Create perplexity_client.py
Browse files- perplexity_client.py +724 -0
perplexity_client.py
ADDED
@@ -0,0 +1,724 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
π§ Perplexity AI Integration for AI Dataset Studio
|
3 |
+
Automatically discovers relevant sources based on project descriptions
|
4 |
+
"""
|
5 |
+
|
6 |
+
import os
|
7 |
+
import requests
|
8 |
+
import json
|
9 |
+
import logging
|
10 |
+
import time
|
11 |
+
import re
|
12 |
+
from typing import List, Dict, Optional, Tuple
|
13 |
+
from urllib.parse import urlparse, urljoin
|
14 |
+
from dataclasses import dataclass
|
15 |
+
from enum import Enum
|
16 |
+
|
17 |
+
# Configure logging
|
18 |
+
logging.basicConfig(level=logging.INFO)
|
19 |
+
logger = logging.getLogger(__name__)
|
20 |
+
|
21 |
+
class SearchType(Enum):
|
22 |
+
"""Types of searches supported by Perplexity AI"""
|
23 |
+
GENERAL = "general"
|
24 |
+
ACADEMIC = "academic"
|
25 |
+
NEWS = "news"
|
26 |
+
SOCIAL = "social"
|
27 |
+
TECHNICAL = "technical"
|
28 |
+
|
29 |
+
@dataclass
|
30 |
+
class SourceResult:
|
31 |
+
"""Structure for individual source results"""
|
32 |
+
url: str
|
33 |
+
title: str
|
34 |
+
description: str
|
35 |
+
relevance_score: float
|
36 |
+
source_type: str
|
37 |
+
domain: str
|
38 |
+
publication_date: Optional[str] = None
|
39 |
+
author: Optional[str] = None
|
40 |
+
|
41 |
+
@dataclass
|
42 |
+
class SearchResults:
|
43 |
+
"""Container for search results"""
|
44 |
+
query: str
|
45 |
+
sources: List[SourceResult]
|
46 |
+
total_found: int
|
47 |
+
search_time: float
|
48 |
+
perplexity_response: str
|
49 |
+
suggestions: List[str]
|
50 |
+
|
51 |
+
class PerplexityClient:
|
52 |
+
"""
|
53 |
+
π§ Perplexity AI Client for Smart Source Discovery
|
54 |
+
|
55 |
+
Features:
|
56 |
+
- Intelligent source discovery based on project descriptions
|
57 |
+
- Multiple search strategies (academic, news, technical, etc.)
|
58 |
+
- Quality filtering and relevance scoring
|
59 |
+
- Rate limiting and error handling
|
60 |
+
- Domain validation and safety checks
|
61 |
+
"""
|
62 |
+
|
63 |
+
def __init__(self, api_key: Optional[str] = None):
|
64 |
+
"""
|
65 |
+
Initialize Perplexity AI client
|
66 |
+
|
67 |
+
Args:
|
68 |
+
api_key: Perplexity API key (if not provided, will try env var)
|
69 |
+
"""
|
70 |
+
self.api_key = api_key or os.getenv('PERPLEXITY_API_KEY')
|
71 |
+
self.base_url = "https://api.perplexity.ai"
|
72 |
+
self.session = requests.Session()
|
73 |
+
|
74 |
+
# Set up headers
|
75 |
+
if self.api_key:
|
76 |
+
self.session.headers.update({
|
77 |
+
'Authorization': f'Bearer {self.api_key}',
|
78 |
+
'Content-Type': 'application/json',
|
79 |
+
'User-Agent': 'AI-Dataset-Studio/1.0'
|
80 |
+
})
|
81 |
+
|
82 |
+
# Rate limiting
|
83 |
+
self.last_request_time = 0
|
84 |
+
self.min_request_interval = 1.0 # Seconds between requests
|
85 |
+
|
86 |
+
# Configuration
|
87 |
+
self.max_retries = 3
|
88 |
+
self.timeout = 30
|
89 |
+
|
90 |
+
logger.info("π§ Perplexity AI client initialized")
|
91 |
+
|
92 |
+
def _validate_api_key(self) -> bool:
|
93 |
+
"""Validate that API key is available and working"""
|
94 |
+
if not self.api_key:
|
95 |
+
logger.error("β No Perplexity API key found. Set PERPLEXITY_API_KEY environment variable.")
