Spaces:
Runtime error
Runtime error
Fix
Browse files
app.py
CHANGED
@@ -22,14 +22,7 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
|
22 |
|
23 |
@tool
|
24 |
def serper_search(query: str) -> str:
|
25 |
-
"""Enhanced web search using Serper API with better result processing
|
26 |
-
|
27 |
-
Args:
|
28 |
-
query: The search query
|
29 |
-
|
30 |
-
Returns:
|
31 |
-
Formatted search results with relevance scoring
|
32 |
-
"""
|
33 |
try:
|
34 |
api_key = os.getenv("SERPER_API_KEY")
|
35 |
if not api_key:
|
@@ -47,7 +40,6 @@ def serper_search(query: str) -> str:
|
|
47 |
data = response.json()
|
48 |
results = []
|
49 |
|
50 |
-
# Process knowledge graph first (highest priority)
|
51 |
if 'knowledgeGraph' in data:
|
52 |
kg = data['knowledgeGraph']
|
53 |
kg_info = f"KNOWLEDGE GRAPH: {kg.get('title', '')} - {kg.get('description', '')}"
|
@@ -56,30 +48,25 @@ def serper_search(query: str) -> str:
|
|
56 |
kg_info += f"\n{key}: {value}"
|
57 |
results.append(kg_info + "\n")
|
58 |
|
59 |
-
# Process organic results with enhanced filtering
|
60 |
if 'organic' in data:
|
61 |
for i, item in enumerate(data['organic'][:7]):
|
62 |
title = item.get('title', '')
|
63 |
snippet = item.get('snippet', '')
|
64 |
link = item.get('link', '')
|
65 |
-
|
66 |
-
# Enhanced result formatting
|
67 |
result_text = f"RESULT {i+1}:\nTitle: {title}\nSnippet: {snippet}\nURL: {link}\n"
|
68 |
|
69 |
-
|
70 |
-
if re.search(r'\d{4}', snippet): # Years
|
71 |
years = re.findall(r'\b(19|20)\d{2}\b', snippet)
|
72 |
if years:
|
73 |
result_text += f"Years mentioned: {', '.join(years)}\n"
|
74 |
|
75 |
-
if re.search(r'\$[\d,]+', snippet):
|
76 |
amounts = re.findall(r'\$[\d,]+(?:\.\d{2})?', snippet)
|
77 |
if amounts:
|
78 |
result_text += f"Amounts: {', '.join(amounts)}\n"
|
79 |
|
80 |
results.append(result_text)
|
81 |
|
82 |
-
# Add people also ask if available
|
83 |
if 'peopleAlsoAsk' in data:
|
84 |
paa = "\nPEOPLE ALSO ASK:\n"
|
85 |
for item in data['peopleAlsoAsk'][:3]:
|
@@ -92,19 +79,10 @@ def serper_search(query: str) -> str:
|
|
92 |
return f"Search error: {str(e)}"
|
93 |
|
94 |
@tool
|
95 |
-
def
|
96 |
-
"""Enhanced Wikipedia search with multiple strategies
|
97 |
-
|
98 |
-
Args:
|
99 |
-
query: Wikipedia search query
|
100 |
-
|
101 |
-
Returns:
|
102 |
-
Comprehensive Wikipedia information
|
103 |
-
"""
|
104 |
try:
|
105 |
results = []
|
106 |
-
|
107 |
-
# Strategy 1: Direct page lookup
|
108 |
clean_query = query.replace(" ", "_")
|
109 |
direct_url = f"https://en.wikipedia.org/api/rest_v1/page/summary/{clean_query}"
|
110 |
|
@@ -116,12 +94,10 @@ def wikipedia_enhanced_search(query: str) -> str:
|
|
116 |
summary = f"WIKIPEDIA DIRECT MATCH:\nTitle: {data.get('title', '')}\n"
|
117 |
summary += f"Extract: {data.get('extract', '')}\n"
|
118 |
|
119 |
-
# Add coordinates if available
|
120 |
if 'coordinates' in data:
|
121 |
coords = data['coordinates']
|
122 |
summary += f"Coordinates: {coords.get('lat', '')}, {coords.get('lon', '')}\n"
|
123 |
|
124 |
-
# Add birth/death dates if available
|
125 |
extract = data.get('extract', '')
|
126 |
birth_match = re.search(r'born[^)]*(\d{1,2}\s+\w+\s+\d{4})', extract, re.IGNORECASE)
|
127 |
if birth_match:
|
@@ -135,7 +111,6 @@ def wikipedia_enhanced_search(query: str) -> str:
|
|
135 |
except:
|
136 |
pass
|
137 |
|
138 |
-
# Strategy 2: Search API for multiple results
|
139 |
search_url = "https://en.wikipedia.org/w/api.php"
|
140 |
search_params = {
|
141 |
"action": "query",
|
@@ -152,14 +127,12 @@ def wikipedia_enhanced_search(query: str) -> str:
|
|
152 |
if 'query' in data and 'search' in data['query']:
|
153 |
search_results = "WIKIPEDIA SEARCH RESULTS:\n"
|
154 |
for item in data['query']['search']:
|
155 |
-
# Clean HTML tags from snippet
|
156 |
snippet = re.sub(r'<[^>]+>', '', item.get('snippet', ''))
|
157 |
search_results += f"• {item['title']}: {snippet}\n"
|
158 |
results.append(search_results)
|
159 |
except:
|
160 |
pass
|
161 |
|
162 |
-
# Strategy 3: Try opensearch for suggestions
|
163 |
opensearch_url = "https://en.wikipedia.org/w/api.php"
|
164 |
opensearch_params = {
|
165 |
"action": "opensearch",
|
@@ -171,7 +144,7 @@ def wikipedia_enhanced_search(query: str) -> str:
|
|
171 |
try:
|
172 |
response = requests.get(opensearch_url, params=opensearch_params, timeout=10)
|
173 |
data = response.json()
|
174 |
-
if len(data) >= 4 and data[1]:
|
175 |
suggestions = "WIKIPEDIA SUGGESTIONS:\n"
|
176 |
for i, (title, desc, url) in enumerate(zip(data[1], data[2], data[3])):
|
177 |
suggestions += f"{i+1}. {title}: {desc}\n"
|
@@ -185,17 +158,9 @@ def wikipedia_enhanced_search(query: str) -> str:
|
|
185 |
return f"Wikipedia search error: {str(e)}"
|
186 |
|
187 |
@tool
|
188 |
-
def
|
189 |
-
"""Enhanced YouTube video analyzer with transcript extraction
|
190 |
-
|
191 |
-
Args:
|
192 |
-
url: YouTube video URL
|
193 |
-
|
194 |
-
Returns:
|
195 |
-
Comprehensive video analysis
|
196 |
-
"""
|
197 |
try:
|
198 |
-
# Extract video ID
|
199 |
video_id_match = re.search(r'(?:v=|/|youtu\.be/)([A-Za-z0-9_-]{11})', url)
|
200 |
if not video_id_match:
|
201 |
return "Invalid YouTube URL format"
|
@@ -203,7 +168,6 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
203 |
video_id = video_id_match.group(1)
|
204 |
results = []
|
205 |
|
206 |
-
# Get basic video info via oEmbed
|
207 |
try:
|
208 |
oembed_url = f"https://www.youtube.com/oembed?url=https://www.youtube.com/watch?v={video_id}&format=json"
|
209 |
response = requests.get(oembed_url, timeout=15)
|
@@ -212,7 +176,6 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
212 |
data = response.json()
|
213 |
basic_info = f"VIDEO INFO:\nTitle: {data.get('title', '')}\nAuthor: {data.get('author_name', '')}\n"
|
214 |
|
215 |
-
# Extract duration if available in title/description patterns
|
216 |
title = data.get('title', '').lower()
|
217 |
if 'minute' in title or 'min' in title:
|
218 |
duration_match = re.search(r'(\d+)\s*(?:minute|min)', title)
|
@@ -223,7 +186,6 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
223 |
except:
|
224 |
pass
|
225 |
|
226 |
-
# Enhanced content analysis through page scraping
|
227 |
try:
|
228 |
video_url = f"https://www.youtube.com/watch?v={video_id}"
|
229 |
headers = {
|
