feat: add VNC streaming support

Removes automatic page closure in take_screenshot and take_screenshot_naive methods
to prevent premature closure of pages that might still be needed in the calling context.
This allows for more flexible page lifecycle management by the caller.

BREAKING CHANGE: Page objects are no longer automatically closed after taking screenshots.
Callers must explicitly handle page closure when appropriate.

.gitignore

@@ -261,4 +261,7 @@ CLAUDE.md
 
 tests/**/test_site
 tests/**/reports
-tests/**/benchmark_reports
+tests/**/benchmark_reports
+
+docs/**/data
+.codecat/

CHANGELOG.md

@@ -5,6 +5,21 @@ All notable changes to Crawl4AI will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.6.2] - 2025-05-02
+
+### Added
+- New `RegexExtractionStrategy` for fast pattern-based extraction without requiring LLM
+  - Built-in patterns for emails, URLs, phone numbers, dates, and more
+  - Support for custom regex patterns
+  - `generate_pattern` utility for LLM-assisted pattern creation (one-time use)
+- Added `fit_html` as a top-level field in `CrawlResult` for optimized HTML extraction
+- Added support for network response body capture in network request tracking
+
+### Changed
+- Updated documentation for no-LLM extraction strategies
+- Enhanced API reference to include RegexExtractionStrategy examples and usage
+- Improved HTML preprocessing with optimized performance for extraction strategies
+
 ## [0.6.1] - 2025-04-24
 
 ### Added

Dockerfile

@@ -43,9 +43,17 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     libjpeg-dev \
     redis-server \
     supervisor \
-    && apt-get clean \ 
+    xvfb \
+    x11vnc \
+    fluxbox \
+    websockify \
+    && apt-get clean \
     && rm -rf /var/lib/apt/lists/*
 
+# Install noVNC for browser-based VNC access
+RUN git clone --depth 1 https://github.com/novnc/noVNC /opt/novnc \
+    && git clone --depth 1 https://github.com/novnc/websockify /opt/novnc/utils/websockify
+
 RUN apt-get update && apt-get install -y --no-install-recommends \
     libglib2.0-0 \
     libnss3 \

crawl4ai/__init__.py

@@ -23,7 +23,8 @@ from .extraction_strategy import (
     CosineStrategy,
     JsonCssExtractionStrategy,
     JsonXPathExtractionStrategy,
-    JsonLxmlExtractionStrategy
+    JsonLxmlExtractionStrategy,
+    RegexExtractionStrategy
 )
 from .chunking_strategy import ChunkingStrategy, RegexChunking
 from .markdown_generation_strategy import DefaultMarkdownGenerator
@@ -105,6 +106,7 @@ __all__ = [
     "JsonCssExtractionStrategy",
     "JsonXPathExtractionStrategy",
     "JsonLxmlExtractionStrategy",
+    "RegexExtractionStrategy",
     "ChunkingStrategy",
     "RegexChunking",
     "DefaultMarkdownGenerator",

crawl4ai/async_crawler_strategy.py

@@ -441,7 +441,7 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
         status_code = 200  # Default for local/raw HTML
         screenshot_data = None
 
-        if url.startswith(("http://", "https://")):
+        if url.startswith(("http://", "https://", "view-source:")):
             return await self._crawl_web(url, config)
 
         elif url.startswith("file://"):
@@ -571,6 +571,14 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
 
             async def handle_response_capture(response):
                 try:
+                    try:
+                        # body = await response.body()
+                        # json_body = await response.json()
+                        text_body = await response.text()
+                    except Exception as e:
+                        body = None
+                        # json_body = None
+                        # text_body = None
                     captured_requests.append({
                         "event_type": "response",
                         "url": response.url,
@@ -579,7 +587,12 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
                         "headers": dict(response.headers), # Convert Header dict
                         "from_service_worker": response.from_service_worker,
                         "request_timing": response.request.timing, # Detailed timing info
-                        "timestamp": time.time()
+                        "timestamp": time.time(),
+                        "body" : {
+                            # "raw": body,
+                            # "json": json_body,
+                            "text": text_body
+                        }
                     })
                 except Exception as e:
                     if self.logger:
@@ -771,7 +784,7 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
             except Error:
                 visibility_info = await self.check_visibility(page)
 
-                if self.config.verbose:
+                if self.browser_config.config.verbose:
                     self.logger.debug(
                         message="Body visibility info: {info}",
                         tag="DEBUG",
@@ -1454,8 +1467,8 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
             buffered = BytesIO()
             img.save(buffered, format="JPEG")
             return base64.b64encode(buffered.getvalue()).decode("utf-8")
-        finally:
-            await page.close()
+        # finally:
+        #     await page.close()
 
     async def take_screenshot_naive(self, page: Page) -> str:
         """
@@ -1488,8 +1501,8 @@ class AsyncPlaywrightCrawlerStrategy(AsyncCrawlerStrategy):
             buffered = BytesIO()
             img.save(buffered, format="JPEG")
             return base64.b64encode(buffered.getvalue()).decode("utf-8")
-        finally:
-            await page.close()
+        # finally:
+        #     await page.close()
 
     async def export_storage_state(self, path: str = None) -> dict:
         """

crawl4ai/async_webcrawler.py

@@ -503,6 +503,8 @@ class AsyncWebCrawler:
             tables = media.pop("tables", [])
             links = result.links.model_dump()
             metadata = result.metadata
+            
+        fit_html = preprocess_html_for_schema(html_content=html, text_threshold= 500, max_size= 300_000)
 
         ################################
         # Generate Markdown            #
@@ -519,7 +521,7 @@ class AsyncWebCrawler:
         html_source_selector = {
             "raw_html": lambda: html,  # The original raw HTML
             "cleaned_html": lambda: cleaned_html,  # The HTML after scraping strategy
-            "fit_html": lambda: preprocess_html_for_schema(html_content=html),  # Preprocessed raw HTML
+            "fit_html": lambda: fit_html,  # The HTML after preprocessing for schema
         }
 
         markdown_input_html = cleaned_html  # Default to cleaned_html
@@ -593,6 +595,7 @@ class AsyncWebCrawler:
             content = {
                 "markdown": markdown_result.raw_markdown,
                 "html": html,
+                "fit_html": fit_html,
                 "cleaned_html": cleaned_html,
                 "fit_markdown": markdown_result.fit_markdown,
             }.get(content_format, markdown_result.raw_markdown)
@@ -600,7 +603,7 @@ class AsyncWebCrawler:
             # Use IdentityChunking for HTML input, otherwise use provided chunking strategy
             chunking = (
                 IdentityChunking()
-                if content_format in ["html", "cleaned_html"]
+                if content_format in ["html", "cleaned_html", "fit_html"]
                 else config.chunking_strategy
             )
             sections = chunking.chunk(content)
@@ -624,6 +627,7 @@ class AsyncWebCrawler:
         return CrawlResult(
             url=url,
             html=html,
+            fit_html=fit_html,
             cleaned_html=cleaned_html,
             markdown=markdown_result,
             media=media,

crawl4ai/browser_profiler.py

@@ -128,7 +128,7 @@ class BrowserProfiler:
         os.makedirs(profile_path, exist_ok=True)
         
         # Print instructions for the user with rich formatting
-        border = "{'='*80}"
+        border = f"{'='*80}"
         self.logger.info("{border}", tag="PROFILE", params={"border": f"\n{border}"}, colors={"border": LogColor.CYAN})
         self.logger.info("Creating browser profile: {profile_name}", tag="PROFILE", params={"profile_name": profile_name}, colors={"profile_name": LogColor.GREEN})
         self.logger.info("Profile directory: {profile_path}", tag="PROFILE", params={"profile_path": profile_path}, colors={"profile_path": LogColor.YELLOW})
@@ -475,7 +475,7 @@ class BrowserProfiler:
                     self.logger.warning("  No profiles found. Create one first with option 1.", tag="PROFILES")
                     continue
                 
