spiderforce4ai 2.1__py3-none-any.whl → 2.4__py3-none-any.whl

spiderforce4ai/__init__.py

@@ -1,10 +1,11 @@
 # spiderforce4ai/__init__.py
 
+from .post_extraction_agent import PostExtractionAgent, PostExtractionConfig, ExtractionTemplate
 import asyncio
 import aiohttp
 import json
 import logging
-from typing import List, Dict, Union, Optional, Tuple
+from typing import List, Dict, Union, Optional, Tuple, Callable, Any
 from dataclasses import dataclass, asdict
 from urllib.parse import urljoin, urlparse
 from pathlib import Path
@@ -23,75 +24,55 @@ from multiprocessing import Pool
 console = Console()
 
 def extract_metadata_headers(markdown: str, url: str = '') -> str:
-    """Extract metadata and headers from markdown content with enhanced SEO formatting."""
+    """Extract metadata and headers from markdown content."""
     lines = markdown.split('\n')
-    extracted = []
-    in_metadata = False
-    metadata = {
-        'title': '',
-        'description': '',
-        'canonical_url': '',
-        'language': ''
-    }
-    first_paragraph = ''
+    metadata = {}
+    headers = []
     
-    # First pass - collect metadata and first paragraph
-    for i, line in enumerate(lines):
-        # Check for metadata block boundaries
-        if line.strip() == '---':
-            if not in_metadata:
-                in_metadata = True
-                continue
-            else:
-                in_metadata = False
-                break
+    def parse_metadata_line(line):
+        """Parse a single metadata line correctly."""
+        first_colon = line.find(':')
+        if first_colon == -1:
+            return None, None
+            
+        key = line[:first_colon].strip()
+        value = line[first_colon + 1:].strip()
         
-        # Extract metadata within the block
-        if in_metadata:
-            if ':' in line:
-                key, value = [part.strip() for part in line.split(':', 1)]
-                key = key.lower()
-                
-                # Handle multi-line values
-                if value.startswith('>'):
-                    value = value[1:].strip()
-                    j = i + 1
-                    while j < len(lines) and lines[j].strip() and not lines[j].strip() == '---':
-                        value += ' ' + lines[j].strip()
-                        j += 1
-                
-                if key == 'title':
-                    metadata['title'] = value
-                elif key in ['description', 'meta_description', 'og:description', 'meta-description']:
-                    metadata['description'] = value
-                elif key in ['canonical_url', 'canonical']:
-                    metadata['canonical_url'] = value
-                elif key in ['language', 'lang']:
-                    metadata['language'] = value
-        elif not in_metadata and not first_paragraph and line.strip() and not line.startswith('#'):
-            first_paragraph = line.strip()
-    
-    # Use first paragraph as fallback description if none found
-    if not metadata['description'] and first_paragraph:
-        metadata['description'] = first_paragraph[:160] + ('...' if len(first_paragraph) > 160 else '')
-    
-    # Add formatted metadata section
-    extracted.append(f"URL: {url}")
-    extracted.append(f"Title: {metadata['title'] or url.split('/')[-2].replace('-', ' ').title()}")
-    extracted.append(f"Description: {metadata['description']}")
-    extracted.append(f"CanonicalUrl: {metadata['canonical_url'] or url}")
-    extracted.append(f"Language: {metadata['language'] or 'en'}")
-    extracted.append("")  # Empty line after metadata
+        # Handle the case where value starts with "URL:" - this means it's a missing description
+        if value.startswith('URL:'):
+            return key, ''
+            
+        return key, value
     
-    # Second pass - process headers
+    # Process each line
     for line in lines:
-        if line.strip().startswith('#'):
+        line = line.strip()
+        if not line:
+            continue
+            
+        # Check if it's a metadata line (contains : but isn't a header)
+        if ':' in line and not line.startswith('#'):
+            key, value = parse_metadata_line(line)
+            if key:
+                metadata[key] = value
+        # Check if it's a header
+        elif line.startswith('#'):
             level = len(line) - len(line.lstrip('#'))
             text = line.lstrip('#').strip()
             if 1 <= level <= 6:
-                extracted.append(f"H{level}: {text}")
+                headers.append(f"H{level}: {text}")
     
-    return '\n'.join(extracted)
+    # Construct output
+    output = []
+    output.append(f"URL: {url}")
+    output.append(f"Title: {metadata.get('Title', url.split('/')[-2].replace('-', ' ').title())}")
+    output.append(f"Description: {metadata.get('Description', '')}")
+    output.append(f"CanonicalUrl: {metadata.get('CanonicalUrl', url)}")
+    output.append(f"Language: {metadata.get('Language', 'en')}")
+    output.append("")  # Empty line
+    output.extend(headers)
+    
+    return '\n'.join(output)
 
 def slugify(url: str) -> str:
     """Convert URL to a valid filename."""
@@ -111,6 +92,7 @@ class CrawlResult:
     error: Optional[str] = None
     timestamp: str = None
     config: Dict = None
+    extraction_result: Optional[Dict] = None  # Store post-extraction results
     