|
96 |
+
return False
|
97 |
+
return True
|
98 |
+
|
99 |
+
def _rate_limit(self):
|
100 |
+
"""Implement rate limiting to respect API limits"""
|
101 |
+
current_time = time.time()
|
102 |
+
time_since_last = current_time - self.last_request_time
|
103 |
+
|
104 |
+
if time_since_last < self.min_request_interval:
|
105 |
+
sleep_time = self.min_request_interval - time_since_last
|
106 |
+
logger.debug(f"β±οΈ Rate limiting: sleeping {sleep_time:.2f}s")
|
107 |
+
time.sleep(sleep_time)
|
108 |
+
|
109 |
+
self.last_request_time = time.time()
|
110 |
+
|
111 |
+
def _make_request(self, payload: Dict) -> Optional[Dict]:
|
112 |
+
"""
|
113 |
+
Make API request to Perplexity with error handling
|
114 |
+
|
115 |
+
Args:
|
116 |
+
payload: Request payload
|
117 |
+
|
118 |
+
Returns:
|
119 |
+
API response or None if failed
|
120 |
+
"""
|
121 |
+
if not self._validate_api_key():
|
122 |
+
return None
|
123 |
+
|
124 |
+
self._rate_limit()
|
125 |
+
|
126 |
+
for attempt in range(self.max_retries):
|
127 |
+
try:
|
128 |
+
logger.debug(f"π‘ Making Perplexity API request (attempt {attempt + 1})")
|
129 |
+
|
130 |
+
response = self.session.post(
|
131 |
+
f"{self.base_url}/chat/completions",
|
132 |
+
json=payload,
|
133 |
+
timeout=self.timeout
|
134 |
+
)
|
135 |
+
|
136 |
+
if response.status_code == 200:
|
137 |
+
logger.debug("β
Perplexity API request successful")
|
138 |
+
return response.json()
|
139 |
+
elif response.status_code == 429:
|
140 |
+
logger.warning("π¦ Rate limit hit, waiting longer...")
|
141 |
+
time.sleep(2 ** attempt) # Exponential backoff
|
142 |
+
continue
|
143 |
+
else:
|
144 |
+
logger.error(f"β API request failed: {response.status_code} - {response.text}")
|
145 |
+
|
146 |
+
except requests.exceptions.Timeout:
|
147 |
+
logger.warning(f"β° Request timeout (attempt {attempt + 1})")
|
148 |
+
except requests.exceptions.RequestException as e:
|
149 |
+
logger.error(f"π Request error: {str(e)}")
|
150 |
+
|
151 |
+
if attempt < self.max_retries - 1:
|
152 |
+
time.sleep(2 ** attempt) # Exponential backoff
|
153 |
+
|
154 |
+
logger.error("β All retry attempts failed")
|
155 |
+
return None
|
156 |
+
|
157 |
+
def discover_sources(
|
158 |
+
self,
|
159 |
+
project_description: str,
|
160 |
+
search_type: SearchType = SearchType.GENERAL,
|
161 |
+
max_sources: int = 20,
|
162 |
+
include_academic: bool = True,
|
163 |
+
include_news: bool = True,
|
164 |
+
domain_filter: Optional[List[str]] = None
|
165 |
+
) -> SearchResults:
|
166 |
+
"""
|
167 |
+
π Discover relevant sources based on project description
|
168 |
+
|
169 |
+
Args:
|
170 |
+
project_description: User's project description
|
171 |
+
search_type: Type of search to perform
|
172 |
+
max_sources: Maximum number of sources to return
|
173 |
+
include_academic: Include academic sources
|
174 |
+
include_news: Include news sources
|
175 |
+
domain_filter: Optional list of domains to focus on
|
176 |
+
|
177 |
+
Returns:
|
178 |
+
SearchResults object with discovered sources
|
179 |
+
"""
|
180 |
+
start_time = time.time()
|
181 |
+
|
182 |
+
logger.info(f"π Discovering sources for: {project_description[:100]}...")