@@ -234,34 +196,28 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
234 |
if response.status_code == 200:
|
235 |
content = response.text
|
236 |
|
237 |
-
# Extract view count
|
238 |
view_match = re.search(r'"viewCount":"(\d+)"', content)
|
239 |
if view_match:
|
240 |
views = int(view_match.group(1))
|
241 |
results.append(f"View count: {views:,}")
|
242 |
|
243 |
-
# Extract upload date
|
244 |
upload_match = re.search(r'"uploadDate":"([^"]+)"', content)
|
245 |
if upload_match:
|
246 |
results.append(f"Upload date: {upload_match.group(1)}")
|
247 |
|
248 |
-
# Look for specific content patterns
|
249 |
content_lower = content.lower()
|
250 |
|
251 |
-
# Bird counting for ornithology videos
|
252 |
if "bird" in content_lower:
|
253 |
bird_numbers = re.findall(r'\b(\d+)\s+(?:bird|species|individual)', content_lower)
|
254 |
if bird_numbers:
|
255 |
results.append(f"Bird counts found: {', '.join(bird_numbers)}")
|
256 |
|
257 |
-
# Duration extraction from JSON-LD
|
258 |
duration_match = re.search(r'"duration":"PT(\d+)M(\d+)S"', content)
|
259 |
if duration_match:
|
260 |
minutes = int(duration_match.group(1))
|
261 |
seconds = int(duration_match.group(2))
|
262 |
results.append(f"Exact duration: {minutes}:{seconds:02d}")
|
263 |
|
264 |
-
# Extract description
|
265 |
desc_patterns = [
|
266 |
r'"description":{"simpleText":"([^"]+)"}',
|
267 |
r'"shortDescription":"([^"]+)"'
|
@@ -270,7 +226,7 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
270 |
for pattern in desc_patterns:
|
271 |
desc_match = re.search(pattern, content)
|
272 |
if desc_match:
|
273 |
-
description = desc_match.group(1)[:500]
|
274 |
results.append(f"Description excerpt: {description}")
|
275 |
break
|
276 |
|
@@ -283,22 +239,13 @@ def youtube_enhanced_analyzer(url: str) -> str:
|
|
283 |
return f"YouTube analysis error: {str(e)}"
|
284 |
|
285 |
@tool
|
286 |
-
def
|
287 |
-
"""Advanced text processing for various linguistic operations
|
288 |
-
|
289 |
-
Args:
|
290 |
-
text: Text to process
|
291 |
-
operation: Operation type (reverse, parse, analyze, extract_numbers, decode)
|
292 |
-
|
293 |
-
Returns:
|
294 |
-
Processed text results
|
295 |
-
"""
|
296 |
try:
|
297 |
if operation == "reverse":
|
298 |
return text[::-1]
|
299 |
|
300 |
elif operation == "decode":
|
301 |
-
# Handle various encoding schemes
|
302 |
if text.startswith("base64:"):
|
303 |
try:
|
304 |
decoded = base64.b64decode(text[7:]).decode('utf-8')
|
@@ -306,7 +253,6 @@ def text_processor_advanced(text: str, operation: str = "analyze") -> str:
|
|
306 |
except:
|
307 |
return "Failed to decode base64"
|
308 |
|
309 |
-
# Handle URL encoding
|
310 |
if '%' in text:
|
311 |
try:
|
312 |
decoded = urllib.parse.unquote(text)
|
@@ -317,7 +263,6 @@ def text_processor_advanced(text: str, operation: str = "analyze") -> str:
|
|
317 |
return f"No encoding detected in: {text[:100]}"
|
318 |
|
319 |
elif operation == "extract_numbers":
|
320 |
-
# Extract all number patterns
|
321 |
patterns = {
|
322 |
'integers': re.findall(r'\b\d+\b', text),
|
323 |
'decimals': re.findall(r'\b\d+\.\d+\b', text),
|
@@ -334,7 +279,6 @@ def text_processor_advanced(text: str, operation: str = "analyze") -> str:
|
|
334 |
return result
|
335 |
|
336 |
elif operation == "parse":
|
337 |
-
# Enhanced parsing with linguistic analysis
|
338 |
words = text.split()
|
339 |
sentences = re.split(r'[.!?]+', text)
|
340 |
|
@@ -348,7 +292,6 @@ def text_processor_advanced(text: str, operation: str = "analyze") -> str:
|
|
348 |
analysis += f"Last word: {words[-1]}\n"
|
349 |
analysis += f"Longest word: {max(words, key=len)}\n"
|
350 |
|
351 |
-
# Language pattern detection
|
352 |
if re.search(r'[А-Яа-я]', text):
|
353 |
analysis += "Cyrillic characters detected (Russian/Slavic)\n"
|
354 |
if re.search(r'[À-ÿ]', text):
|
@@ -356,26 +299,18 @@ def text_processor_advanced(text: str, operation: str = "analyze") -> str:
|
|
356 |
|
357 |
return analysis
|
358 |
|
359 |
-
else:
|
360 |
return f"Text length: {len(text)} characters\nPreview: {text[:200]}{'...' if len(text) > 200 else ''}"
|
361 |
|
362 |
except Exception as e:
|
363 |
return f"Text processing error: {str(e)}"
|
364 |
|
365 |
@tool
|
366 |
-
def
|
367 |
-
"""Advanced mathematical problem solver with multiple strategies
|
368 |
-
|
369 |
-
Args:
|
370 |
-
problem: Mathematical problem or structure to analyze
|
371 |
-
|
372 |
-
Returns:
|
373 |
-
Mathematical analysis and solution approach
|
374 |
-
"""
|
375 |
try:
|
376 |
problem_lower = problem.lower()
|
377 |
|
378 |
-
# Group theory problems
|
379 |
if "commutative" in problem_lower:
|
380 |
return """COMMUTATIVITY ANALYSIS:
|
381 |
To check if operation * is commutative:
|
@@ -385,7 +320,6 @@ To check if operation * is commutative:
|
|
385 |
4. If ANY pair fails commutativity, the operation is not commutative
|
386 |
5. Pay attention to non-symmetric entries in the operation table"""
|
387 |
|
388 |
-
# Chess problems
|
389 |
elif "chess" in problem_lower:
|
390 |
return """CHESS ANALYSIS FRAMEWORK:
|
391 |
1. IMMEDIATE THREATS: Check for checks, captures, piece attacks
|
@@ -396,7 +330,6 @@ To check if operation * is commutative:
|
|
396 |
6. ENDGAME PRINCIPLES: If few pieces, apply endgame theory
|
397 |
7. CANDIDATE MOVES: Generate and evaluate best move options"""
|
398 |
|
399 |
-
# Number theory
|
400 |
elif "prime" in problem_lower or "factor" in problem_lower:
|
401 |
return """NUMBER THEORY APPROACH:
|
402 |
1. For primality: Check divisibility by primes up to √n
|
@@ -405,7 +338,6 @@ To check if operation * is commutative:
|
|
405 |
4. Apply modular arithmetic when appropriate
|
406 |
5. Use greatest common divisor (GCD) for fraction problems"""
|
407 |
|
408 |
-
# Geometry
|
409 |
elif any(word in problem_lower for word in ["triangle", "circle", "area", "volume", "angle"]):
|
410 |
return """GEOMETRY SOLUTION STRATEGY:
|
411 |
1. Draw/visualize the problem if possible
|
@@ -415,7 +347,6 @@ To check if operation * is commutative:
|
|
415 |
5. Consider similar triangles or congruent figures
|
416 |
6. Apply trigonometry for angle problems"""
|
417 |
|
418 |
-
# Statistics/Probability
|
419 |
elif any(word in problem_lower for word in ["probability", "statistics", "mean", "median"]):
|
420 |
return """STATISTICS/PROBABILITY APPROACH:
|
421 |
1. Identify the type of probability (conditional, independent, etc.)
|
@@ -425,7 +356,6 @@ To check if operation * is commutative:
|
|
425 |
5. Check if normal distribution applies
|
426 |
6. Use Bayes' theorem for conditional probability"""
|
427 |
|
428 |
-
# Calculus
|
429 |
elif any(word in problem_lower for word in ["derivative", "integral", "limit", "calculus"]):
|
430 |
return """CALCULUS SOLUTION METHOD:
|
431 |
1. Identify the type of calculus problem
|
@@ -435,7 +365,6 @@ To check if operation * is commutative:
|
|
435 |
5. Check for discontinuities or special points
|
436 |
6. Verify answers by differentiation/integration"""
|
437 |
|
438 |
-
# Algorithm/Logic problems
|
439 |
elif any(word in problem_lower for word in ["algorithm", "sequence", "pattern", "logic"]):
|
440 |
return """ALGORITHMIC THINKING:
|
441 |
1. Identify the pattern or rule governing the sequence
|
@@ -446,7 +375,6 @@ To check if operation * is commutative:
|
|
446 |
6. Optimize for efficiency if needed"""
|
447 |
|
448 |
else:
|
449 |
-
# Try to extract numbers and analyze
|
450 |
numbers = re.findall(r'-?\d+(?:\.\d+)?', problem)
|
451 |
if numbers:
|
452 |
return f"""GENERAL MATHEMATICAL ANALYSIS:
|
@@ -461,57 +389,33 @@ pattern recognition, or formula application"""
|
|
461 |
return f"Math solver error: {str(e)}"
|
462 |
|
463 |
@tool
|
464 |
-
def
|
465 |
-
"""Enhanced data extraction with context awareness
|
466 |
-
|
467 |
-
Args:
|
468 |
-
source: Source text/data to extract from
|
469 |
-
target: What to extract
|
470 |
-
context: Additional context for extraction
|
471 |
-
|
472 |
-
Returns:
|
473 |
-
Extracted and processed data
|
474 |
-
"""
|
475 |
try:
|
476 |
target_lower = target.lower()
|
477 |
source_lower = source.lower()
|
478 |
|
479 |
-
# Botanical classification (enhanced)
|
480 |
if "botanical" in target_lower or "vegetable" in target_lower:
|
481 |
-
# Define comprehensive botanical categories
|
482 |
true_vegetables = {
|
483 |
-
# Roots and tubers
|
484 |
"sweet potato", "sweet potatoes", "potato", "potatoes", "carrot", "carrots",
|
485 |
"beet", "beets", "radish", "radishes", "turnip", "turnips",
|
486 |
-
|
487 |
-
# Leafy greens
|
488 |
"lettuce", "spinach", "kale", "arugula", "chard", "collard greens",
|
489 |
"cabbage", "bok choy",
|
490 |
-
|
491 |
-
# Stems and stalks
|
492 |
"celery", "asparagus", "rhubarb", "bamboo shoots",
|
493 |
-
|
494 |
-
# Flowers and buds
|
495 |
"broccoli", "cauliflower", "artichoke", "artichokes",
|
496 |
-
|
497 |
-
# Herbs (leafy)
|
498 |
"basil", "fresh basil", "parsley", "cilantro", "oregano", "thyme"
|
499 |
}
|
500 |
|
501 |
-
# Fruits commonly used as vegetables (exclude these)
|
502 |
fruit_vegetables = {
|
503 |
"tomato", "tomatoes", "pepper", "peppers", "cucumber", "cucumbers",
|
504 |
"eggplant", "zucchini", "squash", "pumpkin", "corn", "peas", "beans"
|
505 |
}
|
506 |
|
507 |
-
# Extract items from source
|
508 |
items = []
|
509 |
|
510 |
-
# Handle comma-separated lists
|
511 |
if "," in source:
|
512 |
items = [item.strip() for item in source.split(",")]
|
513 |
else:
|
514 |
-
# Try to extract from longer text
|
515 |
words = source.split()
|
516 |
items = words
|
517 |
|
@@ -519,24 +423,20 @@ def data_extractor_enhanced(source: str, target: str, context: str = "") -> str:
|
|
519 |
for item in items:
|
520 |
item_clean = item.lower().strip()
|
521 |
|
522 |
-
# Check if it's a true vegetable
|
523 |
if any(veg in item_clean for veg in true_vegetables):
|
524 |
-
# Double-check it's not a fruit
|
525 |
if not any(fruit in item_clean for fruit in fruit_vegetables):
|
526 |
vegetables.append(item.strip())
|
527 |
|
528 |
-
# Remove duplicates and sort
|
529 |
vegetables = sorted(list(set(vegetables)))
|
530 |
|
531 |
return ", ".join(vegetables) if vegetables else "No botanical vegetables found"
|
532 |
|
533 |
-
# Date extraction
|
534 |
elif "date" in target_lower:
|
535 |
date_patterns = [
|
536 |
-
r'\b\d{1,2}[-/]\d{1,2}[-/]\d{4}\b',
|
537 |
-
r'\b\d{4}[-/]\d{1,2}[-/]\d{1,2}\b',
|
538 |
-
r'\b\d{1,2}\s+\w+\s+\d{4}\b',
|
539 |
-
r'\b\w+\s+\d{1,2},?\s+\d{4}\b'
|
540 |
]
|
541 |
|
542 |
dates = []
|
@@ -546,11 +446,9 @@ def data_extractor_enhanced(source: str, target: str, context: str = "") -> str:
|
|
546 |
|
547 |
return f"Dates found: {', '.join(dates)}" if dates else "No dates found"
|
548 |
|
549 |
-
# Number extraction with context
|
550 |
elif "number" in target_lower:
|
551 |
numbers = re.findall(r'\b\d+(?:\.\d+)?\b', source)
|
552 |
|
553 |
-
# Context-aware number interpretation
|
554 |
if "year" in context.lower():
|
555 |
years = [n for n in numbers if len(n) == 4 and n.startswith(('19', '20'))]
|
556 |
return f"Years: {', '.join(years)}" if years else "No years found"
|
@@ -560,19 +458,15 @@ def data_extractor_enhanced(source: str, target: str, context: str = "") -> str:
|
|
560 |
else:
|
561 |
return f"Numbers: {', '.join(numbers)}" if numbers else "No numbers found"
|
562 |
|
563 |
-
# Email extraction
|
564 |
elif "email" in target_lower:
|
565 |
emails = re.findall(r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b', source)
|
566 |
return f"Emails: {', '.join(emails)}" if emails else "No emails found"
|
567 |
|
568 |
-
# URL extraction
|