-                # Print profile information with colorama formatting
+                # Print profile information 
                 self.logger.info("\nAvailable profiles:", tag="PROFILES")
                 for i, profile in enumerate(profiles):
                     self.logger.info(f"[{i+1}] {profile['name']}", tag="PROFILES")
@@ -607,8 +607,8 @@ class BrowserProfiler:
             os.makedirs(profile_path, exist_ok=True)
         
         # Print initial information
-        border = f"{Fore.CYAN}{'='*80}{Style.RESET_ALL}"
-        self.logger.info(f"\n{border}", tag="CDP")
+        border = f"{'='*80}"
+        self.logger.info("{border}", tag="CDP", params={"border": border}, colors={"border": LogColor.CYAN})
         self.logger.info("Launching standalone browser with CDP debugging", tag="CDP")
         self.logger.info("Browser type: {browser_type}", tag="CDP", params={"browser_type": browser_type}, colors={"browser_type": LogColor.CYAN})
         self.logger.info("Profile path: {profile_path}", tag="CDP", params={"profile_path": profile_path}, colors={"profile_path": LogColor.YELLOW})

crawl4ai/extraction_strategy.py

@@ -1,9 +1,10 @@
 from abc import ABC, abstractmethod
 import inspect
-from typing import Any, List, Dict, Optional
+from typing import Any, List, Dict, Optional, Tuple, Pattern, Union
 from concurrent.futures import ThreadPoolExecutor, as_completed
 import json
 import time
+from enum import IntFlag, auto
 
 from .prompts import PROMPT_EXTRACT_BLOCKS, PROMPT_EXTRACT_BLOCKS_WITH_INSTRUCTION, PROMPT_EXTRACT_SCHEMA_WITH_INSTRUCTION, JSON_SCHEMA_BUILDER_XPATH, PROMPT_EXTRACT_INFERRED_SCHEMA
 from .config import (
@@ -1668,3 +1669,303 @@ class JsonXPathExtractionStrategy(JsonElementExtractionStrategy):
     def _get_element_attribute(self, element, attribute: str):
         return element.get(attribute)
 