|
183 |
+
|
184 |
+
# Build search prompt
|
185 |
+
search_prompt = self._build_search_prompt(
|
186 |
+
project_description,
|
187 |
+
search_type,
|
188 |
+
max_sources,
|
189 |
+
include_academic,
|
190 |
+
include_news,
|
191 |
+
domain_filter
|
192 |
+
)
|
193 |
+
|
194 |
+
# Prepare API payload
|
195 |
+
payload = {
|
196 |
+
"model": "llama-3.1-sonar-large-128k-online",
|
197 |
+
"messages": [
|
198 |
+
{
|
199 |
+
"role": "system",
|
200 |
+
"content": "You are an expert research assistant specializing in finding high-quality, relevant sources for AI/ML dataset creation. Always provide specific URLs, titles, and descriptions."
|
201 |
+
},
|
202 |
+
{
|
203 |
+
"role": "user",
|
204 |
+
"content": search_prompt
|
205 |
+
}
|
206 |
+
],
|
207 |
+
"max_tokens": 4000,
|
208 |
+
"temperature": 0.3,
|
209 |
+
"top_p": 0.9
|
210 |
+
}
|
211 |
+
|
212 |
+
# Make API request
|
213 |
+
response = self._make_request(payload)
|
214 |
+
|
215 |
+
if not response:
|
216 |
+
logger.error("β Failed to get response from Perplexity API")
|
217 |
+
return self._create_empty_results(project_description, time.time() - start_time)
|
218 |
+
|
219 |
+
# Parse response and extract sources
|
220 |
+
try:
|
221 |
+
content = response['choices'][0]['message']['content']
|
222 |
+
sources = self._parse_sources_from_response(content)
|
223 |
+
suggestions = self._extract_suggestions(content)
|
224 |
+
|
225 |
+
search_time = time.time() - start_time
|
226 |
+
|
227 |
+
logger.info(f"β
Found {len(sources)} sources in {search_time:.2f}s")
|
228 |
+
|
229 |
+
return SearchResults(
|
230 |
+
query=project_description,
|
231 |
+
sources=sources[:max_sources],
|
232 |
+
total_found=len(sources),
|
233 |
+
search_time=search_time,
|
234 |
+
perplexity_response=content,
|
235 |
+
suggestions=suggestions
|
236 |
+
)
|
237 |
+
|
238 |
+
except Exception as e:
|
239 |
+
logger.error(f"β Error parsing Perplexity response: {str(e)}")
|
240 |
+
return self._create_empty_results(project_description, time.time() - start_time)
|
241 |
+
|
242 |
+
def _build_search_prompt(
|
243 |
+
self,
|
244 |
+
project_description: str,
|
245 |
+
search_type: SearchType,
|
246 |
+
max_sources: int,
|
247 |
+
include_academic: bool,
|
248 |
+
include_news: bool,
|
249 |
+
domain_filter: Optional[List[str]]
|
250 |
+
) -> str:
|
251 |
+
"""Build optimized search prompt for Perplexity AI"""
|
252 |
+
|
253 |
+
prompt = f"""
|
254 |
+
Find {max_sources} high-quality, diverse sources for an AI/ML dataset creation project:
|
255 |
+
|
256 |
+
PROJECT DESCRIPTION: {project_description}
|
257 |
+
|
258 |
+
SEARCH REQUIREMENTS:
|
259 |
+
- Find sources with extractable text content suitable for ML training
|
260 |
+
- Prioritize sources with structured, high-quality content
|
261 |
+
- Include diverse perspectives and data types
|
262 |
+
- Focus on sources that are legally scrapable (respect robots.txt)
|
263 |
+
|
264 |
+
SEARCH TYPE: {search_type.value}
|
265 |
+
"""
|
266 |
+
|
267 |
+
if include_academic:
|
268 |
+
prompt += "\n- Include academic papers, research articles, and scholarly sources"
|
269 |
+
|
270 |
+
if include_news:
|
271 |
+
prompt += "\n- Include news articles, press releases, and journalistic content"
|
272 |
+
|
273 |
+
if domain_filter:
|
274 |
+
prompt += f"\n- Focus on these domains: {', '.join(domain_filter)}"
|
275 |
+
|
276 |
+
prompt += f"""
|
277 |
+
|
278 |
+
OUTPUT FORMAT:
|
279 |
+
For each source, provide:
|
280 |
+
1. **URL**: Direct link to the content
|
281 |
+
2. **Title**: Clear, descriptive title
|
282 |
+
3. **Description**: 2-3 sentence summary of content and why it's relevant
|
283 |
+
4. **Type**: [academic/news/blog/government/technical/forum/social]
|
284 |
+
5. **Quality Score**: 1-10 rating for dataset suitability
|
285 |
+
|
286 |
+
ADDITIONAL REQUIREMENTS:
|
287 |
+
- Verify URLs are accessible and contain substantial text
|
288 |
+
- Avoid paywalled or login-required content when possible
|
289 |
+
- Prioritize sources with consistent formatting
|
290 |
+
- Include publication dates when available
|
291 |
+
- Suggest related search terms for expanding the dataset
|
292 |
+
|
293 |
+
Please provide concrete, actionable sources that can be immediately scraped for dataset creation.