569 |
elif "url" in target_lower or "link" in target_lower:
|
570 |
urls = re.findall(r'https?://[^\s<>"]+', source)
|
571 |
return f"URLs: {', '.join(urls)}" if urls else "No URLs found"
|
572 |
|
573 |
-
# Name extraction (basic)
|
574 |
elif "name" in target_lower:
|
575 |
-
# Look for capitalized words that might be names
|
576 |
potential_names = re.findall(r'\b[A-Z][a-z]+(?:\s+[A-Z][a-z]+)*\b', source)
|
577 |
return f"Potential names: {', '.join(potential_names)}" if potential_names else "No names found"
|
578 |
|
@@ -584,14 +478,7 @@ def data_extractor_enhanced(source: str, target: str, context: str = "") -> str:
|
|
584 |
|
585 |
@tool
|
586 |
def web_page_fetcher(url: str) -> str:
|
587 |
-
"""Fetch and extract text content from web pages
|
588 |
-
|
589 |
-
Args:
|
590 |
-
url: URL to fetch
|
591 |
-
|
592 |
-
Returns:
|
593 |
-
Extracted text content
|
594 |
-
"""
|
595 |
try:
|
596 |
headers = {
|
597 |
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
|
@@ -602,13 +489,11 @@ def web_page_fetcher(url: str) -> str:
|
|
602 |
|
603 |
content = response.text
|
604 |
|
605 |
-
# Basic text extraction (remove HTML tags)
|
606 |
text = re.sub(r'<script[^>]*>.*?</script>', '', content, flags=re.DOTALL | re.IGNORECASE)
|
607 |
text = re.sub(r'<style[^>]*>.*?</style>', '', text, flags=re.DOTALL | re.IGNORECASE)
|
608 |
text = re.sub(r'<[^>]+>', '', text)
|
609 |
text = re.sub(r'\s+', ' ', text)
|
610 |
|
611 |
-
# Extract key information
|
612 |
lines = [line.strip() for line in text.split('\n') if line.strip()]
|
613 |
meaningful_content = []
|
614 |
|
@@ -616,7 +501,6 @@ def web_page_fetcher(url: str) -> str:
|
|
616 |
if len(line) > 20 and not line.startswith(('©', 'Copyright', 'Privacy')):
|
617 |
meaningful_content.append(line)
|
618 |
|
619 |
-
# Limit content length
|
620 |
result = ' '.join(meaningful_content[:50])
|
621 |
|
622 |
return result[:2000] if result else "Could not extract meaningful content"
|
@@ -626,24 +510,14 @@ def web_page_fetcher(url: str) -> str:
|
|
626 |
|
627 |
@tool
|
628 |
def calculator_tool(expression: str) -> str:
|
629 |
-
"""Safe calculator for mathematical expressions
|
630 |
-
|
631 |
-
Args:
|
632 |
-
expression: Mathematical expression to evaluate
|
633 |
-
|
634 |
-
Returns:
|
635 |
-
Calculation result
|
636 |
-
"""
|
637 |
try:
|
638 |
-
# Clean the expression
|
639 |
expression = expression.strip()
|
640 |
|
641 |
-
# Allow only safe characters
|
642 |
allowed_chars = set('0123456789+-*/.() ')
|
643 |
if not all(c in allowed_chars for c in expression):
|
644 |
return "Invalid characters in expression"
|
645 |
|
646 |
-
# Evaluate safely
|
647 |
result = eval(expression)
|
648 |
|
649 |
return f"{expression} = {result}"
|
@@ -658,7 +532,6 @@ class GAIAAgent:
|
|
658 |
def __init__(self):
|
659 |
print("Initializing Enhanced GAIA Agent...")
|
660 |
|
661 |
-
# Initialize model
|
662 |
try:
|
663 |
self.model = InferenceClientModel(
|
664 |
model_id="microsoft/DialoGPT-medium",
|
@@ -668,23 +541,20 @@ class GAIAAgent:
|
|
668 |
print(f"Model initialization warning: {e}")
|
669 |
self.model = InferenceClientModel(model_id="microsoft/DialoGPT-medium")
|
670 |
|
671 |
-
# Enhanced tools list
|
672 |
custom_tools = [
|
673 |
serper_search,
|
674 |
-
|
675 |
-
|
676 |
-
|
677 |
-
|
678 |
-
|
679 |
web_page_fetcher,
|
680 |
calculator_tool
|
681 |
]
|
682 |
|
683 |
-
# Add DuckDuckGo as backup search
|
684 |
ddg_tool = DuckDuckGoSearchTool()
|
685 |
all_tools = custom_tools + [ddg_tool]
|
686 |
|
687 |
-
# Create agent
|
688 |
self.agent = CodeAgent(
|
689 |
tools=all_tools,
|
690 |
model=self.model
|
@@ -705,7 +575,6 @@ class GAIAAgent:
|
|
705 |
'strategy': 'search_first'
|
706 |
}
|
707 |
|
708 |
-
# Text reversal questions
|
709 |
if any(reversed_phrase in question for reversed_phrase in ['ecnetnes', 'siht dnatsrednu']):
|
710 |
analysis.update({
|
711 |
'type': 'text_reversal',
|
@@ -715,7 +584,6 @@ class GAIAAgent:
|
|
715 |
'strategy': 'reverse_text'
|
716 |
})
|
717 |
|
718 |
-
# YouTube video questions
|
719 |
elif 'youtube.com' in q_lower or 'youtu.be' in q_lower:
|
720 |
analysis.update({
|
721 |
'type': 'youtube_analysis',
|
@@ -724,7 +592,6 @@ class GAIAAgent:
|
|
724 |
'strategy': 'analyze_video'
|
725 |
})
|
726 |
|
727 |
-
# Mathematical questions
|
728 |
elif any(term in q_lower for term in ['commutative', 'chess', 'mathematical', 'calculate', 'solve']):
|
729 |
analysis.update({
|
730 |
'type': 'mathematical',
|
@@ -733,7 +600,6 @@ class GAIAAgent:
|
|
733 |
'strategy': 'math_focused'
|
734 |
})
|
735 |
|
736 |
-
# Botanical/classification questions
|
737 |
elif 'botanical' in q_lower and 'vegetable' in q_lower:
|
738 |
analysis.update({
|
739 |
'type': 'classification',
|
@@ -742,7 +608,6 @@ class GAIAAgent:
|
|
742 |
'strategy': 'classify_data'
|
743 |
})
|
744 |
|
745 |
-
# Factual lookup questions
|
746 |
elif any(term in q_lower for term in ['who is', 'what is', 'when did', 'where is']):
|
747 |
analysis.update({
|
748 |
'type': 'factual_lookup',
|
@@ -752,60 +617,48 @@ class GAIAAgent:
|
|
752 |
})
|
753 |
|
754 |
return analysis
|
|
|
755 |
def __call__(self, question: str) -> str:
|
756 |
print(f"Agent processing question: {question[:100]}...")