     def __post_init__(self):
         if not self.timestamp:
@@ -131,9 +113,14 @@ class CrawlConfig:
     webhook_headers: Optional[Dict[str, str]] = None  # Optional webhook headers
     webhook_payload_template: Optional[str] = None  # Optional custom webhook payload template
     save_reports: bool = False  # Whether to save crawl reports
-    report_file: Optional[Path] = None  # Optional report file location (used only if save_reports is True)
-    combine_to_one_markdown: Optional[str] = None  # 'full' or 'metadata_headers' to combine all pages into one file
-    combined_markdown_file: Optional[Path] = None  # Optional path for combined markdown file
+    report_file: Optional[Path] = None  # Optional report file location
+    combine_to_one_markdown: Optional[str] = None  # 'full' or 'metadata_headers'
+    combined_markdown_file: Optional[Path] = None  # Optional path for combined file
+    
+    # Post-extraction settings
+    post_extraction_agent: Optional[Dict[str, Any]] = None  # LLM configuration
+    post_extraction_agent_save_to_file: Optional[str] = None  # Extraction output file
+    post_agent_transformer_function: Optional[Callable] = None  # Custom transformer
 
     def __post_init__(self):
         # Initialize empty lists/dicts for None values
@@ -161,6 +148,15 @@ class CrawlConfig:
             # Create or clear the combined file
             self.combined_markdown_file.write_text('')
 
+        # Validate post-extraction agent configuration if provided
+        if self.post_extraction_agent:
+            if "messages" not in self.post_extraction_agent:
+                raise ValueError("Post-extraction agent configuration must include 'messages'")
+            if "model" not in self.post_extraction_agent:
+                raise ValueError("Post-extraction agent configuration must include 'model'")
+            if "api_key" not in self.post_extraction_agent:
+                raise ValueError("Post-extraction agent configuration must include 'api_key'")
+
     def to_dict(self) -> Dict:
         """Convert config to dictionary for API requests."""
         payload = {}
@@ -172,52 +168,120 @@ class CrawlConfig:
         if self.remove_selectors_regex:
             payload["remove_selectors_regex"] = self.remove_selectors_regex
         return payload
-
-
+    
 def _send_webhook_sync(result: CrawlResult, config: CrawlConfig) -> None:
     """Synchronous version of webhook sender for parallel processing."""
     if not config.webhook_url:
         return
 
-    # Use custom payload template if provided, otherwise use default
-    if config.webhook_payload_template:
-        # Replace variables in the template
-        payload_str = config.webhook_payload_template.format(
-            url=result.url,
-            status=result.status,
-            markdown=result.markdown if result.status == "success" else None,
-            error=result.error if result.status == "failed" else None,
-            timestamp=result.timestamp,
-            config=config.to_dict()
+    try:
+        # Use custom payload template if provided, otherwise use default
+        if config.webhook_payload_template:
+            # Replace variables in the template
+            payload_str = config.webhook_payload_template.format(
+                url=result.url,
+                status=result.status,
+                markdown=result.markdown if result.status == "success" else None,
+                error=result.error if result.status == "failed" else None,
+                timestamp=result.timestamp,
+                config=config.to_dict(),
+                extraction_result=result.extraction_result if result.extraction_result else None
+            )
+            payload = json.loads(payload_str)  # Parse the formatted JSON string
+        else:
+            # Use default payload format
+            payload = {
+                "url": result.url,
+                "status": result.status,
+                "markdown": result.markdown if result.status == "success" else None,
+                "error": result.error if result.status == "failed" else None,
+                "timestamp": result.timestamp,
+                "config": config.to_dict(),
+                "extraction_result": result.extraction_result if result.extraction_result else None
+            }
+
+        response = requests.post(
+            config.webhook_url,
+            json=payload,
+            headers=config.webhook_headers,
+            timeout=config.webhook_timeout
         )
-        payload = json.loads(payload_str)  # Parse the formatted JSON string
-    else:
-        # Use default payload format
+        response.raise_for_status()
+    except Exception as e:
+        console.print(f"[yellow]Warning: Failed to send webhook for {result.url}: {str(e)}[/yellow]")
+
+async def _send_webhook_async(result: CrawlResult, config: CrawlConfig):
+    """Asynchronous webhook sender."""
+    if not config.webhook_url:
+        return
+
+    try:
+        # Prepare payload similar to sync version
         payload = {
             "url": result.url,
             "status": result.status,
             "markdown": result.markdown if result.status == "success" else None,
             "error": result.error if result.status == "failed" else None,
             "timestamp": result.timestamp,
-            "config": config.to_dict()
+            "config": config.to_dict(),
+            "extraction_result": result.extraction_result if result.extraction_result else None
         }
 
