
| 242 |
+
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
| 243 |
+
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
| 244 |
+
else:
|
| 245 |
+
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
| 246 |
+
|
| 247 |
+
if space_id_startup: # Print repo URLs if SPACE_ID is found
|
| 248 |
+
print(f"✅ SPACE_ID found: {space_id_startup}")
|
| 249 |
+
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
| 250 |
+
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
| 251 |
+
else:
|
| 252 |
+
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
| 253 |
+
|
| 254 |
+
print("-"*(60 + len(" App Starting ")) + "\n")
|
| 255 |
+
|
| 256 |
+
print("Launching Gradio Interface for GAIA Agent Evaluation...")
|
| 257 |
+
# Determine launch parameters based on environment
|
| 258 |
+
is_running_in_space = bool(space_host_startup and space_id_startup)
|
| 259 |
+
|
| 260 |
+
if is_running_in_space:
|
| 261 |
+
# Production settings for Hugging Face Space
|
| 262 |
+
demo.launch(
|
| 263 |
+
debug=False,
|
| 264 |
+
share=False,
|
| 265 |
+
server_name="0.0.0.0"
|
| 266 |
+
)
|
| 267 |
+
else:
|
| 268 |
+
# Development settings for local testing
|
| 269 |
+
demo.launch(
|
| 270 |
+
debug=True,
|
| 271 |
+
share=False
|
| 272 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|