+"""
+RegexExtractionStrategy
+Fast, zero-LLM extraction of common entities via regular expressions.
+"""
+
+_CTRL = {c: rf"\x{ord(c):02x}" for c in map(chr, range(32)) if c not in "\t\n\r"}
+
+_WB_FIX = re.compile(r"\x08")               # stray back-space   →   word-boundary
+_NEEDS_ESCAPE = re.compile(r"(?<!\\)\\(?![\\u])")   # lone backslash
+
+def _sanitize_schema(schema: Dict[str, str]) -> Dict[str, str]:
+    """Fix common JSON-escape goofs coming from LLMs or manual edits."""
+    safe = {}
+    for label, pat in schema.items():
+        # 1️⃣ replace accidental control chars (inc. the infamous back-space)
+        pat = _WB_FIX.sub(r"\\b", pat).translate(_CTRL)
+
+        # 2️⃣ double any single backslash that JSON kept single
+        pat = _NEEDS_ESCAPE.sub(r"\\\\", pat)
+
+        # 3️⃣ quick sanity compile
+        try:
+            re.compile(pat)
+        except re.error as e:
+            raise ValueError(f"Regex for '{label}' won’t compile after fix: {e}") from None
+
+        safe[label] = pat
+    return safe
+
+
+class RegexExtractionStrategy(ExtractionStrategy):
+    """
+    A lean strategy that finds e-mails, phones, URLs, dates, money, etc.,
+    using nothing but pre-compiled regular expressions.
+
+    Extraction returns::
+
+        {
+            "url":   "<page-url>",
+            "label": "<pattern-label>",
+            "value": "<matched-string>",
+            "span":  [start, end]
+        }
+
+    Only `generate_schema()` touches an LLM, extraction itself is pure Python.
+    """
+
+    # -------------------------------------------------------------- #
+    # Built-in patterns exposed as IntFlag so callers can bit-OR them
+    # -------------------------------------------------------------- #
+    class _B(IntFlag):
+        EMAIL           = auto()
+        PHONE_INTL      = auto()
+        PHONE_US        = auto()
+        URL             = auto()
+        IPV4            = auto()
+        IPV6            = auto()
+        UUID            = auto()
+        CURRENCY        = auto()
+        PERCENTAGE      = auto()
+        NUMBER          = auto()
+        DATE_ISO        = auto()
+        DATE_US         = auto()
+        TIME_24H        = auto()
+        POSTAL_US       = auto()
+        POSTAL_UK       = auto()
+        HTML_COLOR_HEX  = auto()
+        TWITTER_HANDLE  = auto()
+        HASHTAG         = auto()
+        MAC_ADDR        = auto()
+        IBAN            = auto()
+        CREDIT_CARD     = auto()
+        NOTHING         = auto()
+        ALL             = (
+            EMAIL | PHONE_INTL | PHONE_US | URL | IPV4 | IPV6 | UUID
+            | CURRENCY | PERCENTAGE | NUMBER | DATE_ISO | DATE_US | TIME_24H
+            | POSTAL_US | POSTAL_UK | HTML_COLOR_HEX | TWITTER_HANDLE
+            | HASHTAG | MAC_ADDR | IBAN | CREDIT_CARD
+        )
+
+    # user-friendly aliases  (RegexExtractionStrategy.Email, .IPv4, …)
+    Email          = _B.EMAIL
+    PhoneIntl      = _B.PHONE_INTL
+    PhoneUS        = _B.PHONE_US
+    Url            = _B.URL
+    IPv4           = _B.IPV4
+    IPv6           = _B.IPV6
+    Uuid           = _B.UUID
+    Currency       = _B.CURRENCY
+    Percentage     = _B.PERCENTAGE
+    Number         = _B.NUMBER
+    DateIso        = _B.DATE_ISO
+    DateUS         = _B.DATE_US
+    Time24h        = _B.TIME_24H
+    PostalUS       = _B.POSTAL_US
+    PostalUK       = _B.POSTAL_UK
+    HexColor       = _B.HTML_COLOR_HEX
+    TwitterHandle  = _B.TWITTER_HANDLE
+    Hashtag        = _B.HASHTAG
+    MacAddr        = _B.MAC_ADDR
+    Iban           = _B.IBAN
+    CreditCard     = _B.CREDIT_CARD
+    All            = _B.ALL
+    Nothing        = _B(0)  # no patterns
+
+    # ------------------------------------------------------------------ #
+    # Built-in pattern catalog
+    # ------------------------------------------------------------------ #
+    DEFAULT_PATTERNS: Dict[str, str] = {
+        # Communication
+        "email":           r"[\w.+-]+@[\w-]+\.[\w.-]+",
+        "phone_intl":      r"\+?\d[\d .()-]{7,}\d",
+        "phone_us":        r"\(?\d{3}\)?[ -. ]?\d{3}[ -. ]?\d{4}",
+        # Web
+        "url":             r"https?://[^\s\"'<>]+",
+        "ipv4":            r"(?:\d{1,3}\.){3}\d{1,3}",
+        "ipv6":            r"[A-F0-9]{1,4}(?::[A-F0-9]{1,4}){7}",
+        # IDs
+        "uuid":            r"[0-9a-f]{8}-[0-9a-f]{4}-[1-5][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}",
+        # Money / numbers
+        "currency":        r"(?:USD|EUR|RM|\$|€|£)\s?\d+(?:[.,]\d{2})?",
+        "percentage":      r"\d+(?:\.\d+)?%",
+        "number":          r"\b\d{1,3}(?:[,.\s]\d{3})*(?:\.\d+)?\b",
+        # Dates / Times
+        "date_iso":        r"\d{4}-\d{2}-\d{2}",
+        "date_us":         r"\d{1,2}/\d{1,2}/\d{2,4}",
+        "time_24h":        r"\b(?:[01]?\d|2[0-3]):[0-5]\d(?:[:.][0-5]\d)?\b",
+        # Misc
+        "postal_us":       r"\b\d{5}(?:-\d{4})?\b",
+        "postal_uk":       r"\b[A-Z]{1,2}\d[A-Z\d]? ?\d[A-Z]{2}\b",
+        "html_color_hex":  r"#[0-9A-Fa-f]{6}\b",
+        "twitter_handle":  r"@[\w]{1,15}",
+        "hashtag":         r"#[\w-]+",
+        "mac_addr":        r"(?:[0-9A-Fa-f]{2}:){5}[0-9A-Fa-f]{2}",
+        "iban":            r"[A-Z]{2}\d{2}[A-Z0-9]{11,30}",
+        "credit_card":     r"\b(?:4\d{12}(?:\d{3})?|5[1-5]\d{14}|3[47]\d{13}|6(?:011|5\d{2})\d{12})\b",
+    }
+
+    _FLAGS = re.IGNORECASE | re.MULTILINE
+    _UNWANTED_PROPS = {
+        "provider": "Use llm_config instead",
+        "api_token": "Use llm_config instead",
+    }
+
+    # ------------------------------------------------------------------ #
+    # Construction
+    # ------------------------------------------------------------------ #
+    def __init__(
+        self,
+        pattern: "_B" = _B.NOTHING,
+        *,
+        custom: Optional[Union[Dict[str, str], List[Tuple[str, str]]]] = None,
+        input_format: str = "fit_html",
+        **kwargs,
+    ) -> None:
+        """
+        Args:
+            patterns: Custom patterns overriding or extending defaults.
+                      Dict[label, regex] or list[tuple(label, regex)].
+            input_format: "html", "markdown" or "text".
+            **kwargs: Forwarded to ExtractionStrategy.
+        """
+        super().__init__(input_format=input_format, **kwargs)
+
+        # 1️⃣  take only the requested built-ins
+        merged: Dict[str, str] = {
+            key: rx
+            for key, rx in self.DEFAULT_PATTERNS.items()
+            if getattr(self._B, key.upper()).value & pattern
+        }
+
+        # 2️⃣  apply user overrides / additions
+        if custom:
+            if isinstance(custom, dict):
+                merged.update(custom)
+            else:  # iterable of (label, regex)
+                merged.update({lbl: rx for lbl, rx in custom})
+
+        self._compiled: Dict[str, Pattern] = {
+            lbl: re.compile(rx, self._FLAGS) for lbl, rx in merged.items()
+        }
+
+    # ------------------------------------------------------------------ #
+    # Extraction
+    # ------------------------------------------------------------------ #
+    def extract(self, url: str, content: str, *q, **kw) -> List[Dict[str, Any]]:
+        # text = self._plain_text(html)
+        out: List[Dict[str, Any]] = []
+
+        for label, cre in self._compiled.items():
+            for m in cre.finditer(content):
+                out.append(
+                    {
+                        "url": url,
+                        "label": label,
+                        "value": m.group(0),
+                        "span": [m.start(), m.end()],
+                    }
+                )
+        return out
+
+    # ------------------------------------------------------------------ #
+    # Helpers
+    # ------------------------------------------------------------------ #
+    def _plain_text(self, content: str) -> str:
+        if self.input_format == "text":
+            return content
+        return BeautifulSoup(content, "lxml").get_text(" ", strip=True)
+
+    # ------------------------------------------------------------------ #
+    # LLM-assisted pattern generator
+    # ------------------------------------------------------------------ #
+    # ------------------------------------------------------------------ #
+    # LLM-assisted one-off pattern builder
+    # ------------------------------------------------------------------ #
+    @staticmethod
+    def generate_pattern(
+        label: str,
+        html: str,
+        *,
+        query: Optional[str] = None,
+        examples: Optional[List[str]] = None,
+        llm_config: Optional[LLMConfig] = None,
+        **kwargs,
+    ) -> Dict[str, str]:
+        """
+        Ask an LLM for a single page-specific regex and return
+            {label: pattern}   ── ready for RegexExtractionStrategy(custom=…)
+        """
+
+        # ── guard deprecated kwargs
+        for k in RegexExtractionStrategy._UNWANTED_PROPS:
+            if k in kwargs:
+                raise AttributeError(
+                    f"{k} is deprecated, {RegexExtractionStrategy._UNWANTED_PROPS[k]}"
+                )
+
+        # ── default LLM config
+        if llm_config is None:
+            llm_config = create_llm_config()
+
+        # ── system prompt – hardened
+        system_msg = (
+            "You are an expert Python-regex engineer.\n"
+            f"Return **one** JSON object whose single key is exactly \"{label}\", "
+            "and whose value is a raw-string regex pattern that works with "
+            "the standard `re` module in Python.\n\n"
+            "Strict rules (obey every bullet):\n"
+            "• If a *user query* is supplied, treat it as the precise semantic target and optimise the "
+            "  pattern to capture ONLY text that answers that query. If the query conflicts with the "
+            "  sample HTML, the HTML wins.\n"
+            "• Tailor the pattern to the *sample HTML* – reproduce its exact punctuation, spacing, "
+            "  symbols, capitalisation, etc. Do **NOT** invent a generic form.\n"
+            "• Keep it minimal and fast: avoid unnecessary capturing, prefer non-capturing `(?: … )`, "
+            "  and guard against catastrophic backtracking.\n"
+            "• Anchor with `^`, `$`, or `\\b` only when it genuinely improves precision.\n"
+            "• Use inline flags like `(?i)` when needed; no verbose flag comments.\n"
+            "• Output must be valid JSON – no markdown, code fences, comments, or extra keys.\n"
+            "• The regex value must be a Python string literal: **double every backslash** "
+            "(e.g. `\\\\b`, `\\\\d`, `\\\\\\\\`).\n\n"
+            "Example valid output:\n"
+            f"{{\"{label}\": \"(?:RM|rm)\\\\s?\\\\d{{1,3}}(?:,\\\\d{{3}})*(?:\\\\.\\\\d{{2}})?\"}}"
+        )
+
+        # ── user message: cropped HTML + optional hints
+        user_parts = ["```html", html[:5000], "```"]  # protect token budget
+        if query:
+            user_parts.append(f"\n\n## Query\n{query.strip()}")
+        if examples:
+            user_parts.append("## Examples\n" + "\n".join(examples[:20]))
+        user_msg = "\n\n".join(user_parts)
+
+        # ── LLM call (with retry/backoff)
+        resp = perform_completion_with_backoff(
+            provider=llm_config.provider,
+            prompt_with_variables="\n\n".join([system_msg, user_msg]),
+            json_response=True,
+            api_token=llm_config.api_token,
+            base_url=llm_config.base_url,
+            extra_args=kwargs,
+        )
+
+        # ── clean & load JSON (fix common escape mistakes *before* json.loads)
+        raw = resp.choices[0].message.content
+        raw = raw.replace("\x08", "\\b")                     # stray back-space → \b
+        raw = re.sub(r'(?<!\\)\\(?![\\u"])', r"\\\\", raw)   # lone \ → \\
+
+        try:
+            pattern_dict = json.loads(raw)
+        except Exception as exc:
+            raise ValueError(f"LLM did not return valid JSON: {raw}") from exc
+
+        # quick sanity-compile
+        for lbl, pat in pattern_dict.items():
+            try:
+                re.compile(pat)
+            except re.error as e:
+                raise ValueError(f"Invalid regex for '{lbl}': {e}") from None
+
+        return pattern_dict