|
294 |
+
"""
|
295 |
+
|
296 |
+
return prompt
|
297 |
+
|
298 |
+
def _parse_sources_from_response(self, content: str) -> List[SourceResult]:
|
299 |
+
"""Parse source information from Perplexity AI response"""
|
300 |
+
sources = []
|
301 |
+
|
302 |
+
# Try to extract structured information
|
303 |
+
# Look for URL patterns
|
304 |
+
url_pattern = r'https?://[^\s<>"{}|\\^`\[\]]+[^\s<>"{}|\\^`\[\].,!?;:]'
|
305 |
+
|
306 |
+
# Split content into sections
|
307 |
+
sections = re.split(r'\n\s*\n', content)
|
308 |
+
|
309 |
+
for section in sections:
|
310 |
+
# Look for URLs in this section
|
311 |
+
urls = re.findall(url_pattern, section, re.IGNORECASE)
|
312 |
+
|
313 |
+
if urls:
|
314 |
+
for url in urls:
|
315 |
+
try:
|
316 |
+
# Clean URL
|
317 |
+
url = url.strip()
|
318 |
+
|
319 |
+
# Extract title (look for text before the URL or after)
|
320 |
+
title = self._extract_title_from_section(section, url)
|
321 |
+
|
322 |
+
# Extract description
|
323 |
+
description = self._extract_description_from_section(section, url)
|
324 |
+
|
325 |
+
# Determine source type
|
326 |
+
source_type = self._determine_source_type(url, section)
|
327 |
+
|
328 |
+
# Calculate relevance score (basic heuristic)
|
329 |
+
relevance_score = self._calculate_relevance_score(section, url)
|
330 |
+
|
331 |
+
# Get domain
|
332 |
+
domain = self._extract_domain(url)
|
333 |
+
|
334 |
+
# Validate URL
|
335 |
+
if self._is_valid_url(url):
|
336 |
+
source = SourceResult(
|
337 |
+
url=url,
|
338 |
+
title=title,
|
339 |
+
description=description,
|
340 |
+
relevance_score=relevance_score,
|
341 |
+
source_type=source_type,
|
342 |
+
domain=domain
|
343 |
+
)
|
344 |
+
sources.append(source)
|
345 |
+
|
346 |
+
except Exception as e:
|
347 |
+
logger.debug(f"β οΈ Error parsing source: {str(e)}")
|
348 |
+
continue
|
349 |
+
|
350 |
+
# Remove duplicates based on URL
|
351 |
+
seen_urls = set()
|
352 |
+
unique_sources = []
|
353 |
+
|
354 |
+
for source in sources:
|
355 |
+
if source.url not in seen_urls:
|
356 |
+
seen_urls.add(source.url)
|
357 |
+
unique_sources.append(source)
|
358 |
+
|
359 |
+
# Sort by relevance score
|
360 |
+
unique_sources.sort(key=lambda x: x.relevance_score, reverse=True)
|
361 |
+
|
362 |
+
return unique_sources
|
363 |
+
|
364 |
+
def _extract_title_from_section(self, section: str, url: str) -> str:
|
365 |
+
"""Extract title from section text"""
|
366 |
+
lines = section.split('\n')
|
367 |
+
|
368 |
+
for line in lines:
|
369 |
+
if url in line:
|
370 |
+
# Look for title patterns
|
371 |
+
title_patterns = [
|
372 |
+
r'\*\*([^*]+)\*\*', # **Title**
|
373 |
+
r'#{1,6}\s*([^\n]+)', # # Title
|
374 |
+
r'Title:\s*([^\n]+)', # Title: Something
|
375 |
+
r'([^:\n]+):?\s*' + re.escape(url), # Title: URL
|
376 |
+
]
|
377 |
+
|
378 |
+
for pattern in title_patterns:
|
379 |
+
match = re.search(pattern, line, re.IGNORECASE)
|
380 |
+
if match:
|
381 |
+
return match.group(1).strip()
|
382 |
+
|
383 |
+
# Fallback: use domain name
|
384 |
+
return self._extract_domain(url)