+        async with httpx.AsyncClient() as client:
+            response = await client.post(
+                config.webhook_url,
+                json=payload,
+                headers=config.webhook_headers,
+                timeout=config.webhook_timeout
+            )
+            response.raise_for_status()
+    except Exception as e:
+        console.print(f"[yellow]Warning: Failed to send webhook for {result.url}: {str(e)}[/yellow]")
+
+async def _save_markdown_async(url: str, markdown: str, config: CrawlConfig):
+    """Save markdown content to file and/or append to combined file asynchronously."""
     try:
-        response = requests.post(
-            config.webhook_url,
-            json=payload,
-            headers=config.webhook_headers,
-            timeout=config.webhook_timeout
-        )
-        response.raise_for_status()
+        # Save individual file if not combining or if combining in full mode
+        if not config.combine_to_one_markdown or config.combine_to_one_markdown == 'full':
+            filename = f"{slugify(url)}.md"
+            filepath = config.output_dir / filename
+            async with aiofiles.open(filepath, 'w', encoding='utf-8') as f:
+                await f.write(markdown)
+
+        # Handle combined markdown file
+        if config.combine_to_one_markdown:
+            content = markdown if config.combine_to_one_markdown == 'full' else extract_metadata_headers(markdown, url)
+            combined_content = f"\n----PAGE----\n{url}\n\n{content}\n----PAGE END----\n"
+            
+            async with aiofiles.open(config.combined_markdown_file, 'a', encoding='utf-8') as f:
+                await f.write(combined_content)
+    except Exception as e:
+        console.print(f"[red]Error saving markdown for {url}: {str(e)}[/red]")
+
+def _save_markdown_sync(url: str, markdown: str, config: CrawlConfig) -> None:
+    """Synchronous version of markdown saver for parallel processing."""
+    try:
+        # Save individual file if not combining or if combining in full mode
+        if not config.combine_to_one_markdown or config.combine_to_one_markdown == 'full':
+            filepath = config.output_dir / f"{slugify(url)}.md"
+            with open(filepath, 'w', encoding='utf-8') as f:
+                f.write(markdown)
+        
+        # Handle combined markdown file
+        if config.combine_to_one_markdown:
+            content = markdown if config.combine_to_one_markdown == 'full' else extract_metadata_headers(markdown, url)
+            combined_content = f"\n----PAGE----\n{url}\n\n{content}\n----PAGE END----\n"
+            
+            with open(config.combined_markdown_file, 'a', encoding='utf-8') as f:
+                f.write(combined_content)
     except Exception as e:
-        print(f"Warning: Failed to send webhook for {result.url}: {str(e)}")
+        console.print(f"[red]Error saving markdown for {url}: {str(e)}[/red]")
 
-# Module level function for multiprocessing
 def _process_url_parallel(args: Tuple[str, str, CrawlConfig]) -> CrawlResult:
     """Process a single URL for parallel processing."""
     url, base_url, config = args
     try:
+        # Make the conversion request
         endpoint = f"{base_url}/convert"
         payload = {
             "url": url,
@@ -232,7 +296,6 @@ def _process_url_parallel(args: Tuple[str, str, CrawlConfig]) -> CrawlResult:
                 error=f"HTTP {response.status_code}: {response.text}",
                 config=config.to_dict()
             )
-            # Send webhook for failed result
             _send_webhook_sync(result, config)
             return result
         
@@ -240,19 +303,7 @@ def _process_url_parallel(args: Tuple[str, str, CrawlConfig]) -> CrawlResult:
         
         # Save markdown if output directory is configured
         if config.output_dir:
-            # Save individual file if not combining or if combining in full mode
-            if not config.combine_to_one_markdown or config.combine_to_one_markdown == 'full':
-                filepath = config.output_dir / f"{slugify(url)}.md"
-                with open(filepath, 'w', encoding='utf-8') as f:
-                    f.write(markdown)
-            
-            # Handle combined markdown file
-            if config.combine_to_one_markdown:
-                content = markdown if config.combine_to_one_markdown == 'full' else extract_metadata_headers(markdown, url)
-                combined_content = f"\n----PAGE----\n{url}\n\n{content}\n----PAGE END----\n"
-                
-                with open(config.combined_markdown_file, 'a', encoding='utf-8') as f:
-                    f.write(combined_content)
+            _save_markdown_sync(url, markdown, config)
         
         result = CrawlResult(
             url=url,
@@ -261,6 +312,28 @@ def _process_url_parallel(args: Tuple[str, str, CrawlConfig]) -> CrawlResult:
             config=config.to_dict()
         )
         
+        # Handle post-extraction if configured
+        if config.post_extraction_agent:
+            try:
+                post_config = PostExtractionConfig(
+                    model=config.post_extraction_agent["model"],
+                    messages=config.post_extraction_agent["messages"],
+                    api_key=config.post_extraction_agent["api_key"],
+                    max_tokens=config.post_extraction_agent.get("max_tokens", 1000),
+                    temperature=config.post_extraction_agent.get("temperature", 0.7),
+                    base_url=config.post_extraction_agent.get("base_url"),
+                    combine_output=bool(config.post_extraction_agent_save_to_file),
+                    output_file=config.post_extraction_agent_save_to_file,
+                    custom_transform_function=config.post_agent_transformer_function
+                )
+                
+                agent = PostExtractionAgent(post_config)
+                extraction_result = asyncio.run(agent.process_content(url, markdown))
+                if extraction_result:
+                    result.extraction_result = extraction_result
+            except Exception as e:
+                console.print(f"[red]Error in post-extraction processing for {url}: {str(e)}[/red]")
+        
         # Send webhook for successful result
         _send_webhook_sync(result, config)
         
@@ -281,6 +354,60 @@ def _process_url_parallel(args: Tuple[str, str, CrawlConfig]) -> CrawlResult:
         _send_webhook_sync(result, config)
         return result
 
