signalwire-agents 0.1.23__py3-none-any.whl → 0.1.24__py3-none-any.whl

signalwire_agents/skills/spider/skill.py

@@ -0,0 +1,479 @@
+"""Spider skill for fast web scraping with SignalWire AI Agents."""
+import re
+import logging
+from typing import Dict, Any, Optional, List
+from urllib.parse import urljoin, urlparse
+import requests
+from lxml import html
+from lxml.etree import XPathEvalError
+
+from signalwire_agents.core.skill_base import SkillBase
+from signalwire_agents.core.function_result import SwaigFunctionResult
+
+
+class SpiderSkill(SkillBase):
+    """Fast web scraping skill optimized for speed and token efficiency."""
+    
+    SKILL_NAME = "spider"
+    SKILL_DESCRIPTION = "Fast web scraping and crawling capabilities"
+    SKILL_VERSION = "1.0.0"
+    REQUIRED_PACKAGES = ["lxml"]  # beautifulsoup4 and requests are in base dependencies
+    REQUIRED_ENV_VARS = []  # No required env vars by default
+    SUPPORTS_MULTIPLE_INSTANCES = True
+    
+    # Compiled regex for performance
+    WHITESPACE_REGEX = re.compile(r'\s+')
+    
+    def __init__(self, agent, params: Dict[str, Any]):
+        """Initialize the spider skill with configuration parameters."""
+        super().__init__(agent, params)
+        
+        # Performance settings
+        self.delay = self.params.get('delay', 0.1)
+        self.concurrent_requests = self.params.get('concurrent_requests', 5)
+        self.timeout = self.params.get('timeout', 5)
+        
+        # Crawling limits
+        self.max_pages = self.params.get('max_pages', 1)
+        self.max_depth = self.params.get('max_depth', 0)
+        
+        # Content processing
+        self.extract_type = self.params.get('extract_type', 'fast_text')
+        self.max_text_length = self.params.get('max_text_length', 3000)
+        self.clean_text = self.params.get('clean_text', True)
+        
+        # Features
+        self.cache_enabled = self.params.get('cache_enabled', True)
+        self.follow_robots_txt = self.params.get('follow_robots_txt', False)
+        self.user_agent = self.params.get('user_agent', 'Spider/1.0 (SignalWire AI Agent)')
+        
+        # Optional headers
+        self.headers = self.params.get('headers', {})
+        self.headers['User-Agent'] = self.user_agent
+        
+        # Session for connection pooling
+        self.session = requests.Session()
+        self.session.headers.update(self.headers)
+        
+        # Cache for responses
+        self.cache = {} if self.cache_enabled else None
+        
+        # XPath expressions for unwanted elements
+        self.remove_xpaths = [
+            '//script', '//style', '//nav', '//header', 
+            '//footer', '//aside', '//noscript'
+        ]
+    
+    def get_instance_key(self) -> str:
+        """Return unique key for this skill instance."""
+        tool_name = self.params.get('tool_name', self.SKILL_NAME)
+        return f"{self.SKILL_NAME}_{tool_name}"
+    
+    def setup(self) -> bool:
+        """Validate configuration and setup the skill."""
+        # Validate delay is reasonable
+        if self.delay < 0:
+            self.logger.error("Delay cannot be negative")
+            return False
+            
+        # Validate concurrent requests
+        if not 1 <= self.concurrent_requests <= 20:
+            self.logger.error("Concurrent requests must be between 1 and 20")
+            return False
+            
+        # Validate max pages and depth
+        if self.max_pages < 1:
+            self.logger.error("Max pages must be at least 1")
+            return False
+            
+        if self.max_depth < 0:
+            self.logger.error("Max depth cannot be negative")
+            return False
+            
+        self.logger.info(f"Spider skill configured: delay={self.delay}s, max_pages={self.max_pages}, max_depth={self.max_depth}")
+        return True
+    
+    def register_tools(self) -> None:
+        """Register the web scraping tools with the agent."""