|
385 |
+
|
386 |
+
def _extract_description_from_section(self, section: str, url: str) -> str:
|
387 |
+
"""Extract description from section text"""
|
388 |
+
# Remove the URL line and look for descriptive text
|
389 |
+
lines = section.split('\n')
|
390 |
+
description_lines = []
|
391 |
+
|
392 |
+
for line in lines:
|
393 |
+
if url not in line and line.strip():
|
394 |
+
# Skip markdown headers and bullets
|
395 |
+
clean_line = re.sub(r'^[#*\-\d\.]+\s*', '', line.strip())
|
396 |
+
if len(clean_line) > 20: # Meaningful content
|
397 |
+
description_lines.append(clean_line)
|
398 |
+
|
399 |
+
description = ' '.join(description_lines)
|
400 |
+
|
401 |
+
# Truncate if too long
|
402 |
+
if len(description) > 200:
|
403 |
+
description = description[:200] + "..."
|
404 |
+
|
405 |
+
return description or "High-quality source for dataset creation"
|
406 |
+
|
407 |
+
def _determine_source_type(self, url: str, section: str) -> str:
|
408 |
+
"""Determine the type of source based on URL and context"""
|
409 |
+
url_lower = url.lower()
|
410 |
+
section_lower = section.lower()
|
411 |
+
|
412 |
+
# Academic sources
|
413 |
+
if any(domain in url_lower for domain in [
|
414 |
+
'arxiv.org', 'scholar.google', 'pubmed', 'ieee.org',
|
415 |
+
'acm.org', 'springer.com', 'elsevier.com', 'nature.com',
|
416 |
+
'sciencedirect.com', 'jstor.org'
|
417 |
+
]):
|
418 |
+
return 'academic'
|
419 |
+
|
420 |
+
# News sources
|
421 |
+
if any(domain in url_lower for domain in [
|
422 |
+
'cnn.com', 'bbc.com', 'reuters.com', 'ap.org', 'nytimes.com',
|
423 |
+
'washingtonpost.com', 'theguardian.com', 'bloomberg.com',
|
424 |
+
'techcrunch.com', 'wired.com'
|
425 |
+
]):
|
426 |
+
return 'news'
|
427 |
+
|
428 |
+
# Government sources
|
429 |
+
if '.gov' in url_lower or 'government' in section_lower:
|
430 |
+
return 'government'
|
431 |
+
|
432 |
+
# Technical/Documentation
|
433 |
+
if any(domain in url_lower for domain in [
|
434 |
+
'docs.', 'documentation', 'github.com', 'stackoverflow.com',
|
435 |
+
'medium.com', 'dev.to'
|
436 |
+
]):
|
437 |
+
return 'technical'
|
438 |
+
|
439 |
+
# Social media
|
440 |
+
if any(domain in url_lower for domain in [
|
441 |
+
'twitter.com', 'reddit.com', 'linkedin.com', 'facebook.com'
|
442 |
+
]):
|
443 |
+
return 'social'
|
444 |
+
|
445 |
+
# Default to blog
|
446 |
+
return 'blog'
|
447 |
+
|
448 |
+
def _calculate_relevance_score(self, section: str, url: str) -> float:
|
449 |
+
"""Calculate relevance score for a source (0-10)"""
|
450 |
+
score = 5.0 # Base score
|
451 |
+
|
452 |
+
section_lower = section.lower()
|
453 |
+
url_lower = url.lower()
|
454 |
+
|
455 |
+
# Boost for quality indicators
|
456 |
+
quality_indicators = [
|
457 |
+
'research', 'study', 'analysis', 'comprehensive', 'detailed',
|
458 |
+
'expert', 'professional', 'authoritative', 'peer-reviewed',
|
459 |
+
'dataset', 'data', 'machine learning', 'ai', 'artificial intelligence'
|
460 |
+
]
|
461 |
+
|
462 |
+
for indicator in quality_indicators:
|
463 |
+
if indicator in section_lower:
|
464 |
+
score += 0.5
|
465 |
+
|
466 |
+
# Boost for academic sources
|
467 |
+
if any(domain in url_lower for domain in ['arxiv.org', 'scholar.google', 'pubmed']):
|
468 |
+
score += 2.0
|
469 |
+
|
470 |
+
# Boost for government sources
|
471 |
+
if '.gov' in url_lower:
|
472 |
+
score += 1.5
|
473 |
+
|
474 |
+
# Penalize for social media
|
475 |
+
if any(domain in url_lower for domain in ['twitter.com', 'facebook.com']):
|
476 |
+
score -= 1.0
|
477 |
+
|
478 |
+
# Cap at 10
|
479 |
+
return min(score, 10.0)
|
480 |
+
|
481 |
+
def _extract_domain(self, url: str) -> str:
|
482 |
+
"""Extract domain from URL"""
|
483 |
+
try:
|
484 |
+
parsed = urlparse(url)
|
485 |
+
return parsed.netloc
|
486 |
+
except:
|
487 |
+
return "unknown"
|
488 |
+
|
489 |
+
def _is_valid_url(self, url: str) -> bool:
|
490 |
+
"""Validate URL format and basic accessibility"""
|
491 |
+
try:
|
492 |
+
parsed = urlparse(url)
|
493 |
+
return all([parsed.scheme, parsed.netloc])
|
494 |
+
except:
|
495 |
+
return False
|
496 |
+
|
497 |
+
def _extract_suggestions(self, content: str) -> List[str]:
|
498 |
+
"""Extract search suggestions from Perplexity response"""
|
499 |
+
suggestions = []
|
500 |
+
|
501 |
+
# Look for suggestion patterns
|
502 |
+
suggestion_patterns = [
|
503 |
+
r'related search terms?:?\s*([^\n]+)',
|
504 |
+
r'you might also search for:?\s*([^\n]+)',
|
505 |
+
r'additional keywords?:?\s*([^\n]+)',
|
506 |
+
r'suggestions?:?\s*([^\n]+)'
|
507 |
+
]
|
508 |
+
|
509 |
+
for pattern in suggestion_patterns:
|
510 |
+
matches = re.findall(pattern, content, re.IGNORECASE)
|
511 |
+
for match in matches:
|
512 |
+
# Split by common delimiters
|
513 |
+
terms = re.split(r'[,;|]', match)
|
514 |
+
suggestions.extend([term.strip().strip('"\'') for term in terms if term.strip()])
|
515 |
+
|
516 |
+
return suggestions[:10] # Limit to 10 suggestions
|
517 |
+
|
518 |
+
def _create_empty_results(self, query: str, search_time: float) -> SearchResults:
|
519 |
+
"""Create empty results object for failed searches"""
|
520 |
+
return SearchResults(
|
521 |
+
query=query,
|
522 |
+
sources=[],
|
523 |
+
total_found=0,
|
524 |
+
search_time=search_time,
|
525 |
+
perplexity_response="",
|
526 |
+
suggestions=[]
|
527 |
+
)
|
528 |
+
|
529 |
+
def search_with_keywords(self, keywords: List[str], search_type: SearchType = SearchType.GENERAL) -> SearchResults:
|
530 |
+
"""
|
531 |
+
π Search using specific keywords
|
532 |
+
|
533 |
+
Args:
|
534 |
+
keywords: List of search keywords
|
535 |
+
search_type: Type of search to perform
|
536 |
+
|
537 |
+
Returns:
|
538 |
+
SearchResults object
|
539 |
+
"""
|
540 |
+
query = " ".join(keywords)
|
541 |
+
return self.discover_sources(
|
542 |
+
project_description=f"Find sources related to: {query}",
|
543 |
+
search_type=search_type
|
544 |
+
)
|
545 |
+
|
546 |
+
def get_domain_sources(self, domain: str, topic: str, max_sources: int = 10) -> SearchResults:
|
547 |
+
"""
|
548 |
+
π Find sources from a specific domain
|
549 |
+
|
550 |
+
Args:
|
551 |
+