+async def _save_report_async(results: List[CrawlResult], config: CrawlConfig, retry_stats: Dict = None):
+    """Save crawl report to JSON file asynchronously."""
+    if not config.report_file:
+        return
+
+    # Separate successful and failed results
+    successful_results = [r for r in results if r.status == "success"]
+    failed_results = [r for r in results if r.status == "failed"]
+
+    report = {
+        "timestamp": datetime.now().isoformat(),
+        "config": config.to_dict(),
+        "results": {
+            "successful": [asdict(r) for r in successful_results],
+            "failed": [asdict(r) for r in failed_results]
+        },
+        "summary": {
+            "total": len(results),
+            "successful": len(successful_results),
+            "failed": len(failed_results),
+            "retry_info": retry_stats or {}
+        }
+    }
+
+    async with aiofiles.open(config.report_file, 'w', encoding='utf-8') as f:
+        await f.write(json.dumps(report, indent=2))
+
+def _save_report_sync(results: List[CrawlResult], config: CrawlConfig, retry_stats: Dict = None) -> None:
+    """Synchronous version of report saver."""
+    if not config.report_file:
+        return
+
+    # Create report similar to async version
+    successful_results = [r for r in results if r.status == "success"]
+    failed_results = [r for r in results if r.status == "failed"]
+    
+    report = {
+        "timestamp": datetime.now().isoformat(),
+        "config": config.to_dict(),
+        "results": {
+            "successful": [asdict(r) for r in successful_results],
+            "failed": [asdict(r) for r in failed_results]
+        },
+        "summary": {
+            "total": len(results),
+            "successful": len(successful_results),
+            "failed": len(failed_results),
+            "retry_info": retry_stats or {}
+        }
+    }
+
+    with open(config.report_file, 'w', encoding='utf-8') as f:
+        json.dump(report, f, indent=2)
+
 class SpiderForce4AI:
     """Main class for interacting with SpiderForce4AI service."""
 
@@ -289,6 +416,7 @@ class SpiderForce4AI:
         self.session = None
         self._executor = ThreadPoolExecutor()
         self.crawl_results: List[CrawlResult] = []
+        self._retry_stats = {}
 
     async def _ensure_session(self):
         """Ensure aiohttp session exists."""
@@ -300,215 +428,6 @@ class SpiderForce4AI:
         if self.session and not self.session.closed:
             await self.session.close()
 