crawl4ai/models.py

@@ -129,6 +129,7 @@ class MarkdownGenerationResult(BaseModel):
 class CrawlResult(BaseModel):
     url: str
     html: str
+    fit_html: Optional[str] = None
     success: bool
     cleaned_html: Optional[str] = None
     media: Dict[str, List[Dict]] = {}

crawl4ai/utils.py

@@ -2735,33 +2735,67 @@ def preprocess_html_for_schema(html_content, text_threshold=100, attr_value_thre
             # Also truncate tail text if present
             if element.tail and len(element.tail.strip()) > text_threshold:
                 element.tail = element.tail.strip()[:text_threshold] + '...'
-        
-        # 4. Find repeated patterns and keep only a few examples
-        # This is a simplistic approach - more sophisticated pattern detection could be implemented
-        pattern_elements = {}
-        for element in tree.xpath('//*[contains(@class, "")]'):
-            parent = element.getparent()
+
+        # 4. Detect duplicates and drop them in a single pass
+        seen: dict[tuple, None] = {}
+        for el in list(tree.xpath('//*[@class]')):          # snapshot once, XPath is fast
+            parent = el.getparent()
             if parent is None:
                 continue
-                
-            # Create a signature based on tag and classes
-            classes = element.get('class', '')
-            if not classes:
+
+            cls = el.get('class')
+            if not cls:
                 continue
-            signature = f"{element.tag}.{classes}"
-            
-            if signature in pattern_elements:
-                pattern_elements[signature].append(element)
+
+            # ── build signature ───────────────────────────────────────────
+            h = xxhash.xxh64()                              # stream, no big join()
+            for txt in el.itertext():
+                h.update(txt)
+            sig = (el.tag, cls, h.intdigest())             # tuple cheaper & hashable
+
+            # ── first seen? keep – else drop ─────────────
+            if sig in seen and parent is not None:
+                parent.remove(el)                           # duplicate
             else:
-                pattern_elements[signature] = [element]
+                seen[sig] = None
         
-        # Keep only 3 examples of each repeating pattern
-        for signature, elements in pattern_elements.items():
-            if len(elements) > 3:
-                # Keep the first 2 and last elements
-                for element in elements[2:-1]:
-                    if element.getparent() is not None:
-                        element.getparent().remove(element)
+        # # 4. Find repeated patterns and keep only a few examples
+        # # This is a simplistic approach - more sophisticated pattern detection could be implemented
+        # pattern_elements = {}
+        # for element in tree.xpath('//*[contains(@class, "")]'):
+        #     parent = element.getparent()
+        #     if parent is None:
+        #         continue
+                
+        #     # Create a signature based on tag and classes
+        #     classes = element.get('class', '')
+        #     if not classes:
+        #         continue
+        #     innert_text = ''.join(element.xpath('.//text()'))
+        #     innert_text_hash = xxhash.xxh64(innert_text.encode()).hexdigest()
+        #     signature = f"{element.tag}.{classes}.{innert_text_hash}"
+            
+        #     if signature in pattern_elements:
+        #         pattern_elements[signature].append(element)
+        #     else:
+        #         pattern_elements[signature] = [element]
+        
+        # # Keep only first examples of each repeating pattern
+        # for signature, elements in pattern_elements.items():
+        #     if len(elements) > 1:
+        #         # Keep the first element and remove the rest
+        #         for element in elements[1:]:
+        #             if element.getparent() is not None:
+        #                 element.getparent().remove(element)
+
+
+        # # Keep only 3 examples of each repeating pattern
+        # for signature, elements in pattern_elements.items():
+        #     if len(elements) > 3:
+        #         # Keep the first 2 and last elements
+        #         for element in elements[2:-1]:
+        #             if element.getparent() is not None:
+        #                 element.getparent().remove(element)
         
         # 5. Convert back to string
         result = etree.tostring(tree, encoding='unicode', method='html')

deploy/docker/README.md

@@ -17,6 +17,7 @@
   - [Screenshot Endpoint](#screenshot-endpoint)
   - [PDF Export Endpoint](#pdf-export-endpoint)
   - [JavaScript Execution Endpoint](#javascript-execution-endpoint)
+  - [Browser VNC Endpoint](#browser-vnc-endpoint)
   - [Library Context Endpoint](#library-context-endpoint)
 - [MCP (Model Context Protocol) Support](#mcp-model-context-protocol-support)
   - [What is MCP?](#what-is-mcp)
@@ -377,6 +378,20 @@ Executes JavaScript snippets on the specified URL and returns the full crawl res
 
 - `scripts`: List of JavaScript snippets to execute sequentially
 
+### Browser VNC Endpoint
+
+```
+GET /vnc
+```
+
+Opens a browser-based VNC session for interacting with the container's desktop environment. Use `/vnc/url` to retrieve only the iframe URL.
+
+```
+GET /vnc/url
+```
+
+Returns a JSON object containing the URL of the embedded noVNC client.
+
 ---
 
 ## Dockerfile Parameters

deploy/docker/api.py

@@ -1,8 +1,10 @@
 import os
 import json
 import asyncio
-from typing import List, Tuple
+from typing import List, Tuple, Dict
 from functools import partial
+from uuid import uuid4
+from datetime import datetime
 
 import logging
 from typing import Optional, AsyncGenerator
@@ -272,7 +274,9 @@ async def handle_llm_request(
 async def handle_task_status(
     redis: aioredis.Redis,
     task_id: str,
-    base_url: str
+    base_url: str,
+    *,
+    keep: bool = False
 ) -> JSONResponse:
     """Handle task status check requests."""
     task = await redis.hgetall(f"task:{task_id}")
@@ -286,7 +290,7 @@ async def handle_task_status(
     response = create_task_response(task, task_id, base_url)
 
     if task["status"] in [TaskStatus.COMPLETED, TaskStatus.FAILED]:
-        if should_cleanup_task(task["created_at"]):
+        if not keep and should_cleanup_task(task["created_at"]):
             await redis.delete(f"task:{task_id}")
 
     return JSONResponse(response)
@@ -520,4 +524,48 @@ async def handle_stream_crawl_request(
         raise HTTPException(
             status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
             detail=str(e)
-        )
+        )
+        
+async def handle_crawl_job(
+    redis,
+    background_tasks: BackgroundTasks,
+    urls: List[str],
+    browser_config: Dict,
+    crawler_config: Dict,
+    config: Dict,
+) -> Dict:
+    """
+    Fire-and-forget version of handle_crawl_request.
+    Creates a task in Redis, runs the heavy work in a background task,
+    lets /crawl/job/{task_id} polling fetch the result.
+    """
+    task_id = f"crawl_{uuid4().hex[:8]}"
+    await redis.hset(f"task:{task_id}", mapping={
+        "status": TaskStatus.PROCESSING,         # <-- keep enum values consistent
+        "created_at": datetime.utcnow().isoformat(),
+        "url": json.dumps(urls),                 # store list as JSON string
+        "result": "",
+        "error": "",
+    })
+
+    async def _runner():
+        try:
+            result = await handle_crawl_request(
+                urls=urls,
+                browser_config=browser_config,
+                crawler_config=crawler_config,
+                config=config,
+            )
+            await redis.hset(f"task:{task_id}", mapping={
+                "status": TaskStatus.COMPLETED,
+                "result": json.dumps(result),
+            })
+            await asyncio.sleep(5)  # Give Redis time to process the update
+        except Exception as exc:
+            await redis.hset(f"task:{task_id}", mapping={
+                "status": TaskStatus.FAILED,
+                "error": str(exc),
+            })
+
+    background_tasks.add_task(_runner)
+    return {"task_id": task_id}