+        # Tool name prefix for multiple instances
+        tool_prefix = self.params.get('tool_name', '')
+        if tool_prefix:
+            tool_prefix = f"{tool_prefix}_"
+        
+        # Register scrape_url tool
+        self.agent.define_tool(
+            name=f"{tool_prefix}scrape_url",
+            description="Extract text content from a single web page",
+            parameters={
+                "url": {
+                    "type": "string",
+                    "description": "The URL to scrape"
+                }
+            },
+            required=["url"],
+            handler=self._scrape_url_handler,
+            **self.swaig_fields
+        )
+        
+        # Register crawl_site tool
+        self.agent.define_tool(
+            name=f"{tool_prefix}crawl_site",
+            description="Crawl multiple pages starting from a URL",
+            parameters={
+                "start_url": {
+                    "type": "string",
+                    "description": "Starting URL for the crawl"
+                }
+            },
+            required=["start_url"],
+            handler=self._crawl_site_handler,
+            **self.swaig_fields
+        )
+        
+        # Register extract_structured_data tool
+        self.agent.define_tool(
+            name=f"{tool_prefix}extract_structured_data",
+            description="Extract specific data from a web page using selectors",
+            parameters={
+                "url": {
+                    "type": "string",
+                    "description": "The URL to scrape"
+                }
+            },
+            required=["url"],
+            handler=self._extract_structured_handler,
+            **self.swaig_fields
+        )
+    
+    def _fetch_url(self, url: str) -> Optional[requests.Response]:
+        """Fetch a URL with caching and error handling."""
+        # Check cache first
+        if self.cache_enabled and url in self.cache:
+            self.logger.debug(f"Cache hit for {url}")
+            return self.cache[url]
+        
+        try:
+            response = self.session.get(url, timeout=self.timeout)
+            response.raise_for_status()
+            
+            # Cache successful responses
+            if self.cache_enabled:
+                self.cache[url] = response
+                
+            return response
+            
+        except requests.exceptions.Timeout:
+            self.logger.error(f"Timeout fetching {url}")
+            return None
+        except requests.exceptions.RequestException as e:
+            self.logger.error(f"Error fetching {url}: {e}")
+            return None
+    
+    def _fast_text_extract(self, response: requests.Response) -> str:
+        """Ultra-fast text extraction using lxml."""
+        try:
+            # Parse HTML with lxml
+            tree = html.fromstring(response.content)
+            
+            # Remove unwanted elements in one pass
+            for xpath in self.remove_xpaths:
+                for elem in tree.xpath(xpath):
+                    elem.drop_tree()
+            
+            # Extract text
+            text = tree.text_content()
+            
+            # Clean whitespace if requested
+            if self.clean_text:
+                text = self.WHITESPACE_REGEX.sub(' ', text).strip()
+            
+            # Smart truncation
+            if len(text) > self.max_text_length:
+                keep_start = self.max_text_length * 2 // 3
+                keep_end = self.max_text_length // 3
+                text = (
+                    text[:keep_start] + 
+                    "\n\n[...CONTENT TRUNCATED...]\n\n" + 
+                    text[-keep_end:]
+                )
+            
+            return text
+            
+        except Exception as e:
+            self.logger.error(f"Error extracting text: {e}")
+            return ""
+    
+    def _markdown_extract(self, response: requests.Response) -> str:
+        """Extract content in markdown format."""
+        try:
+            from bs4 import BeautifulSoup
+            
+            soup = BeautifulSoup(response.content, 'html.parser')
+            
+            # Remove unwanted tags
+            for tag in ['script', 'style', 'nav', 'header', 'footer', 'aside']:
+                for elem in soup.find_all(tag):
+                    elem.decompose()
+            
+            # Convert to markdown-like format
+            text_parts = []
+            
+            # Title
+            title = soup.find('title')
+            if title:
+                text_parts.append(f"# {title.get_text().strip()}\n")
+            
+            # Main content
+            for elem in soup.find_all(['h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'p', 'li', 'code', 'pre']):
+                if elem.name.startswith('h'):
+                    level = int(elem.name[1])
+                    text_parts.append(f"\n{'#' * level} {elem.get_text().strip()}\n")
+                elif elem.name == 'p':
+                    text_parts.append(f"\n{elem.get_text().strip()}\n")
+                elif elem.name == 'li':
+                    text_parts.append(f"- {elem.get_text().strip()}")
+                elif elem.name in ['code', 'pre']:
+                    text_parts.append(f"\n```\n{elem.get_text().strip()}\n```\n")
+            
+            text = '\n'.join(text_parts)
+            
+            # Truncate if needed
+            if len(text) > self.max_text_length:
+                text = text[:self.max_text_length] + "\n\n[...TRUNCATED...]"