-    async def _save_markdown(self, url: str, markdown: str, output_dir: Path):
-        """Save markdown content to file and/or append to combined file."""
-        # Save individual file if not combining or if combining in full mode
-        if not self.config.combine_to_one_markdown or self.config.combine_to_one_markdown == 'full':
-            filename = f"{slugify(url)}.md"
-            filepath = output_dir / filename
-            async with aiofiles.open(filepath, 'w', encoding='utf-8') as f:
-                await f.write(markdown)
-
-        # Handle combined markdown file
-        if self.config.combine_to_one_markdown:
-            content = markdown if config.combine_to_one_markdown == 'full' else extract_metadata_headers(markdown, url)
-            combined_content = f"\n----PAGE----\n{url}\n\n{content}\n----PAGE END----\n"
-            
-            async with aiofiles.open(self.config.combined_markdown_file, 'a', encoding='utf-8') as f:
-                await f.write(combined_content)
-
-
-
-    def crawl_sitemap_server_parallel(self, sitemap_url: str, config: CrawlConfig) -> List[CrawlResult]:
-        """
-        Crawl sitemap URLs using server-side parallel processing.
-        """
-        print(f"Fetching sitemap from {sitemap_url}...")
-        
-        # Fetch sitemap
-        try:
-            response = requests.get(sitemap_url, timeout=config.timeout)
-            response.raise_for_status()
-            sitemap_text = response.text
-        except Exception as e:
-            print(f"Error fetching sitemap: {str(e)}")
-            raise
-
-        # Parse sitemap
-        try:
-            root = ET.fromstring(sitemap_text)
-            namespace = {'ns': root.tag.split('}')[0].strip('{')}
-            urls = [loc.text for loc in root.findall('.//ns:loc', namespace)]
-            print(f"Found {len(urls)} URLs in sitemap")
-        except Exception as e:
-            print(f"Error parsing sitemap: {str(e)}")
-            raise
-
-        # Process URLs using server-side parallel endpoint
-        return self.crawl_urls_server_parallel(urls, config)
-
-
-    def crawl_urls_server_parallel(self, urls: List[str], config: CrawlConfig) -> List[CrawlResult]:
-        """
-        Crawl multiple URLs using server-side parallel processing.
-        This uses the /convert_parallel endpoint which handles parallelization on the server.
-        """
-        print(f"Sending {len(urls)} URLs for parallel processing...")
-        
-        try:
-            endpoint = f"{self.base_url}/convert_parallel"
-            
-            # Prepare payload
-            payload = {
-                "urls": urls,
-                **config.to_dict()
-            }
-            
-            # Send request
-            response = requests.post(
-                endpoint, 
-                json=payload, 
-                timeout=config.timeout
-            )
-            response.raise_for_status()
-            
-            # Process results
-            results = []
-            server_results = response.json()  # Assuming server returns JSON array of results
-            
-            for url_result in server_results:
-                result = CrawlResult(
-                    url=url_result["url"],
-                    status=url_result.get("status", "failed"),
-                    markdown=url_result.get("markdown"),
-                    error=url_result.get("error"),
-                    config=config.to_dict()
-                )
-                
-                # Save markdown if successful and output dir is configured
-                if result.status == "success" and config.output_dir and result.markdown:
-                    filepath = config.output_dir / f"{slugify(result.url)}.md"
-                    with open(filepath, 'w', encoding='utf-8') as f:
-                        f.write(result.markdown)
-                
-                # Send webhook if configured
-                if config.webhook_url:
-                    _send_webhook_sync(result, config)
-                    
-                results.append(result)
-            
-            # Calculate statistics
-            successful = len([r for r in results if r.status == "success"])
-            failed = len([r for r in results if r.status == "failed"])
-
-            # Print summary
-            print(f"\nParallel processing completed:")
-            print(f"✓ Successful: {successful}")
-            print(f"✗ Failed: {failed}")
-
-            # Save report if enabled
-            if config.save_reports and config.report_file:
-                self._retry_stats = {
-                    "initial_failures": failed,
-                    "failure_ratio": (failed / len(urls)) * 100,
-                    "retry_successful": 0,  # No retries in server parallel mode
-                    "retry_failed": failed
-                }
-                self._save_report_sync(results, config)
-                console.print(f"📊 Report saved to: {config.report_file}")
-            
-            return results
-                
-        except Exception as e:
-            print(f"Error during parallel processing: {str(e)}")
-            # Create failed results for all URLs
-            return [
-                CrawlResult(
-                    url=url,
-                    status="failed",
-                    error=str(e),
-                    config=config.to_dict()
-                ) for url in urls
-            ]
-
-
-    async def _send_webhook(self, result: CrawlResult, config: CrawlConfig):
-        """Send webhook with crawl results."""
-        if not config.webhook_url:
-            return
-
-        payload = {
-            "url": result.url,
-            "status": result.status,
-            "markdown": result.markdown if result.status == "success" else None,
-            "error": result.error if result.status == "failed" else None,
-            "timestamp": result.timestamp,
-            "config": config.to_dict()
-        }
-
-        try:
-            async with httpx.AsyncClient() as client:
-                response = await client.post(
-                    config.webhook_url,
-                    json=payload,
-                    timeout=config.webhook_timeout
-                )
-                response.raise_for_status()
-        except Exception as e:
-            console.print(f"[yellow]Warning: Failed to send webhook for {result.url}: {str(e)}[/yellow]")
-
-    def _save_report_sync(self, results: List[CrawlResult], config: CrawlConfig) -> None:
-        """Save crawl report synchronously."""
-        # Separate successful and failed results
-        successful_results = [r for r in results if r.status == "success"]
-        failed_results = [r for r in results if r.status == "failed"]
-        
-        # Create report with only final state
-        report = {
-            "timestamp": datetime.now().isoformat(),
-            "config": config.to_dict(),
-            "results": {
-                "successful": [asdict(r) for r in successful_results],
-                "failed": [asdict(r) for r in failed_results]  # Only truly failed URLs after retries
-            },
-            "summary": {
-                "total": len(results),
-                "successful": len(successful_results),
-                "failed": len(failed_results),
-                "retry_info": getattr(self, '_retry_stats', {})  # Include retry statistics if available
-            }
-        }
-
-        with open(config.report_file, 'w', encoding='utf-8') as f:
-            json.dump(report, f, indent=2)
-
-    async def _save_report(self, config: CrawlConfig):
-        """Save crawl report to JSON file."""
-        if not config.report_file:
-            return
-
-        # Separate successful and failed results
-        successful_results = [r for r in self.crawl_results if r.status == "success"]
-        failed_results = [r for r in self.crawl_results if r.status == "failed"]
-
-        report = {
-            "timestamp": datetime.now().isoformat(),
-            "config": config.to_dict(),
-            "results": {
-                "successful": [asdict(r) for r in successful_results],
-                "failed": [asdict(r) for r in failed_results]  # Only truly failed URLs after retries
-            },
-            "summary": {
-                "total": len(self.crawl_results),
-                "successful": len(successful_results),
-                "failed": len(failed_results),
-                "retry_info": getattr(self, '_retry_stats', {})  # Include retry statistics if available
-            }
-        }
-
-        async with aiofiles.open(config.report_file, 'w', encoding='utf-8') as f:
-            await f.write(json.dumps(report, indent=2))
-
     async def crawl_url_async(self, url: str, config: CrawlConfig) -> CrawlResult:
         """Crawl a single URL asynchronously."""
         await self._ensure_session()
@@ -539,9 +458,31 @@ class SpiderForce4AI:
                     )
 
                     if config.output_dir:
-                        await self._save_markdown(url, markdown, config.output_dir)
+                        await _save_markdown_async(url, markdown, config)
                     
-                    await self._send_webhook(result, config)
+                    # Handle post-extraction if configured
+                    if config.post_extraction_agent and result.status == "success":
+                        try:
+                            post_config = PostExtractionConfig(
+                                model=config.post_extraction_agent["model"],
+                                messages=config.post_extraction_agent["messages"],
+                                api_key=config.post_extraction_agent["api_key"],
+                                max_tokens=config.post_extraction_agent.get("max_tokens", 1000),
+                                temperature=config.post_extraction_agent.get("temperature", 0.7),
+                                base_url=config.post_extraction_agent.get("base_url"),
+                                combine_output=bool(config.post_extraction_agent_save_to_file),
+                                output_file=config.post_extraction_agent_save_to_file,
+                                custom_transform_function=config.post_agent_transformer_function
+                            )
+                            
+                            agent = PostExtractionAgent(post_config)
+                            extraction_result = await agent.process_content(url, markdown)
+                            if extraction_result:
+                                result.extraction_result = extraction_result
+                        except Exception as e:
+                            console.print(f"[red]Error in post-extraction processing for {url}: {str(e)}[/red]")
+                    
+                    await _send_webhook_async(result, config)
                 
                 self.crawl_results.append(result)
                 return result
@@ -561,18 +502,18 @@ class SpiderForce4AI:
         return asyncio.run(self.crawl_url_async(url, config))
 
     async def _retry_failed_urls(self, failed_results: List[CrawlResult], config: CrawlConfig, progress=None) -> List[CrawlResult]:
-        """Retry failed URLs once."""
+        """Retry failed URLs with optional progress tracking."""
         if not failed_results:
             return []
 
         failed_count = len(failed_results)
-        total_count = len([r for r in self.crawl_results])
+        total_count = len(self.crawl_results)
         failure_ratio = (failed_count / total_count) * 100
         
         console.print(f"\n[yellow]Retrying failed URLs: {failed_count} ({failure_ratio:.1f}% failed)[/yellow]")
         retry_results = []
         