deploy/docker/config.yml

@@ -3,7 +3,7 @@ app:
   title: "Crawl4AI API"
   version: "1.0.0"
   host: "0.0.0.0"
-  port: 11235
+  port: 11234
   reload: False
   workers: 1
   timeout_keep_alive: 300

deploy/docker/job.py

@@ -0,0 +1,99 @@
+"""
+Job endpoints (enqueue + poll) for long-running LL​M extraction and raw crawl.
+Relies on the existing Redis task helpers in api.py
+"""
+
+from typing import Dict, Optional, Callable
+from fastapi import APIRouter, BackgroundTasks, Depends, Request
+from pydantic import BaseModel, HttpUrl
+
+from api import (
+    handle_llm_request,
+    handle_crawl_job,
+    handle_task_status,
+)
+
+# ------------- dependency placeholders -------------
+_redis = None        # will be injected from server.py
+_config = None
+_token_dep: Callable = lambda: None  # dummy until injected
+
+# public router
+router = APIRouter()
+
+
+# === init hook called by server.py =========================================
+def init_job_router(redis, config, token_dep) -> APIRouter:
+    """Inject shared singletons and return the router for mounting."""
+    global _redis, _config, _token_dep
+    _redis, _config, _token_dep = redis, config, token_dep
+    return router
+
+
+# ---------- payload models --------------------------------------------------
+class LlmJobPayload(BaseModel):
+    url:    HttpUrl
+    q:      str
+    schema: Optional[str] = None
+    cache:  bool = False
+
+
+class CrawlJobPayload(BaseModel):
+    urls:           list[HttpUrl]
+    browser_config: Dict = {}
+    crawler_config: Dict = {}
+
+
+# ---------- LL​M job ---------------------------------------------------------
+@router.post("/llm/job", status_code=202)
+async def llm_job_enqueue(
+        payload: LlmJobPayload,
+        background_tasks: BackgroundTasks,
+        request: Request,
+        _td: Dict = Depends(lambda: _token_dep()),   # late-bound dep
+):
+    return await handle_llm_request(
+        _redis,
+        background_tasks,
+        request,
+        str(payload.url),
+        query=payload.q,
+        schema=payload.schema,
+        cache=payload.cache,
+        config=_config,
+    )
+
+
+@router.get("/llm/job/{task_id}")
+async def llm_job_status(
+    request: Request,
+    task_id: str,
+    _td: Dict = Depends(lambda: _token_dep())
+):
+    return await handle_task_status(_redis, task_id)
+
+
+# ---------- CRAWL job -------------------------------------------------------
+@router.post("/crawl/job", status_code=202)
+async def crawl_job_enqueue(
+        payload: CrawlJobPayload,
+        background_tasks: BackgroundTasks,
+        _td: Dict = Depends(lambda: _token_dep()),
+):
+    return await handle_crawl_job(
+        _redis,
+        background_tasks,
+        [str(u) for u in payload.urls],
+        payload.browser_config,
+        payload.crawler_config,
+        config=_config,
+    )
+
+
+@router.get("/crawl/job/{task_id}")
+async def crawl_job_status(
+    request: Request,
+    task_id: str,
+    _td: Dict = Depends(lambda: _token_dep())
+):
+    return await handle_task_status(_redis, task_id, base_url=str(request.base_url))

deploy/docker/schemas.py

@@ -0,0 +1,42 @@
+from typing import List, Optional, Dict
+from enum import Enum
+from pydantic import BaseModel, Field
+from utils import FilterType
+
+
+class CrawlRequest(BaseModel):
+    urls: List[str] = Field(min_length=1, max_length=100)
+    browser_config: Optional[Dict] = Field(default_factory=dict)
+    crawler_config: Optional[Dict] = Field(default_factory=dict)
+
+class MarkdownRequest(BaseModel):
+    """Request body for the /md endpoint."""
+    url: str                    = Field(...,  description="Absolute http/https URL to fetch")
+    f:   FilterType             = Field(FilterType.FIT,
+                                        description="Content‑filter strategy: FIT, RAW, BM25, or LLM")
+    q:   Optional[str] = Field(None,  description="Query string used by BM25/LLM filters")
+    c:   Optional[str] = Field("0",   description="Cache‑bust / revision counter")
+
+
+class RawCode(BaseModel):
+    code: str
+
+class HTMLRequest(BaseModel):
+    url: str
+    
+class ScreenshotRequest(BaseModel):
+    url: str
+    screenshot_wait_for: Optional[float] = 2
+    output_path: Optional[str] = None
+
+class PDFRequest(BaseModel):
+    url: str
+    output_path: Optional[str] = None
+
+
+class JSEndpointRequest(BaseModel):
+    url: str
+    scripts: List[str] = Field(
+        ...,
+        description="List of separated JavaScript snippets to execute"
+    )

deploy/docker/server.py

@@ -12,7 +12,7 @@ from crawl4ai import AsyncWebCrawler, BrowserConfig, CrawlerRunConfig
 from auth import create_access_token, get_token_dependency, TokenRequest
 from pydantic import BaseModel
 from typing import Optional, List, Dict
-from fastapi import Request, Depends 
+from fastapi import Request, Depends
 from fastapi.responses import FileResponse
 import base64
 import re
@@ -22,8 +22,22 @@ from api import (
     handle_stream_crawl_request, handle_crawl_request,
     stream_results
 )
+from schemas import (
+    CrawlRequest,
+    MarkdownRequest,
+    RawCode,
+    HTMLRequest,
+    ScreenshotRequest,
+    PDFRequest,
+    JSEndpointRequest,
+)
+
 from utils import (
-    FilterType, load_config, setup_logging, verify_email_domain
+    FilterType,
+    load_config,
+    setup_logging,
+    verify_email_domain,
+    get_base_url,
 )
 import os
 import sys
@@ -37,23 +51,17 @@ from fastapi import (
     FastAPI, HTTPException, Request, Path, Query, Depends
 )
 from rank_bm25 import BM25Okapi
-
-def chunk_code_functions(code: str) -> List[str]:
-    tree = ast.parse(code)
-    lines = code.splitlines()
-    chunks = []
-    for node in tree.body:
-        if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
-            start = node.lineno - 1
-            end = getattr(node, 'end_lineno', start + 1)
-            chunks.append("\n".join(lines[start:end]))
-    return chunks
 from fastapi.responses import (
-    StreamingResponse, RedirectResponse, PlainTextResponse, JSONResponse
+    StreamingResponse,
+    RedirectResponse,
+    PlainTextResponse,
+    JSONResponse,
+    HTMLResponse,
 )
 from fastapi.middleware.httpsredirect import HTTPSRedirectMiddleware
 from fastapi.middleware.trustedhost import TrustedHostMiddleware
 from fastapi.staticfiles import StaticFiles
+from job import init_job_router
 
 from mcp_bridge import attach_mcp, mcp_resource, mcp_template, mcp_tool
 
@@ -129,13 +137,31 @@ app.mount(
     name="play",
 )
 