+            
+            return text
+            
+        except ImportError:
+            self.logger.warning("BeautifulSoup not available, falling back to fast_text")
+            return self._fast_text_extract(response)
+        except Exception as e:
+            self.logger.error(f"Error in markdown extraction: {e}")
+            return self._fast_text_extract(response)
+    
+    def _structured_extract(self, response: requests.Response, selectors: Dict[str, str] = None) -> Dict[str, Any]:
+        """Extract structured data using selectors."""
+        try:
+            tree = html.fromstring(response.content)
+            result = {
+                "url": response.url,
+                "status_code": response.status_code,
+                "title": "",
+                "data": {}
+            }
+            
+            # Get title
+            title_elem = tree.xpath('//title/text()')
+            if title_elem:
+                result["title"] = title_elem[0].strip()
+            
+            # Extract using provided selectors
+            if selectors:
+                for field, selector in selectors.items():
+                    try:
+                        if selector.startswith('/'):  # XPath
+                            values = tree.xpath(selector)
+                        else:  # CSS selector
+                            from lxml.cssselect import CSSSelector
+                            sel = CSSSelector(selector)
+                            values = sel(tree)
+                        
+                        # Extract text from elements
+                        if values:
+                            if len(values) == 1:
+                                result["data"][field] = values[0].text_content().strip()
+                            else:
+                                result["data"][field] = [v.text_content().strip() for v in values]
+                    except (XPathEvalError, Exception) as e:
+                        self.logger.warning(f"Error with selector {selector}: {e}")
+                        result["data"][field] = None
+            
+            return result
+            
+        except Exception as e:
+            self.logger.error(f"Error in structured extraction: {e}")
+            return {"error": str(e)}
+    
+    def _scrape_url_handler(self, args: Dict[str, Any], raw_data: Dict[str, Any]) -> SwaigFunctionResult:
+        """Handle single page scraping."""
+        url = args.get('url', '').strip()
+        if not url:
+            return SwaigFunctionResult("Please provide a URL to scrape")
+        
+        # Validate URL
+        parsed = urlparse(url)
+        if not parsed.scheme or not parsed.netloc:
+            return SwaigFunctionResult(f"Invalid URL: {url}")
+        
+        # Fetch the page
+        response = self._fetch_url(url)
+        if not response:
+            return SwaigFunctionResult(f"Failed to fetch {url}")
+        
+        # Extract content based on configured type (not from args)
+        extract_type = self.extract_type
+        
+        try:
+            if extract_type == 'structured':
+                # For structured extraction, use predefined selectors from config if available
+                selectors = self.params.get('selectors', {})
+                result = self._structured_extract(response, selectors)
+                return SwaigFunctionResult(f"Extracted structured data from {url}: {result}")
+            elif extract_type == 'markdown':
+                content = self._markdown_extract(response)
+            else:  # fast_text (default)
+                content = self._fast_text_extract(response)
+            
+            if not content:
+                return SwaigFunctionResult(f"No content extracted from {url}")
+            
+            # Format response
+            char_count = len(content)
+            header = f"Content from {url} ({char_count} characters):\n\n"
+            
+            return SwaigFunctionResult(header + content)
+            
+        except Exception as e:
+            self.logger.error(f"Error processing {url}: {e}")
+            return SwaigFunctionResult(f"Error processing {url}: {str(e)}")
+    
+    def _crawl_site_handler(self, args: Dict[str, Any], raw_data: Dict[str, Any]) -> SwaigFunctionResult:
+        """Handle multi-page crawling."""