-        # Create a new progress bar if one wasn't provided
+        # Create or use provided progress bar
         should_close_progress = progress is None
         if progress is None:
             progress = Progress(
@@ -616,6 +557,7 @@ class SpiderForce4AI:
     async def crawl_urls_async(self, urls: List[str], config: CrawlConfig) -> List[CrawlResult]:
         """Crawl multiple URLs asynchronously with progress bar."""
         await self._ensure_session()
+        post_extraction_results = {}
         
         with Progress(
             SpinnerColumn(),
@@ -624,52 +566,60 @@ class SpiderForce4AI:
             TaskProgressColumn(),
             console=console
         ) as progress:
-            task = progress.add_task("[cyan]Crawling URLs...", total=len(urls))
+            crawl_task = progress.add_task("[cyan]Crawling URLs...", total=len(urls))
             
             async def crawl_with_progress(url):
                 result = await self.crawl_url_async(url, config)
-                progress.update(task, advance=1, description=f"[cyan]Crawled: {url}")
+                progress.update(crawl_task, advance=1, description=f"[cyan]Crawled: {url}")
                 return result
 
+            # Set up concurrency control
             semaphore = asyncio.Semaphore(config.max_concurrent_requests)
+            
             async def crawl_with_semaphore(url):
                 async with semaphore:
                     result = await crawl_with_progress(url)
                     await asyncio.sleep(config.request_delay)
                     return result
 
+            # Perform initial crawl
             initial_results = await asyncio.gather(*[crawl_with_semaphore(url) for url in urls])
             
-            # Identify failed URLs
+            # Handle failed URLs
             failed_results = [r for r in initial_results if r.status == "failed"]
-            
-            # Calculate initial failure ratio
             initial_failed = len(failed_results)
             total_urls = len(urls)
             failure_ratio = (initial_failed / total_urls) * 100
 
             # Retry failed URLs if ratio is acceptable
-            if failed_results:
-                if failure_ratio > 20:
-                    console.print(f"\n[red]Failure ratio too high ({failure_ratio:.1f}%) - aborting retry due to possible server overload[/red]")
-                    results = initial_results
-                else:
-                    retry_results = await self._retry_failed_urls(failed_results, config, progress)
-                    # Update results list by replacing failed results with successful retries
-                    results = initial_results.copy()
-                    for retry_result in retry_results:
-                        for i, result in enumerate(results):
-                            if result.url == retry_result.url:
-                                results[i] = retry_result
-                                break
-            else:
-                results = initial_results
+            results = initial_results
+            retry_successful = 0
             
-            # Calculate final statistics before saving report
+            if failed_results and failure_ratio <= 20:
+                retry_results = await self._retry_failed_urls(failed_results, config, progress)
+                retry_successful = len([r for r in retry_results if r.status == "success"])
+                
+                # Update results list
+                for retry_result in retry_results:
+                    for i, result in enumerate(results):
+                        if result.url == retry_result.url:
+                            results[i] = retry_result
+                            break
+
+            # Calculate final statistics
             final_successful = len([r for r in results if r.status == "success"])
             final_failed = len([r for r in results if r.status == "failed"])
             
-            # Print detailed summary
+            # Update retry stats
+            self._retry_stats = {
+                "initial_failures": initial_failed,
+                "failure_ratio": failure_ratio,
+                "retry_successful": retry_successful if initial_failed > 0 else 0,
+                "retry_failed": final_failed,
+                "post_extraction_successful": len(post_extraction_results) if post_extraction_results else 0
+            }
+
+            # Print summary
             console.print(f"\n[green]Crawling Summary:[/green]")
             console.print(f"Total URLs processed: {total_urls}")
             console.print(f"Initial failures: {initial_failed} ({failure_ratio:.1f}%)")
@@ -678,18 +628,11 @@ class SpiderForce4AI:
             console.print(f"  ✗ Failed: {final_failed}")
             
             if initial_failed > 0:
-                retry_successful = initial_failed - final_failed
                 console.print(f"Retry success rate: {retry_successful}/{initial_failed} ({(retry_successful/initial_failed)*100:.1f}%)")
             
-            # Save final report after all retries are complete
+            # Save final report
             if config.save_reports:
-                self._retry_stats = {
-                    "initial_failures": initial_failed,
-                    "failure_ratio": failure_ratio,
-                    "retry_successful": retry_successful if initial_failed > 0 else 0,
-                    "retry_failed": final_failed
-                }
-                await self._save_report(config)
+                await _save_report_async(results, config, self._retry_stats)
                 console.print(f"📊 Report saved to: {config.report_file}")
             
             return results
@@ -726,32 +669,21 @@ class SpiderForce4AI:
         return asyncio.run(self.crawl_sitemap_async(sitemap_url, config))
 