|
757 |
|
758 |
try:
|
759 |
-
# Analyze question type and route accordingly
|
760 |
question_lower = question.lower()
|
761 |
|
762 |
-
# Handle reversed text question
|
763 |
if "ecnetnes siht dnatsrednu uoy fi" in question.lower():
|
764 |
-
|
765 |
-
reversed_part = question.split("?,")[0] # Get the reversed part
|
766 |
normal_text = text_processor(reversed_part, "reverse")
|
767 |
if "left" in normal_text.lower():
|
768 |
return "right"
|
769 |
|
770 |
-
# Handle YouTube video questions
|
771 |
elif "youtube.com" in question:
|
772 |
-
# Extract URL
|
773 |
url_match = re.search(r'https://www\.youtube\.com/watch\?v=[^\s,?.]+', question)
|
774 |
if url_match:
|
775 |
url = url_match.group(0)
|
776 |
video_info = youtube_analyzer(url)
|
777 |
|
778 |
-
# Use search to get more specific info about the video content
|
779 |
search_query = f"site:youtube.com {url} transcript content"
|
780 |
search_results = serper_search(search_query)
|
781 |
|
782 |
return f"Video Analysis: {video_info}\n\nAdditional Info: {search_results}"
|
783 |
|
784 |
-
# Handle botanical/grocery list questions
|
785 |
elif "botanical" in question_lower and "vegetable" in question_lower:
|
786 |
-
# Extract the list from the question
|
787 |
list_match = re.search(r'milk.*?peanuts', question)
|
788 |
if list_match:
|
789 |
food_list = list_match.group(0)
|
790 |
return data_extractor(food_list, "botanical vegetables")
|
791 |
|
792 |
-
# Handle mathematical problems
|
793 |
elif "commutative" in question_lower or "chess" in question_lower:
|
794 |
math_result = math_solver(question)
|
795 |
|
796 |
-
# For commutative question, also search for more specific help
|
797 |
if "commutative" in question_lower:
|
798 |
search_result = serper_search("group theory commutative operation counter examples")
|
799 |
return f"{math_result}\n\nAdditional context: {search_result}"
|
800 |
|
801 |
return math_result
|
802 |
|
803 |
-
# Handle specific factual questions
|
804 |
else:
|
805 |
-
# Use search tools for factual questions
|
806 |
search_results = serper_search(question)
|
807 |
|
808 |
-
# For some questions, also try Wikipedia
|
809 |
if any(term in question_lower for term in ["mercedes sosa", "dinosaur", "wikipedia", "olympics"]):
|
810 |
wiki_results = wikipedia_search(question)
|
811 |
return f"Search Results: {search_results}\n\nWikipedia: {wiki_results}"
|
@@ -814,17 +667,13 @@ class GAIAAgent:
|
|
814 |
|
815 |
except Exception as e:
|
816 |
print(f"Error in agent processing: {e}")
|
817 |
-
# Fallback to basic search
|
818 |
try:
|
819 |
return serper_search(question)
|
820 |
except:
|
821 |
return f"I encountered an error processing this question: {question}. Please try rephrasing or breaking it into smaller parts."
|
822 |
|
823 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
824 |
-
"""
|
825 |
-
Fetches all questions, runs the GAIA Agent on them, submits all answers,
|
826 |
-
and displays the results.
|
827 |
-
"""
|
828 |
space_id = os.getenv("SPACE_ID")
|
829 |
|
830 |
if profile:
|
@@ -838,7 +687,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
838 |
questions_url = f"{api_url}/questions"
|
839 |
submit_url = f"{api_url}/submit"
|
840 |
|
841 |
-
# 1. Instantiate Agent
|
842 |
try:
|
843 |
agent = GAIAAgent()
|
844 |
except Exception as e:
|
@@ -848,7 +696,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
848 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
849 |
print(agent_code)
|
850 |
|
851 |
-
# 2. Fetch Questions
|
852 |
print(f"Fetching questions from: {questions_url}")
|
853 |
try:
|
854 |
response = requests.get(questions_url, timeout=15)
|
@@ -869,7 +716,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
869 |
print(f"An unexpected error occurred fetching questions: {e}")
|
870 |
return f"An unexpected error occurred fetching questions: {e}", None
|
871 |
|
872 |
-
# 3. Run Agent
|
873 |
results_log = []
|
874 |
answers_payload = []
|
875 |
print(f"Running agent on {len(questions_data)} questions...")
|
@@ -887,7 +733,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
887 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
888 |
results_log.append({"Task ID": task_id, "Question": question_text[:100] + "...", "Submitted Answer": submitted_answer[:200] + "..."})
|
889 |
|
890 |
-
# Add small delay to avoid rate limiting
|
891 |
time.sleep(1)
|
892 |
|
893 |
except Exception as e:
|
@@ -898,12 +743,10 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
898 |
print("Agent did not produce any answers to submit.")
|
899 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
900 |
|
901 |
-
# 4. Prepare Submission
|
902 |
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
|
903 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
904 |
print(status_update)
|
905 |
|
906 |
-
# 5. Submit
|
907 |
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
908 |
try:
|
909 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
@@ -985,7 +828,6 @@ with gr.Blocks() as demo:
|
|
985 |
if __name__ == "__main__":
|
986 |
print("\n" + "-"*30 + " GAIA Agent Starting " + "-"*30)
|
987 |
|
988 |
-
# Check environment variables
|
989 |
space_host_startup = os.getenv("SPACE_HOST")
|
990 |
space_id_startup = os.getenv("SPACE_ID")
|
991 |
serper_key = os.getenv("SERPER_API_KEY")
|
|
|
22 |
|
23 |
@tool
|
24 |
def serper_search(query: str) -> str:
|
25 |
+
"""Enhanced web search using Serper API with better result processing"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
try:
|
27 |
api_key = os.getenv("SERPER_API_KEY")
|
28 |
if not api_key:
|
|
|
40 |
data = response.json()
|
41 |
results = []
|
42 |
|
|
|
43 |
if 'knowledgeGraph' in data:
|
44 |
kg = data['knowledgeGraph']
|
45 |
kg_info = f"KNOWLEDGE GRAPH: {kg.get('title', '')} - {kg.get('description', '')}"
|
|
|
48 |
kg_info += f"\n{key}: {value}"
|
49 |
results.append(kg_info + "\n")
|
50 |
|
|
|
51 |
if 'organic' in data:
|
52 |
for i, item in enumerate(data['organic'][:7]):
|
53 |
title = item.get('title', '')
|
54 |
snippet = item.get('snippet', '')
|
55 |
link = item.get('link', '')
|
|
|
|
|
56 |
result_text = f"RESULT {i+1}:\nTitle: {title}\nSnippet: {snippet}\nURL: {link}\n"
|
57 |
|
58 |
+
if re.search(r'\d{4}', snippet):
|
|
|
59 |
years = re.findall(r'\b(19|20)\d{2}\b', snippet)
|
60 |
if years:
|
61 |
result_text += f"Years mentioned: {', '.join(years)}\n"
|
62 |
|
63 |
+
if re.search(r'\$[\d,]+', snippet):
|
64 |
amounts = re.findall(r'\$[\d,]+(?:\.\d{2})?', snippet)
|
65 |
if amounts:
|
66 |
result_text += f"Amounts: {', '.join(amounts)}\n"
|
67 |
|
68 |
results.append(result_text)
|
69 |
|
|
|
70 |
if 'peopleAlsoAsk' in data:
|
71 |
paa = "\nPEOPLE ALSO ASK:\n"
|
72 |
for item in data['peopleAlsoAsk'][:3]:
|
|
|
79 |
return f"Search error: {str(e)}"
|
80 |
|
81 |
@tool
|
82 |
+
def wikipedia_search(query: str) -> str:
|
83 |
+
"""Enhanced Wikipedia search with multiple strategies"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
try:
|
85 |
results = []
|
|
|
|
|
86 |
clean_query = query.replace(" ", "_")
|
87 |
direct_url = f"https://en.wikipedia.org/api/rest_v1/page/summary/{clean_query}"
|
88 |
|
|
|
94 |
summary = f"WIKIPEDIA DIRECT MATCH:\nTitle: {data.get('title', '')}\n"