+        start_url = args.get('start_url', '').strip()
+        if not start_url:
+            return SwaigFunctionResult("Please provide a starting URL for the crawl")
+        
+        # Use configured parameters (not from args)
+        max_depth = self.max_depth
+        max_pages = self.max_pages
+        follow_patterns = self.params.get('follow_patterns', [])
+        
+        # Validate parameters
+        if max_depth < 0:
+            return SwaigFunctionResult("Max depth cannot be negative")
+        if max_pages < 1:
+            return SwaigFunctionResult("Max pages must be at least 1")
+        
+        # Simple breadth-first crawl
+        visited = set()
+        to_visit = [(start_url, 0)]  # (url, depth)
+        results = []
+        
+        while to_visit and len(visited) < max_pages:
+            if not to_visit:
+                break
+                
+            url, depth = to_visit.pop(0)
+            
+            # Skip if already visited or depth exceeded
+            if url in visited or depth > max_depth:
+                continue
+            
+            # Fetch and process page
+            response = self._fetch_url(url)
+            if not response:
+                continue
+            
+            visited.add(url)
+            
+            # Extract content
+            content = self._fast_text_extract(response)
+            if content:
+                results.append({
+                    'url': url,
+                    'depth': depth,
+                    'content_length': len(content),
+                    'summary': content[:500] + '...' if len(content) > 500 else content
+                })
+            
+            # Extract links if not at max depth
+            if depth < max_depth:
+                try:
+                    tree = html.fromstring(response.content)
+                    links = tree.xpath('//a[@href]/@href')
+                    
+                    for link in links:
+                        absolute_url = urljoin(url, link)
+                        
+                        # Check if we should follow this link
+                        if follow_patterns:
+                            if not any(re.search(pattern, absolute_url) for pattern in follow_patterns):
+                                continue
+                        
+                        # Only follow same domain by default
+                        if urlparse(absolute_url).netloc == urlparse(start_url).netloc:
+                            if absolute_url not in visited:
+                                to_visit.append((absolute_url, depth + 1))
+                                
+                except Exception as e:
+                    self.logger.warning(f"Error extracting links from {url}: {e}")
+            
+            # Respect delay between requests
+            if self.delay > 0 and len(visited) < max_pages:
+                import time
+                time.sleep(self.delay)
+        
+        # Format results
+        if not results:
+            return SwaigFunctionResult(f"No pages could be crawled from {start_url}")
+        
+        summary = f"Crawled {len(results)} pages from {urlparse(start_url).netloc}:\n\n"
+        
+        for i, result in enumerate(results, 1):
+            summary += f"{i}. {result['url']} (depth: {result['depth']}, {result['content_length']} chars)\n"
+            summary += f"   Summary: {result['summary'][:100]}...\n\n"
+        
+        total_chars = sum(r['content_length'] for r in results)
+        summary += f"\nTotal content: {total_chars:,} characters across {len(results)} pages"
+        
+        return SwaigFunctionResult(summary)
+    
+    def _extract_structured_handler(self, args: Dict[str, Any], raw_data: Dict[str, Any]) -> SwaigFunctionResult:
+        """Handle structured data extraction."""
+        url = args.get('url', '').strip()
+        
+        if not url:
+            return SwaigFunctionResult("Please provide a URL")
+        
+        # Use configured selectors from params
+        selectors = self.params.get('selectors', {})
+        if not selectors:
+            return SwaigFunctionResult("No selectors configured for structured data extraction")
+        
+        # Fetch the page
+        response = self._fetch_url(url)
+        if not response:
+            return SwaigFunctionResult(f"Failed to fetch {url}")
+        
+        # Extract structured data
+        result = self._structured_extract(response, selectors)
+        
+        if 'error' in result:
+            return SwaigFunctionResult(f"Error extracting data: {result['error']}")
+        
+        # Format the response
+        output = f"Extracted data from {url}:\n\n"
+        output += f"Title: {result.get('title', 'N/A')}\n\n"
+        
+        if result.get('data'):
+            output += "Data:\n"
+            for field, value in result['data'].items():
+                output += f"- {field}: {value}\n"
+        else:
+            output += "No data extracted with provided selectors"
+        
+        return SwaigFunctionResult(output)
+    
+    def get_hints(self) -> List[str]:
+        """Return speech recognition hints for this skill."""
+        return [
+            "scrape", "crawl", "extract", "web page", "website",
+            "get content from", "fetch data from", "spider"
+        ]
+    
+    def cleanup(self) -> None:
+        """Clean up resources when skill is unloaded."""
+        if hasattr(self, 'session'):
+            self.session.close()
+        if hasattr(self, 'cache'):
+            self.cache.clear()
+        self.logger.info("Spider skill cleaned up")