-# Optional nice‑to‑have: opening the root shows the playground
+# Serve noVNC static files if available
+VNC_DIR = pathlib.Path("/opt/novnc")
+if VNC_DIR.exists():
+    app.mount("/novnc", StaticFiles(directory=VNC_DIR, html=True), name="novnc")
 
 
 @app.get("/")
 async def root():
     return RedirectResponse("/playground")
 
+
+@app.get("/vnc")
+async def vnc_page(request: Request):
+    """Return a simple page embedding the noVNC client."""
+    url = f"{get_base_url(request)}/novnc/vnc.html?autoconnect=true&resize=scale"
+    html = f"<iframe src='{url}' width='1024' height='768' style='border:none'></iframe>"
+    return HTMLResponse(f"<html><body>{html}</body></html>")
+
+
+@app.get("/vnc/url")
+async def vnc_url(request: Request):
+    """Return the direct URL to the noVNC client."""
+    url = f"{get_base_url(request)}/novnc/vnc.html?autoconnect=true&resize=scale"
+    return {"url": url}
+
 # ─────────────────── infra / middleware  ─────────────────────
 redis = aioredis.from_url(config["redis"].get("uri", "redis://localhost"))
 
@@ -211,48 +237,10 @@ def _safe_eval_config(expr: str) -> dict:
     return obj.dump()
 
 
-# ───────────────────────── Schemas ───────────────────────────
-class CrawlRequest(BaseModel):
-    urls: List[str] = Field(min_length=1, max_length=100)
-    browser_config: Optional[Dict] = Field(default_factory=dict)
-    crawler_config: Optional[Dict] = Field(default_factory=dict)
-
-# ────────────── Schemas ──────────────
-class MarkdownRequest(BaseModel):
-    """Request body for the /md endpoint."""
-    url: str                    = Field(...,  description="Absolute http/https URL to fetch")
-    f:   FilterType             = Field(FilterType.FIT,
-                                        description="Content‑filter strategy: FIT, RAW, BM25, or LLM")
-    q:   Optional[str] = Field(None,  description="Query string used by BM25/LLM filters")
-    c:   Optional[str] = Field("0",   description="Cache‑bust / revision counter")
-
-
-class RawCode(BaseModel):
-    code: str
-
-class HTMLRequest(BaseModel):
-    url: str
-    
-class ScreenshotRequest(BaseModel):
-    url: str
-    screenshot_wait_for: Optional[float] = 2
-    output_path: Optional[str] = None
-
-class PDFRequest(BaseModel):
-    url: str
-    output_path: Optional[str] = None
-
-
-class JSEndpointRequest(BaseModel):
-    url: str
-    scripts: List[str] = Field(
-        ...,
-        description="List of separated JavaScript snippets to execute"
-    )
+# ── job router ──────────────────────────────────────────────
+app.include_router(init_job_router(redis, config, token_dep))
 
 # ──────────────────────── Endpoints ──────────────────────────
-
-
 @app.post("/token")
 async def get_token(req: TokenRequest):
     if not verify_email_domain(req.email):
@@ -278,7 +266,8 @@ async def get_markdown(
     _td: Dict = Depends(token_dep),
 ):
     if not body.url.startswith(("http://", "https://")):
-        raise HTTPException(400, "URL must be absolute and start with http/https")
+        raise HTTPException(
+            400, "URL must be absolute and start with http/https")
     markdown = await handle_markdown_request(
         body.url, body.f, body.q, body.c, config
     )
@@ -314,12 +303,13 @@ async def generate_html(
 
 # Screenshot endpoint
 
+
 @app.post("/screenshot")
 @limiter.limit(config["rate_limiting"]["default_limit"])
 @mcp_tool("screenshot")
 async def generate_screenshot(
     request: Request,
-    body: ScreenshotRequest, 
+    body: ScreenshotRequest,
     _td: Dict = Depends(token_dep),
 ):
     """
@@ -327,7 +317,8 @@ async def generate_screenshot(
     Use when you need an image snapshot of the rendered page. Its recommened to provide an output path to save the screenshot.
     Then in result instead of the screenshot you will get a path to the saved file.
     """
-    cfg = CrawlerRunConfig(screenshot=True, screenshot_wait_for=body.screenshot_wait_for)
+    cfg = CrawlerRunConfig(
+        screenshot=True, screenshot_wait_for=body.screenshot_wait_for)
     async with AsyncWebCrawler(config=BrowserConfig()) as crawler:
         results = await crawler.arun(url=body.url, config=cfg)
     screenshot_data = results[0].screenshot
@@ -341,12 +332,13 @@ async def generate_screenshot(
 