     def crawl_sitemap_parallel(self, sitemap_url: str, config: CrawlConfig) -> List[CrawlResult]:
-        """Crawl sitemap URLs in parallel using multiprocessing (no asyncio required)."""
-        print(f"Fetching sitemap from {sitemap_url}...")
-        
-        # Fetch sitemap
+        """Crawl sitemap URLs in parallel using multiprocessing."""
+        # Fetch and parse sitemap
         try:
             response = requests.get(sitemap_url, timeout=config.timeout)
             response.raise_for_status()
-            sitemap_text = response.text
-        except Exception as e:
-            print(f"Error fetching sitemap: {str(e)}")
-            raise
-
-        # Parse sitemap
-        try:
-            root = ET.fromstring(sitemap_text)
+            root = ET.fromstring(response.text)
             namespace = {'ns': root.tag.split('}')[0].strip('{')}
             urls = [loc.text for loc in root.findall('.//ns:loc', namespace)]
-            print(f"Found {len(urls)} URLs in sitemap")
+            console.print(f"[green]Found {len(urls)} URLs in sitemap[/green]")
         except Exception as e:
-            print(f"Error parsing sitemap: {str(e)}")
+            console.print(f"[red]Error processing sitemap: {str(e)}[/red]")
             raise
 
-        # Prepare arguments for parallel processing
+        # Process URLs in parallel
         process_args = [(url, self.base_url, config) for url in urls]
-
-        # Create process pool and execute crawls
         results = []
 
         with Pool(processes=config.max_concurrent_requests) as pool:
@@ -762,81 +694,186 @@ class SpiderForce4AI:
                 TextColumn("[progress.percentage]{task.percentage:>3.0f}%"),
                 TextColumn("({task.completed}/{task.total})"),
             ) as progress:
-                task = progress.add_task("Crawling URLs...", total=len(urls))
+                task = progress.add_task("[cyan]Crawling URLs...", total=len(urls))
                 
                 for result in pool.imap_unordered(_process_url_parallel, process_args):
                     results.append(result)
                     progress.update(task, advance=1)
                     status = "✓" if result.status == "success" else "✗"
-                    progress.description = f"Last: {status} {result.url}"
+                    progress.description = f"[cyan]Last: {status} {result.url}"
 
-        # Calculate initial failure statistics
+        # Calculate statistics and handle retries
         failed_results = [r for r in results if r.status == "failed"]
         initial_failed = len(failed_results)
-        total_urls = len(urls)
-        failure_ratio = (initial_failed / total_urls) * 100
+        failure_ratio = (initial_failed / len(urls)) * 100
+        retry_successful = 0
 
-        # Retry failed URLs if ratio is acceptable
-        if failed_results:
-            if failure_ratio > 20:
-                console.print(f"\n[red]Failure ratio too high ({failure_ratio:.1f}%) - aborting retry due to possible server overload[/red]")
-            else:
-                failed_count = len(failed_results)
-                failure_ratio = (failed_count / total_urls) * 100
-                console.print(f"\n[yellow]Retrying failed URLs: {failed_count} ({failure_ratio:.1f}% failed)[/yellow]")
-                for result in failed_results:
-                    new_result = _process_url_parallel((result.url, self.base_url, config))
-                    
-                    # Save markdown and trigger webhook for successful retries
-                    if new_result.status == "success":
-                        console.print(f"[green]✓ Retry successful: {result.url}[/green]")
-                        # Save markdown if output directory is configured
-                        if config.output_dir and new_result.markdown:
-                            filepath = config.output_dir / f"{slugify(new_result.url)}.md"
-                            with open(filepath, 'w', encoding='utf-8') as f:
-                                f.write(new_result.markdown)
-                        # Send webhook for successful retry
-                        _send_webhook_sync(new_result, config)
-                    else:
-                        console.print(f"[red]✗ Retry failed: {result.url} - {new_result.error}[/red]")
-                        # Send webhook for failed retry
-                        _send_webhook_sync(new_result, config)
-                    
-                    # Update results list
-                    for i, r in enumerate(results):
-                        if r.url == new_result.url:
-                            results[i] = new_result
-                            break
+        if failed_results and failure_ratio <= 20:
+            console.print(f"\n[yellow]Retrying {initial_failed} failed URLs...[/yellow]")
+            for result in failed_results:
+                new_result = _process_url_parallel((result.url, self.base_url, config))
+                if new_result.status == "success":
+                    retry_successful += 1
+                    console.print(f"[green]✓ Retry successful: {result.url}[/green]")
+                else:
+                    console.print(f"[red]✗ Retry failed: {result.url}[/red]")
+                
+                # Update results list
+                for i, r in enumerate(results):
+                    if r.url == new_result.url:
+                        results[i] = new_result
+                        break
 
         # Calculate final statistics
         final_successful = len([r for r in results if r.status == "success"])
         final_failed = len([r for r in results if r.status == "failed"])
 
-        # Print detailed summary
+        # Print summary
         console.print(f"\n[green]Crawling Summary:[/green]")
-        console.print(f"Total URLs processed: {total_urls}")
+        console.print(f"Total URLs processed: {len(urls)}")
         console.print(f"Initial failures: {initial_failed} ({failure_ratio:.1f}%)")
         console.print(f"Final results:")
         console.print(f"  ✓ Successful: {final_successful}")
         console.print(f"  ✗ Failed: {final_failed}")
-
+        
         if initial_failed > 0:
-            retry_successful = initial_failed - final_failed
             console.print(f"Retry success rate: {retry_successful}/{initial_failed} ({(retry_successful/initial_failed)*100:.1f}%)")
 