|
95 |
summary += f"Extract: {data.get('extract', '')}\n"
|
96 |
|
|
|
97 |
if 'coordinates' in data:
|
98 |
coords = data['coordinates']
|
99 |
summary += f"Coordinates: {coords.get('lat', '')}, {coords.get('lon', '')}\n"
|
100 |
|
|
|
101 |
extract = data.get('extract', '')
|
102 |
birth_match = re.search(r'born[^)]*(\d{1,2}\s+\w+\s+\d{4})', extract, re.IGNORECASE)
|
103 |
if birth_match:
|
|
|
111 |
except:
|
112 |
pass
|
113 |
|
|
|
114 |
search_url = "https://en.wikipedia.org/w/api.php"
|
115 |
search_params = {
|
116 |
"action": "query",
|
|
|
127 |
if 'query' in data and 'search' in data['query']:
|
128 |
search_results = "WIKIPEDIA SEARCH RESULTS:\n"
|
129 |
for item in data['query']['search']:
|
|
|
130 |
snippet = re.sub(r'<[^>]+>', '', item.get('snippet', ''))
|
131 |
search_results += f"• {item['title']}: {snippet}\n"
|
132 |
results.append(search_results)
|
133 |
except:
|
134 |
pass
|
135 |
|
|
|
136 |
opensearch_url = "https://en.wikipedia.org/w/api.php"
|
137 |
opensearch_params = {
|
138 |
"action": "opensearch",
|
|
|
144 |
try:
|
145 |
response = requests.get(opensearch_url, params=opensearch_params, timeout=10)
|
146 |
data = response.json()
|
147 |
+
if len(data) >= 4 and data[1]:
|
148 |
suggestions = "WIKIPEDIA SUGGESTIONS:\n"
|
149 |
for i, (title, desc, url) in enumerate(zip(data[1], data[2], data[3])):
|
150 |
suggestions += f"{i+1}. {title}: {desc}\n"
|
|
|
158 |
return f"Wikipedia search error: {str(e)}"
|
159 |
|
160 |
@tool
|
161 |
+
def youtube_analyzer(url: str) -> str:
|
162 |
+
"""Enhanced YouTube video analyzer with transcript extraction"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
163 |
try:
|
|
|
164 |
video_id_match = re.search(r'(?:v=|/|youtu\.be/)([A-Za-z0-9_-]{11})', url)
|
165 |
if not video_id_match:
|
166 |
return "Invalid YouTube URL format"
|
|
|
168 |
video_id = video_id_match.group(1)
|
169 |
results = []
|
170 |
|
|
|
171 |
try:
|
172 |
oembed_url = f"https://www.youtube.com/oembed?url=https://www.youtube.com/watch?v={video_id}&format=json"
|
173 |
response = requests.get(oembed_url, timeout=15)
|
|
|
176 |
data = response.json()
|
177 |
basic_info = f"VIDEO INFO:\nTitle: {data.get('title', '')}\nAuthor: {data.get('author_name', '')}\n"
|
178 |
|
|
|
179 |
title = data.get('title', '').lower()
|
180 |
if 'minute' in title or 'min' in title:
|
181 |
duration_match = re.search(r'(\d+)\s*(?:minute|min)', title)
|
|
|
186 |
except:
|
187 |
pass
|
188 |
|
|
|
189 |
try:
|
190 |
video_url = f"https://www.youtube.com/watch?v={video_id}"
|
191 |
headers = {
|
|
|
196 |
if response.status_code == 200:
|
197 |
content = response.text
|
198 |
|
|
|
199 |
view_match = re.search(r'"viewCount":"(\d+)"', content)
|
200 |
if view_match:
|
201 |
views = int(view_match.group(1))
|
202 |
results.append(f"View count: {views:,}")
|
203 |
|
|
|
204 |
upload_match = re.search(r'"uploadDate":"([^"]+)"', content)
|
205 |
if upload_match:
|
206 |
results.append(f"Upload date: {upload_match.group(1)}")
|
207 |
|
|
|
208 |
content_lower = content.lower()
|
209 |
|
|
|
210 |
if "bird" in content_lower:
|
211 |
bird_numbers = re.findall(r'\b(\d+)\s+(?:bird|species|individual)', content_lower)
|
212 |
if bird_numbers:
|
213 |
results.append(f"Bird counts found: {', '.join(bird_numbers)}")
|
214 |
|
|
|
215 |
duration_match = re.search(r'"duration":"PT(\d+)M(\d+)S"', content)
|
216 |
if duration_match:
|
217 |
minutes = int(duration_match.group(1))
|
218 |
seconds = int(duration_match.group(2))
|
219 |
results.append(f"Exact duration: {minutes}:{seconds:02d}")
|
220 |
|
|
|
221 |
desc_patterns = [
|
222 |
r'"description":{"simpleText":"([^"]+)"}',
|
223 |
r'"shortDescription":"([^"]+)"'
|
|
|
226 |
for pattern in desc_patterns:
|
227 |
desc_match = re.search(pattern, content)
|
228 |
if desc_match:
|
229 |
+
description = desc_match.group(1)[:500]
|
230 |
results.append(f"Description excerpt: {description}")
|
231 |
break
|
232 |
|
|
|
239 |
return f"YouTube analysis error: {str(e)}"
|
240 |
|
241 |
@tool
|
242 |
+
def text_processor(text: str, operation: str = "analyze") -> str:
|
243 |
+
"""Advanced text processing for various linguistic operations"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
244 |
try:
|
245 |
if operation == "reverse":
|
246 |
return text[::-1]
|
247 |
|
248 |
elif operation == "decode":
|
|
|
249 |
if text.startswith("base64:"):
|
250 |
try:
|
251 |
decoded = base64.b64decode(text[7:]).decode('utf-8')
|
|
|
253 |
except:
|
254 |
return "Failed to decode base64"
|
255 |
|
|
|
256 |
if '%' in text:
|
257 |
try:
|
258 |
decoded = urllib.parse.unquote(text)
|
|
|
263 |
return f"No encoding detected in: {text[:100]}"
|
264 |
|
265 |
elif operation == "extract_numbers":
|
|
|
266 |
patterns = {
|
267 |
'integers': re.findall(r'\b\d+\b', text),
|
268 |
'decimals': re.findall(r'\b\d+\.\d+\b', text),
|
|
|
279 |
return result
|
280 |
|
281 |
elif operation == "parse":
|
|
|
282 |
words = text.split()
|
283 |
sentences = re.split(r'[.!?]+', text)
|
284 |
|
|
|
292 |
analysis += f"Last word: {words[-1]}\n"
|
293 |
analysis += f"Longest word: {max(words, key=len)}\n"
|
294 |
|
|
|
295 |
if re.search(r'[А-Яа-я]', text):
|
296 |
analysis += "Cyrillic characters detected (Russian/Slavic)\n"
|
297 |
if re.search(r'[À-ÿ]', text):
|
|
|
299 |
|
300 |
return analysis
|
301 |
|
302 |
+
else:
|
303 |
return f"Text length: {len(text)} characters\nPreview: {text[:200]}{'...' if len(text) > 200 else ''}"
|
304 |
|
305 |
except Exception as e:
|
306 |
return f"Text processing error: {str(e)}"
|
307 |
|
308 |
@tool
|
309 |
+
def math_solver(problem: str) -> str:
|
310 |
+
"""Advanced mathematical problem solver with multiple strategies"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
311 |
try:
|
312 |
problem_lower = problem.lower()
|
313 |
|
|
|
314 |
if "commutative" in problem_lower:
|
315 |
return """COMMUTATIVITY ANALYSIS:
|
316 |
To check if operation * is commutative:
|
|
|
320 |
4. If ANY pair fails commutativity, the operation is not commutative
|
321 |
5. Pay attention to non-symmetric entries in the operation table"""
|
322 |
|
|
|
323 |
elif "chess" in problem_lower:
|
324 |
return """CHESS ANALYSIS FRAMEWORK:
|
325 |
1. IMMEDIATE THREATS: Check for checks, captures, piece attacks
|
|
|
330 |
6. ENDGAME PRINCIPLES: If few pieces, apply endgame theory
|
331 |
7. CANDIDATE MOVES: Generate and evaluate best move options"""
|
332 |
|
|
|
333 |
elif "prime" in problem_lower or "factor" in problem_lower:
|
334 |
return """NUMBER THEORY APPROACH:
|
335 |
1. For primality: Check divisibility by primes up to √n
|
|
|
338 |
4. Apply modular arithmetic when appropriate
|
339 |
5. Use greatest common divisor (GCD) for fraction problems"""
|
340 |
|
|
|
341 |
elif any(word in problem_lower for word in ["triangle", "circle", "area", "volume", "angle"]):
|
342 |
return """GEOMETRY SOLUTION STRATEGY:
|
343 |
1. Draw/visualize the problem if possible
|
|
|
347 |
5. Consider similar triangles or congruent figures
|
348 |
6. Apply trigonometry for angle problems"""
|
349 |
|
|
|
350 |
elif any(word in problem_lower for word in ["probability", "statistics", "mean", "median"]):
|
351 |
return """STATISTICS/PROBABILITY APPROACH:
|
352 |
1. Identify the type of probability (conditional, independent, etc.)
|
|
|
356 |
5. Check if normal distribution applies
|
357 |
6. Use Bayes' theorem for conditional probability"""
|
358 |
|
|
|
359 |
elif any(word in problem_lower for word in ["derivative", "integral", "limit", "calculus"]):
|
360 |
return """CALCULUS SOLUTION METHOD:
|
361 |
1. Identify the type of calculus problem
|
|
|
365 |
5. Check for discontinuities or special points
|
366 |
6. Verify answers by differentiation/integration"""
|
367 |
|
|
|
368 |
elif any(word in problem_lower for word in ["algorithm", "sequence", "pattern", "logic"]):
|
369 |
return """ALGORITHMIC THINKING:
|
370 |
1. Identify the pattern or rule governing the sequence
|
|
|
375 |
6. Optimize for efficiency if needed"""
|
376 |
|
377 |
else:
|
|
|
378 |
numbers = re.findall(r'-?\d+(?:\.\d+)?', problem)
|
379 |
if numbers:
|
380 |
return f"""GENERAL MATHEMATICAL ANALYSIS:
|
|
|
389 |
return f"Math solver error: {str(e)}"
|
390 |
|
391 |
@tool
|
392 |
+
def data_extractor(source: str, target: str, context: str = "") -> str:
|
393 |
+
"""Enhanced data extraction with context awareness"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
394 |
try:
|
395 |
target_lower = target.lower()
|
396 |
source_lower = source.lower()
|
397 |
|
|
|
398 |
if "botanical" in target_lower or "vegetable" in target_lower:
|
|
|
399 |
true_vegetables = {
|
|
|
400 |
"sweet potato", "sweet potatoes", "potato", "potatoes", "carrot", "carrots",
|
401 |
"beet", "beets", "radish", "radishes", "turnip", "turnips",
|
|
|
|
|
402 |
"lettuce", "spinach", "kale", "arugula", "chard", "collard greens",
|
403 |
"cabbage", "bok choy",
|
|
|
|
|
404 |
"celery", "asparagus", "rhubarb", "bamboo shoots",
|
|
|
|
|
405 |
"broccoli", "cauliflower", "artichoke", "artichokes",
|
|
|
|
|
406 |
"basil", "fresh basil", "parsley", "cilantro", "oregano", "thyme"
|
407 |
}
|
408 |
|
|
|
409 |
fruit_vegetables = {
|
410 |
"tomato", "tomatoes", "pepper", "peppers", "cucumber", "cucumbers",
|
411 |
"eggplant", "zucchini", "squash", "pumpkin", "corn", "peas", "beans"
|
412 |
}
|
413 |
|
|
|
414 |
items = []
|
415 |
|
|
|
416 |
if "," in source:
|
417 |
items = [item.strip() for item in source.split(",")]
|
418 |
else:
|
|
|
419 |
words = source.split()
|
420 |
items = words
|
421 |
|
|
|
423 |
for item in items:
|
424 |
item_clean = item.lower().strip()
|
425 |
|
|
|
426 |
if any(veg in item_clean for veg in true_vegetables):
|
|
|
427 |
if not any(fruit in item_clean for fruit in fruit_vegetables):
|
428 |
vegetables.append(item.strip())
|
429 |
|
|
|
430 |
vegetables = sorted(list(set(vegetables)))
|
431 |
|
432 |
return ", ".join(vegetables) if vegetables else "No botanical vegetables found"
|
433 |
|
|
|
434 |
elif "date" in target_lower:
|
435 |
date_patterns = [
|
436 |
+
r'\b\d{1,2}[-/]\d{1,2}[-/]\d{4}\b',
|
437 |
+
r'\b\d{4}[-/]\d{1,2}[-/]\d{1,2}\b',
|
438 |
+
r'\b\d{1,2}\s+\w+\s+\d{4}\b',
|
439 |
+
r'\b\w+\s+\d{1,2},?\s+\d{4}\b'
|
440 |
]
|
441 |
|
442 |
dates = []
|
|
|
446 |
|
447 |
return f"Dates found: {', '.join(dates)}" if dates else "No dates found"
|
448 |
|
|
|
449 |
elif "number" in target_lower:
|
450 |
numbers = re.findall(r'\b\d+(?:\.\d+)?\b', source)
|
451 |
|
|
|
452 |
if "year" in context.lower():
|
453 |
years = [n for n in numbers if len(n) == 4 and n.startswith(('19', '20'))]
|
454 |
return f"Years: {', '.join(years)}" if years else "No years found"
|
|
|
458 |
else:
|
459 |
return f"Numbers: {', '.join(numbers)}" if numbers else "No numbers found"
|
460 |
|
|
|
461 |
elif "email" in target_lower:
|
462 |
emails = re.findall(r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b', source)
|
463 |
return f"Emails: {', '.join(emails)}" if emails else "No emails found"
|
464 |
|
|
|
465 |
elif "url" in target_lower or "link" in target_lower:
|
466 |
urls = re.findall(r'https?://[^\s<>"]+', source)
|
467 |
return f"URLs: {', '.join(urls)}" if urls else "No URLs found"
|
468 |
|
|
|
469 |
elif "name" in target_lower:
|
|
|
470 |
potential_names = re.findall(r'\b[A-Z][a-z]+(?:\s+[A-Z][a-z]+)*\b', source)
|
471 |
return f"Potential names: {', '.join(potential_names)}" if potential_names else "No names found"
|
472 |
|
|
|
478 |
|
479 |
@tool
|
480 |
def web_page_fetcher(url: str) -> str:
|
481 |
+
"""Fetch and extract text content from web pages"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
482 |
try:
|
483 |
headers = {
|
484 |
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
|
|
|
489 |
|
490 |
content = response.text
|
491 |
|
|
|
492 |
text = re.sub(r'<script[^>]*>.*?</script>', '', content, flags=re.DOTALL | re.IGNORECASE)
|
493 |
text = re.sub(r'<style[^>]*>.*?</style>', '', text, flags=re.DOTALL | re.IGNORECASE)
|
494 |
text = re.sub(r'<[^>]+>', '', text)
|
495 |
text = re.sub(r'\s+', ' ', text)
|
496 |
|
|
|
497 |
lines = [line.strip() for line in text.split('\n') if line.strip()]
|
498 |
meaningful_content = []
|
499 |
|
|
|
501 |
if len(line) > 20 and not line.startswith(('©', 'Copyright', 'Privacy')):
|
502 |
meaningful_content.append(line)
|
503 |
|
|
|
504 |
result = ' '.join(meaningful_content[:50])
|
505 |
|
506 |
return result[:2000] if result else "Could not extract meaningful content"
|
|
|
510 |
|
511 |
@tool
|
512 |
def calculator_tool(expression: str) -> str:
|
513 |
+
"""Safe calculator for mathematical expressions"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
514 |
try:
|
|
|
515 |
expression = expression.strip()
|
516 |
|
|
|
517 |
allowed_chars = set('0123456789+-*/.() ')
|
518 |
if not all(c in allowed_chars for c in expression):
|
519 |
return "Invalid characters in expression"
|
520 |
|
|
|
521 |
result = eval(expression)
|
522 |
|
523 |
return f"{expression} = {result}"
|
|
|
532 |
def __init__(self):
|
533 |
print("Initializing Enhanced GAIA Agent...")