 # PDF endpoint
 
+
 @app.post("/pdf")
 @limiter.limit(config["rate_limiting"]["default_limit"])
 @mcp_tool("pdf")
 async def generate_pdf(
     request: Request,
-    body: PDFRequest, 
+    body: PDFRequest,
     _td: Dict = Depends(token_dep),
 ):
     """
@@ -384,7 +376,7 @@ async def execute_js(
         Your script will replace '{script}' and execute in the browser context. So provide either an IIFE or a sync/async function that returns a value.
     Return Format:
         - The return result is an instance of CrawlResult, so you have access to markdown, links, and other stuff. If this is enough, you don't need to call again for other endpoints.
-        
+
         ```python
         class CrawlResult(BaseModel):
             url: str
@@ -418,7 +410,7 @@ async def execute_js(
             fit_markdown: Optional[str] = None
             fit_html: Optional[str] = None
         ```
-        
+
     """
     cfg = CrawlerRunConfig(js_code=body.scripts)
     async with AsyncWebCrawler(config=BrowserConfig()) as crawler:
@@ -507,6 +499,7 @@ async def crawl_stream(
         },
     )
 
+
 def chunk_code_functions(code_md: str) -> List[str]:
     """Extract each function/class from markdown code blocks per file."""
     pattern = re.compile(
@@ -530,6 +523,7 @@ def chunk_code_functions(code_md: str) -> List[str]:
                 chunks.append(f"# File: {file_path}\n{snippet}")
     return chunks
 
+
 def chunk_doc_sections(doc: str) -> List[str]:
     lines = doc.splitlines(keepends=True)
     sections = []
@@ -545,6 +539,7 @@ def chunk_doc_sections(doc: str) -> List[str]:
         sections.append("".join(current))
     return sections
 
+
 @app.get("/ask")
 @limiter.limit(config["rate_limiting"]["default_limit"])
 @mcp_tool("ask")
@@ -552,21 +547,24 @@ async def get_context(
     request: Request,
     _td: Dict = Depends(token_dep),
     context_type: str = Query("all", regex="^(code|doc|all)$"),
-    query: Optional[str] = Query(None, description="search query to filter chunks"),
-    score_ratio: float = Query(0.5, ge=0.0, le=1.0, description="min score as fraction of max_score"),
-    max_results: int = Query(20, ge=1, description="absolute cap on returned chunks"),
+    query: Optional[str] = Query(
+        None, description="search query to filter chunks"),
+    score_ratio: float = Query(
+        0.5, ge=0.0, le=1.0, description="min score as fraction of max_score"),
+    max_results: int = Query(
+        20, ge=1, description="absolute cap on returned chunks"),
 ):
     """
     This end point is design for any questions about Crawl4ai library. It returns a plain text markdown with extensive information about Crawl4ai. 
     You can use this as a context for any AI assistant. Use this endpoint for AI assistants to retrieve library context for decision making or code generation tasks.
     Alway is BEST practice you provide a query to filter the context. Otherwise the lenght of the response will be very long.
-    
+
     Parameters:
     - context_type: Specify "code" for code context, "doc" for documentation context, or "all" for both.
     - query: RECOMMENDED search query to filter paragraphs using BM25. You can leave this empty to get all the context.
     - score_ratio: Minimum score as a fraction of the maximum score for filtering results.
     - max_results: Maximum number of results to return. Default is 20.
-    
+
     Returns:
     - JSON response with the requested context.
     - If "code" is specified, returns the code context.
@@ -576,7 +574,7 @@ async def get_context(
     # load contexts
     base = os.path.dirname(__file__)
     code_path = os.path.join(base, "c4ai-code-context.md")
-    doc_path  = os.path.join(base, "c4ai-doc-context.md")
+    doc_path = os.path.join(base, "c4ai-doc-context.md")
     if not os.path.exists(code_path) or not os.path.exists(doc_path):
         raise HTTPException(404, "Context files not found")
 
@@ -626,7 +624,7 @@ async def get_context(
         ]
 
     return JSONResponse(results)
-    
+
 
 # attach MCP layer (adds /mcp/ws, /mcp/sse, /mcp/schema)
 print(f"MCP server running on {config['app']['host']}:{config['app']['port']}")

deploy/docker/supervisord.conf

@@ -20,9 +20,53 @@ user=appuser                    ; Run gunicorn as our non-root user
 autorestart=true
 priority=20
 environment=PYTHONUNBUFFERED=1  ; Ensure Python output is sent straight to logs
+environment=DISPLAY=:99
 stdout_logfile=/dev/stdout      ; Redirect gunicorn stdout to container stdout
 stdout_logfile_maxbytes=0
 stderr_logfile=/dev/stderr      ; Redirect gunicorn stderr to container stderr
 stderr_logfile_maxbytes=0
 
+[program:xvfb]
+command=/usr/bin/Xvfb :99 -screen 0 1280x720x24
+user=appuser
+autorestart=true
+priority=5
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+
+[program:fluxbox]
+command=/usr/bin/fluxbox
+user=appuser
+autorestart=true
+priority=6
+environment=DISPLAY=:99
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+
+[program:x11vnc]
+command=/usr/bin/x11vnc -display :99 -nopw -forever -shared -rfbport 5900 -quiet
+user=appuser
+autorestart=true
+priority=7
+environment=DISPLAY=:99
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+
+[program:websockify]
+command=/usr/bin/websockify 6080 localhost:5900 --web /opt/novnc
+user=appuser
+autorestart=true
+priority=8
+environment=DISPLAY=:99
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+
 # Optional: Add filebeat or other logging agents here if needed

deploy/docker/utils.py

@@ -45,10 +45,10 @@ def datetime_handler(obj: any) -> Optional[str]:
         return obj.isoformat()
     raise TypeError(f"Object of type {type(obj)} is not JSON serializable")
 
-def should_cleanup_task(created_at: str) -> bool:
+def should_cleanup_task(created_at: str, ttl_seconds: int = 3600) -> bool:
     """Check if task should be cleaned up based on creation time."""
     created = datetime.fromisoformat(created_at)
-    return (datetime.now() - created).total_seconds() > 3600
+    return (datetime.now() - created).total_seconds() > ttl_seconds
 
 def decode_redis_hash(hash_data: Dict[bytes, bytes]) -> Dict[str, str]:
     """Decode Redis hash data from bytes to strings."""

docs/apps/linkdin/README.md

@@ -16,12 +16,12 @@ prospect‑wizard/
 
 ### 1.1  Install dependencies
 ```bash
-pip install crawl4ai openai sentence-transformers networkx pandas vis-network rich
+pip install crawl4ai litellm sentence-transformers pandas rich
 ```
 
 ### 1.2  Create / warm a LinkedIn browser profile
 ```bash
-crwl profiler
+crwl profiles
 ```
 1. The interactive shell shows **New profile** – hit **enter**.
 2. Choose a name, e.g. `profile_linkedin_uc`.
@@ -34,16 +34,16 @@ crwl profiler
 ## 2  Discovery – scrape companies & people
 
 ```bash
-python c4ai_discover.py full \ 
-  --query "health insurance management" \ 
+python c4ai_discover.py full \
+  --query "health insurance management" \
   --geo 102713980 \               # Malaysia geoUrn
-  --title_filters "" \            # or "Product,Engineering"
-  --max_companies 10 \            # default set small for workshops
-  --max_people 20 \               # \^ same
-  --profile-name profile_linkedin_uc \ 
-  --outdir ./data \ 
-  --concurrency 2 \ 
-  --log_level debug
+  --title-filters "" \            # or "Product,Engineering"
+  --max-companies 10 \            # default set small for workshops
+  --max-people 20 \               # \^ same
+  --profile-name profile_linkedin_uc \
+  --outdir ./data \
+  --concurrency 2 \
+  --log-level debug
 ```
 **Outputs** in `./data/`:
 * `companies.jsonl` – one JSON per company
@@ -66,14 +66,15 @@ _See more: <https://www.linkedin.com/search/results/companies/?geoUrn=XXX> – t
 ## 3  Insights – embeddings, org‑charts, decision makers
 