domain: Target domain (e.g., "nature.com")
|
552 |
+
topic: Topic to search for
|
553 |
+
max_sources: Maximum sources to return
|
554 |
+
|
555 |
+
Returns:
|
556 |
+
SearchResults object
|
557 |
+
"""
|
558 |
+
return self.discover_sources(
|
559 |
+
project_description=f"Find articles about {topic} from {domain}",
|
560 |
+
domain_filter=[domain],
|
561 |
+
max_sources=max_sources
|
562 |
+
)
|
563 |
+
|
564 |
+
def validate_sources(self, sources: List[SourceResult]) -> List[SourceResult]:
|
565 |
+
"""
|
566 |
+
β
Validate and filter sources for quality and accessibility
|
567 |
+
|
568 |
+
Args:
|
569 |
+
sources: List of source results to validate
|
570 |
+
|
571 |
+
Returns:
|
572 |
+
Filtered list of valid sources
|
573 |
+
"""
|
574 |
+
valid_sources = []
|
575 |
+
|
576 |
+
for source in sources:
|
577 |
+
try:
|
578 |
+
# Basic URL validation
|
579 |
+
if not self._is_valid_url(source.url):
|
580 |
+
logger.debug(f"β οΈ Invalid URL: {source.url}")
|
581 |
+
continue
|
582 |
+
|
583 |
+
# Check if domain is accessible (basic check)
|
584 |
+
domain = self._extract_domain(source.url)
|
585 |
+
if not domain or domain == "unknown":
|
586 |
+
logger.debug(f"β οΈ Unknown domain: {source.url}")
|
587 |
+
continue
|
588 |
+
|
589 |
+
# Quality score threshold
|
590 |
+
if source.relevance_score < 3.0:
|
591 |
+
logger.debug(f"β οΈ Low quality score: {source.url}")
|
592 |
+
continue
|
593 |
+
|
594 |
+
valid_sources.append(source)
|
595 |
+
|
596 |
+
except Exception as e:
|
597 |
+
logger.debug(f"β οΈ Error validating source {source.url}: {str(e)}")
|
598 |
+
continue
|
599 |
+
|
600 |
+
logger.info(f"β
Validated {len(valid_sources)} out of {len(sources)} sources")
|
601 |
+
return valid_sources
|
602 |
+
|
603 |
+
def export_sources(self, results: SearchResults, format: str = "json") -> str:
|
604 |
+
"""
|
605 |
+
π Export search results to various formats
|
606 |
+
|
607 |
+
Args:
|
608 |
+
results: SearchResults object to export
|
609 |
+
format: Export format ("json", "csv", "markdown")
|
610 |
+
|
611 |
+
Returns:
|
612 |
+
Exported data as string
|
613 |
+
"""
|
614 |
+
if format.lower() == "json":
|
615 |
+
return self._export_json(results)
|
616 |
+
elif format.lower() == "csv":
|
617 |
+
return self._export_csv(results)
|
618 |
+
elif format.lower() == "markdown":
|
619 |
+
return self._export_markdown(results)
|
620 |
+
else:
|
621 |
+
raise ValueError(f"Unsupported export format: {format}")
|
622 |
+
|
623 |
+
def _export_json(self, results: SearchResults) -> str:
|
624 |
+
"""Export results as JSON"""
|
625 |
+
data = {
|
626 |
+
"query": results.query,
|
627 |
+
"total_found": results.total_found,
|
628 |
+
"search_time": results.search_time,
|
629 |
+
"sources": [
|
630 |
+
{
|
631 |
+
"url": source.url,
|
632 |
+
"title": source.title,
|
633 |
+
"description": source.description,
|
634 |
+
"relevance_score": source.relevance_score,
|
635 |
+
"source_type": source.source_type,
|
636 |
+
"domain": source.domain,
|
637 |
+
"publication_date": source.publication_date,
|