-        # Save final report after all retries are complete
+# Save report
         if config.save_reports:
             self._retry_stats = {
                 "initial_failures": initial_failed,
                 "failure_ratio": failure_ratio,
-                "retry_successful": retry_successful if initial_failed > 0 else 0,
+                "retry_successful": retry_successful,
                 "retry_failed": final_failed
             }
-            self._save_report_sync(results, config)
+            _save_report_sync(results, config, self._retry_stats)
             console.print(f"📊 Report saved to: {config.report_file}")
 
         return results
 
+    def crawl_urls_server_parallel(self, urls: List[str], config: CrawlConfig) -> List[CrawlResult]:
+        """
+        Crawl multiple URLs using server-side parallel processing.
+        This uses the /convert_parallel endpoint which handles parallelization on the server.
+        """
+        console.print(f"[cyan]Sending {len(urls)} URLs for parallel processing...[/cyan]")
+        
+        try:
+            endpoint = f"{self.base_url}/convert_parallel"
+            
+            # Prepare payload
+            payload = {
+                "urls": urls,
+                **config.to_dict()
+            }
+            
+            # Send request
+            response = requests.post(
+                endpoint, 
+                json=payload, 
+                timeout=config.timeout
+            )
+            response.raise_for_status()
+            
+            # Process results
+            results = []
+            server_results = response.json()
+            
+            for url_result in server_results:
+                result = CrawlResult(
+                    url=url_result["url"],
+                    status=url_result.get("status", "failed"),
+                    markdown=url_result.get("markdown"),
+                    error=url_result.get("error"),
+                    config=config.to_dict()
+                )
+                
+                # Save markdown if successful and output dir is configured
+                if result.status == "success" and config.output_dir and result.markdown:
+                    _save_markdown_sync(result.url, result.markdown, config)
+                
+                # Handle post-extraction if configured
+                if config.post_extraction_agent and result.status == "success":
+                    try:
+                        post_config = PostExtractionConfig(
+                            model=config.post_extraction_agent["model"],
+                            messages=config.post_extraction_agent["messages"],
+                            api_key=config.post_extraction_agent["api_key"],
+                            max_tokens=config.post_extraction_agent.get("max_tokens", 1000),
+                            temperature=config.post_extraction_agent.get("temperature", 0.7),
+                            base_url=config.post_extraction_agent.get("base_url"),
+                            combine_output=bool(config.post_extraction_agent_save_to_file),
+                            output_file=config.post_extraction_agent_save_to_file,
+                            custom_transform_function=config.post_agent_transformer_function
+                        )
+                        
+                        agent = PostExtractionAgent(post_config)
+                        extraction_result = asyncio.run(agent.process_content(result.url, result.markdown))
+                        if extraction_result:
+                            result.extraction_result = extraction_result
+                    except Exception as e:
+                        console.print(f"[red]Error in post-extraction processing for {result.url}: {str(e)}[/red]")
+                
+                # Send webhook if configured
+                _send_webhook_sync(result, config)
+                results.append(result)
+            
+            # Calculate statistics
+            successful = len([r for r in results if r.status == "success"])
+            failed = len([r for r in results if r.status == "failed"])
+            
+            # Print summary
+            console.print("\n[green]Parallel processing completed:[/green]")
+            console.print(f"✓ Successful: {successful}")
+            console.print(f"✗ Failed: {failed}")
+
+            # Save report if enabled
+            if config.save_reports:
+                self._retry_stats = {
+                    "initial_failures": failed,
+                    "failure_ratio": (failed / len(urls)) * 100,
+                    "retry_successful": 0,  # No retries in server parallel mode
+                    "retry_failed": failed
+                }
+                _save_report_sync(results, config, self._retry_stats)
+                console.print(f"📊 Report saved to: {config.report_file}")
+            
+            return results
+                
+        except Exception as e:
+            console.print(f"[red]Error during parallel processing: {str(e)}[/red]")
+            # Create failed results for all URLs
+            return [
+                CrawlResult(
+                    url=url,
+                    status="failed",
+                    error=str(e),
+                    config=config.to_dict()
+                ) for url in urls
+            ]
+
+    def crawl_sitemap_server_parallel(self, sitemap_url: str, config: CrawlConfig) -> List[CrawlResult]:
+        """
+        Crawl sitemap URLs using server-side parallel processing.
+        """
+        console.print(f"[cyan]Fetching sitemap from {sitemap_url}...[/cyan]")
+        
+        try:
+            response = requests.get(sitemap_url, timeout=config.timeout)
+            response.raise_for_status()
+            root = ET.fromstring(response.text)
+            namespace = {'ns': root.tag.split('}')[0].strip('{')}
+            urls = [loc.text for loc in root.findall('.//ns:loc', namespace)]
+            console.print(f"[green]Found {len(urls)} URLs in sitemap[/green]")
+            
+            # Process URLs using server-side parallel endpoint
+            return self.crawl_urls_server_parallel(urls, config)
+            
+        except Exception as e:
+            console.print(f"[red]Error processing sitemap: {str(e)}[/red]")
+            raise
+
     async def __aenter__(self):
         """Async context manager entry."""
         await self._ensure_session()
@@ -854,3 +891,7 @@ class SpiderForce4AI:
         """Sync context manager exit."""
         self._executor.shutdown(wait=True)
 
+# Version info
+#__version__ = "2.3.1"
+#__author__ = "Piotr Tamulewicz"
+#__email__ = "pt@petertam.pro"