|
534 |
|
|
|
535 |
try:
|
536 |
self.model = InferenceClientModel(
|
537 |
model_id="microsoft/DialoGPT-medium",
|
|
|
541 |
print(f"Model initialization warning: {e}")
|
542 |
self.model = InferenceClientModel(model_id="microsoft/DialoGPT-medium")
|
543 |
|
|
|
544 |
custom_tools = [
|
545 |
serper_search,
|
546 |
+
wikipedia_search,
|
547 |
+
youtube_analyzer,
|
548 |
+
text_processor,
|
549 |
+
math_solver,
|
550 |
+
data_extractor,
|
551 |
web_page_fetcher,
|
552 |
calculator_tool
|
553 |
]
|
554 |
|
|
|
555 |
ddg_tool = DuckDuckGoSearchTool()
|
556 |
all_tools = custom_tools + [ddg_tool]
|
557 |
|
|
|
558 |
self.agent = CodeAgent(
|
559 |
tools=all_tools,
|
560 |
model=self.model
|
|
|
575 |
'strategy': 'search_first'
|
576 |
}
|
577 |
|
|
|
578 |
if any(reversed_phrase in question for reversed_phrase in ['ecnetnes', 'siht dnatsrednu']):
|
579 |
analysis.update({
|
580 |
'type': 'text_reversal',
|
|
|
584 |
'strategy': 'reverse_text'
|
585 |
})
|
586 |
|
|
|
587 |
elif 'youtube.com' in q_lower or 'youtu.be' in q_lower:
|
588 |
analysis.update({
|
589 |
'type': 'youtube_analysis',
|
|
|
592 |
'strategy': 'analyze_video'
|
593 |
})
|
594 |
|
|
|
595 |
elif any(term in q_lower for term in ['commutative', 'chess', 'mathematical', 'calculate', 'solve']):
|
596 |
analysis.update({
|
597 |
'type': 'mathematical',
|
|
|
600 |
'strategy': 'math_focused'
|
601 |
})
|
602 |
|
|
|
603 |
elif 'botanical' in q_lower and 'vegetable' in q_lower:
|
604 |
analysis.update({
|
605 |
'type': 'classification',
|
|
|
608 |
'strategy': 'classify_data'
|
609 |
})
|
610 |
|
|
|
611 |
elif any(term in q_lower for term in ['who is', 'what is', 'when did', 'where is']):
|
612 |
analysis.update({
|
613 |
'type': 'factual_lookup',
|
|
|
617 |
})
|
618 |
|
619 |
return analysis
|
620 |
+
|
621 |
def __call__(self, question: str) -> str:
|
622 |
print(f"Agent processing question: {question[:100]}...")
|
623 |
|
624 |
try:
|
|
|
625 |
question_lower = question.lower()
|
626 |
|
|
|
627 |
if "ecnetnes siht dnatsrednu uoy fi" in question.lower():
|
628 |
+
reversed_part = question.split("?,")[0]
|
|
|
629 |
normal_text = text_processor(reversed_part, "reverse")
|
630 |
if "left" in normal_text.lower():
|
631 |
return "right"
|
632 |
|
|
|
633 |
elif "youtube.com" in question:
|
|
|
634 |
url_match = re.search(r'https://www\.youtube\.com/watch\?v=[^\s,?.]+', question)
|
635 |
if url_match:
|
636 |
url = url_match.group(0)
|
637 |
video_info = youtube_analyzer(url)
|
638 |
|
|
|
639 |
search_query = f"site:youtube.com {url} transcript content"
|
640 |
search_results = serper_search(search_query)
|
641 |
|
642 |
return f"Video Analysis: {video_info}\n\nAdditional Info: {search_results}"
|
643 |
|
|
|
644 |
elif "botanical" in question_lower and "vegetable" in question_lower:
|
|
|
645 |
list_match = re.search(r'milk.*?peanuts', question)
|
646 |
if list_match:
|
647 |
food_list = list_match.group(0)
|
648 |
return data_extractor(food_list, "botanical vegetables")
|
649 |
|
|
|
650 |
elif "commutative" in question_lower or "chess" in question_lower:
|
651 |
math_result = math_solver(question)
|
652 |
|
|
|
653 |
if "commutative" in question_lower:
|
654 |
search_result = serper_search("group theory commutative operation counter examples")
|
655 |
return f"{math_result}\n\nAdditional context: {search_result}"
|
656 |
|
657 |
return math_result
|
658 |
|
|
|
659 |
else:
|
|
|
660 |
search_results = serper_search(question)
|
661 |
|
|
|
662 |
if any(term in question_lower for term in ["mercedes sosa", "dinosaur", "wikipedia", "olympics"]):
|
663 |
wiki_results = wikipedia_search(question)
|
664 |
return f"Search Results: {search_results}\n\nWikipedia: {wiki_results}"
|
|
|
667 |
|
668 |
except Exception as e:
|
669 |
print(f"Error in agent processing: {e}")
|
|
|
670 |
try:
|
671 |
return serper_search(question)
|
672 |
except:
|
673 |
return f"I encountered an error processing this question: {question}. Please try rephrasing or breaking it into smaller parts."
|
674 |
|
675 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
676 |
+
"""Fetches all questions, runs the GAIA Agent on them, submits all answers"""
|
|
|
|
|
|
|
677 |
space_id = os.getenv("SPACE_ID")
|
678 |
|
679 |
if profile:
|
|
|
687 |
questions_url = f"{api_url}/questions"
|
688 |
submit_url = f"{api_url}/submit"
|
689 |
|
|
|
690 |
try:
|
691 |
agent = GAIAAgent()
|
692 |
except Exception as e:
|
|
|
696 |
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
697 |
print(agent_code)
|
698 |
|
|
|
699 |
print(f"Fetching questions from: {questions_url}")
|
700 |
try:
|
701 |
response = requests.get(questions_url, timeout=15)
|
|
|
716 |
print(f"An unexpected error occurred fetching questions: {e}")
|
717 |
return f"An unexpected error occurred fetching questions: {e}", None
|
718 |
|
|
|
719 |
results_log = []
|
720 |
answers_payload = []
|
721 |
print(f"Running agent on {len(questions_data)} questions...")
|
|
|
733 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
734 |
results_log.append({"Task ID": task_id, "Question": question_text[:100] + "...", "Submitted Answer": submitted_answer[:200] + "..."})
|
735 |
|
|
|
736 |
time.sleep(1)
|
737 |
|
738 |
except Exception as e:
|
|
|
743 |
print("Agent did not produce any answers to submit.")
|
744 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
745 |
|
|
|
746 |
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
|
747 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
748 |
print(status_update)
|
749 |
|
|
|
750 |
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
751 |
try:
|
752 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
|
|
828 |
if __name__ == "__main__":
|
829 |
print("\n" + "-"*30 + " GAIA Agent Starting " + "-"*30)
|
830 |
|
|
|
831 |
space_host_startup = os.getenv("SPACE_HOST")
|
832 |
space_id_startup = os.getenv("SPACE_ID")
|
833 |
serper_key = os.getenv("SERPER_API_KEY")
|