638 |
+
"author": source.author
|
639 |
+
}
|
640 |
+
for source in results.sources
|
641 |
+
],
|
642 |
+
"suggestions": results.suggestions
|
643 |
+
}
|
644 |
+
return json.dumps(data, indent=2)
|
645 |
+
|
646 |
+
def _export_csv(self, results: SearchResults) -> str:
|
647 |
+
"""Export results as CSV"""
|
648 |
+
import csv
|
649 |
+
from io import StringIO
|
650 |
+
|
651 |
+
output = StringIO()
|
652 |
+
writer = csv.writer(output)
|
653 |
+
|
654 |
+
# Write header
|
655 |
+
writer.writerow([
|
656 |
+
"URL", "Title", "Description", "Relevance Score",
|
657 |
+
"Source Type", "Domain", "Publication Date", "Author"
|
658 |
+
])
|
659 |
+
|
660 |
+
# Write data
|
661 |
+
for source in results.sources:
|
662 |
+
writer.writerow([
|
663 |
+
source.url,
|
664 |
+
source.title,
|
665 |
+
source.description,
|
666 |
+
source.relevance_score,
|
667 |
+
source.source_type,
|
668 |
+
source.domain,
|
669 |
+
source.publication_date or "",
|
670 |
+
source.author or ""
|
671 |
+
])
|
672 |
+
|
673 |
+
return output.getvalue()
|
674 |
+
|
675 |
+
def _export_markdown(self, results: SearchResults) -> str:
|
676 |
+
"""Export results as Markdown"""
|
677 |
+
md = f"# Search Results for: {results.query}\n\n"
|
678 |
+
md += f"**Total Sources Found:** {results.total_found}\n"
|
679 |
+
md += f"**Search Time:** {results.search_time:.2f} seconds\n\n"
|
680 |
+
|
681 |
+
md += "## Sources\n\n"
|
682 |
+
|
683 |
+
for i, source in enumerate(results.sources, 1):
|
684 |
+
md += f"### {i}. {source.title}\n\n"
|
685 |
+
md += f"**URL:** {source.url}\n"
|
686 |
+
md += f"**Type:** {source.source_type}\n"
|
687 |
+
md += f"**Domain:** {source.domain}\n"
|
688 |
+
md += f"**Relevance Score:** {source.relevance_score}/10\n"
|
689 |
+
md += f"**Description:** {source.description}\n\n"
|
690 |
+
|
691 |
+
if results.suggestions:
|
692 |
+
md += "## Related Search Suggestions\n\n"
|
693 |
+
for suggestion in results.suggestions:
|
694 |
+
md += f"- {suggestion}\n"
|
695 |
+
|
696 |
+
return md
|
697 |
+
|
698 |
+
# Example usage and testing functions
|
699 |
+
def test_perplexity_client():
|
700 |
+
"""Test function for Perplexity client"""
|
701 |
+
client = PerplexityClient()
|
702 |
+
|
703 |
+
if not client._validate_api_key():
|
704 |
+
print("β No API key found. Set PERPLEXITY_API_KEY environment variable.")
|
705 |
+
return
|
706 |
+
|
707 |
+
# Test search
|
708 |
+
results = client.discover_sources(
|
709 |
+
project_description="Create a dataset for sentiment analysis of product reviews",
|
710 |
+
search_type=SearchType.GENERAL,
|
711 |
+
max_sources=10
|
712 |
+
)
|
713 |
+
|
714 |
+
print(f"π Found {len(results.sources)} sources")
|
715 |
+
for source in results.sources[:3]:
|
716 |
+
print(f" - {source.title}: {source.url}")
|
717 |
+
|
718 |
+
# Test export
|
719 |
+
json_export = client.export_sources(results, "json")
|
720 |
+
print(f"π JSON export: {len(json_export)} characters")
|
721 |
+
|
722 |
+
if __name__ == "__main__":
|
723 |
+
# Test the client
|
724 |
+
test_perplexity_client()
|