 ```bash
-python c4ai_insights.py \ 
-  --in  ./data \ 
-  --out ./data \ 
-  --embed_model all-MiniLM-L6-v2 \ 
-  --top_k 10 \ 
-  --openai_model gpt-4.1 \ 
-  --max_llm_tokens 8024 \ 
-  --llm_temperature 1.0 \ 
+python c4ai_insights.py \
+  --in ./data \
+  --out ./data \
+  --embed-model all-MiniLM-L6-v2 \
+  --llm-provider gemini/gemini-2.0-flash \
+  --llm-api-key "" \
+  --top-k 10 \
+  --max-llm-tokens 8024 \
+  --llm-temperature 1.0 \
   --workers 4
 ```
 Emits next to the Stage‑1 files:
@@ -121,6 +122,6 @@ The page fetches `data/company_graph.json` and the `org_chart_*.json` files auto
 ---
 
 ### TL;DR
-`crwl profiler` → `c4ai_discover.py` → `c4ai_insights.py` → open `graph_view_template.html`.  
+`crwl profiles` → `c4ai_discover.py` → `c4ai_insights.py` → open `graph_view_template.html`.  
 Live long and `import crawl4ai`.
 

docs/apps/linkdin/c4ai_discover.py

@@ -43,7 +43,6 @@ from rich.console import Console
 from rich.logging import RichHandler
 
 from datetime import datetime, UTC
-from itertools import cycle
 from textwrap import dedent
 from types import SimpleNamespace
 from typing import Dict, List, Optional
@@ -272,7 +271,7 @@ def build_arg_parser() -> argparse.ArgumentParser:
         parser.add_argument("--title-filters", default="Product,Engineering", help="comma list of job keywords")
         parser.add_argument("--max-companies", type=int, default=1000)
         parser.add_argument("--max-people", type=int, default=500)
-        parser.add_argument("--profile-path", default=str(pathlib.Path.home() / ".crawl4ai/profiles/profile_linkedin_uc"))
+        parser.add_argument("--profile-name", default=str(pathlib.Path.home() / ".crawl4ai/profiles/profile_linkedin_uc"))
         parser.add_argument("--outdir", default="./output")
         parser.add_argument("--concurrency", type=int, default=4)
         parser.add_argument("--log-level", default="info", choices=["debug", "info", "warn", "error"])
@@ -355,8 +354,7 @@ async def async_main(opts):
         user_agent_generator_config= {
             "platforms": "mobile",
             "os": "Android"
-        },
-        verbose=False,
+        }
     )
     crawler = AsyncWebCrawler(config=bc)
     
@@ -366,7 +364,7 @@ async def async_main(opts):
     # crawler = await next_crawler().start()
     try:
         # Build LinkedIn search URL
-        search_url = f"https://www.linkedin.com/search/results/companies/?keywords={quote(opts.query)}&geoUrn={opts.geo}"
+        search_url = f'https://www.linkedin.com/search/results/companies/?keywords={quote(opts.query)}&companyHqGeo="{opts.geo}"'
         logging.info("Seed URL => %s", search_url)
 
         companies: List[Dict] = []
@@ -425,14 +423,13 @@ def main():
     if cli_opts.debug:
         opts = detect_debug_defaults(force=True)
     else:
-        env_defaults = detect_debug_defaults()
         env_defaults = detect_debug_defaults()
         opts = env_defaults if env_defaults else cli_opts
 
     if not getattr(opts, "cmd", None):
         opts.cmd = "full"
 
-    exit_code = asyncio.run(async_main(opts))
+    exit_code = asyncio.run(async_main(cli_opts))
     sys.exit(exit_code)
 
 

docs/apps/linkdin/c4ai_insights.py

@@ -20,7 +20,7 @@ from __future__ import annotations
 # Imports & Third-party
 # ───────────────────────────────────────────────────────────────────────────────
 
-import argparse, asyncio, json, os, sys, pathlib, random, time, csv
+import argparse, asyncio, json, pathlib, random
 from datetime import datetime, UTC
 from types import SimpleNamespace
 from pathlib import Path
@@ -30,7 +30,7 @@ from rich.console import Console
 from rich.logging import RichHandler
 from rich.progress import Progress, SpinnerColumn, BarColumn, TextColumn, TimeElapsedColumn
 import logging
-from jinja2 import Environment, FileSystemLoader, select_autoescape
+
 
 BASE_DIR = pathlib.Path(__file__).resolve().parent
 
@@ -43,7 +43,7 @@ import numpy as np
 import pandas as pd
 import hashlib
 
-from openai import OpenAI                    # same SDK you pre-loaded
+from litellm import completion   #Support any LLM Provider
 
 # ───────────────────────────────────────────────────────────────────────────────
 # Utils
@@ -70,11 +70,11 @@ def dev_defaults() -> SimpleNamespace:
         out_dir="./insights_debug",
         embed_model="all-MiniLM-L6-v2",
         top_k=10,
-        openai_model="gpt-4.1",
+        llm_provider="openai/gpt-4.1", 
+        llm_api_key=None,
         max_llm_tokens=8000,
         llm_temperature=1.0,
-        workers=4,           # parallel processing
-        stub=False,          # manual
+        workers=4
     )
 
 # ───────────────────────────────────────────────────────────────────────────────
@@ -166,7 +166,7 @@ def build_company_graph(companies, embeds:np.ndarray, top_k:int) -> Dict[str,Any
 # ───────────────────────────────────────────────────────────────────────────────
 # Org-chart via LLM
 # ───────────────────────────────────────────────────────────────────────────────
-async def infer_org_chart_llm(company, people, client:OpenAI, model_name:str, max_tokens:int, temperature:float, stub:bool):
+async def infer_org_chart_llm(company, people, llm_provider:str, api_key:str, max_tokens:int, temperature:float, stub:bool=False, base_url:str=None):
     if stub:
         # Tiny fake org-chart when debugging offline
         chief = random.choice(people)
@@ -202,15 +202,20 @@ Here is a JSON list of employees:
 Return JSON: {{ "nodes":[{{id,name,title,dept,yoe_total,yoe_current,seniority_score,decision_score,avatar_url,profile_url}}], "edges":[{{source,target,type,confidence}}] }}
 """}
     ]
-    resp = client.chat.completions.create(
-        model=model_name,
+    resp = completion(
+        model=llm_provider,
         messages=prompt,
         max_tokens=max_tokens,
         temperature=temperature,
-        response_format={"type":"json_object"}
+        response_format={"type":"json_object"},
+        api_key=api_key,
+        base_url=base_url
     )
     chart = json.loads(resp.choices[0].message.content)
-    chart["meta"] = dict(model=model_name, generated_at=datetime.now(UTC).isoformat())
+    chart["meta"] = dict(
+        model=llm_provider,
+        generated_at=datetime.now(UTC).isoformat()
+    )
     return chart
 
 # ───────────────────────────────────────────────────────────────────────────────
@@ -270,15 +275,11 @@ async def run(opts):
     logging.info(f"[bold cyan]Loaded[/] {len(companies)} companies, {len(people)} people")
 
     logging.info("[bold]⇢[/] Embedding company descriptions…")
-    # embeds = embed_descriptions(companies, opts.embed_model, opts)
+    embeds = embed_descriptions(companies, opts.embed_model, opts)
     
     logging.info("[bold]⇢[/] Building similarity graph")
-    # company_graph = build_company_graph(companies, embeds, opts.top_k)
-    # dump_json(company_graph, out_dir/"company_graph.json")
-
-    # OpenAI client (only built if not debugging)
-    stub = bool(opts.stub)
-    client = OpenAI() if not stub else None
+    company_graph = build_company_graph(companies, embeds, opts.top_k)
+    dump_json(company_graph, out_dir/"company_graph.json")
 
     # Filter companies that need processing
     to_process = []
@@ -311,14 +312,14 @@ async def run(opts):
             async def process_one(comp):
                 handle = comp["handle"].strip("/").replace("/","_")
                 persons = [p for p in people if p["company_handle"].strip("/") == comp["handle"].strip("/")]
-                
                 chart = await infer_org_chart_llm(
                     comp, persons,
-                    client=client if client else OpenAI(api_key="sk-debug"),
-                    model_name=opts.openai_model,
+                    llm_provider=opts.llm_provider,
+                    api_key=opts.llm_api_key or None,
                     max_tokens=opts.max_llm_tokens,
                     temperature=opts.llm_temperature,
-                    stub=stub,
+                    stub=opts.stub or False,
+                    base_url=opts.llm_base_url or None
                 )
                 chart["meta"]["company"] = comp["name"]
                 
@@ -354,18 +355,22 @@ def build_arg_parser():
     p = argparse.ArgumentParser(description="Build graphs & visualisation from Stage-1 output")
     p.add_argument("--in",       dest="in_dir",  required=False, help="Stage-1 output dir", default=".")
     p.add_argument("--out",      dest="out_dir", required=False, help="Destination dir",   default=".")
-    p.add_argument("--embed_model", default="all-MiniLM-L6-v2")
-    p.add_argument("--top_k", type=int, default=10, help="Top-k neighbours per company")
-    p.add_argument("--openai_model", default="gpt-4.1")
-    p.add_argument("--max_llm_tokens", type=int, default=8024)
-    p.add_argument("--llm_temperature", type=float, default=1.0)
+    p.add_argument("--embed-model", default="all-MiniLM-L6-v2")
+    p.add_argument("--top-k", type=int, default=10, help="Top-k neighbours per company")
+    p.add_argument("--llm-provider", default="openai/gpt-4.1", 
+                  help="LLM model to use in format 'provider/model_name' (e.g., 'anthropic/claude-3')")
+    p.add_argument("--llm-api-key", help="API key for LLM provider (defaults to env vars)")
+    p.add_argument("--llm-base-url", help="Base URL for LLM API endpoint")
+    p.add_argument("--max-llm-tokens", type=int, default=8024)
+    p.add_argument("--llm-temperature", type=float, default=1.0)
     p.add_argument("--stub", action="store_true", help="Skip OpenAI call and generate tiny fake org charts")
     p.add_argument("--workers", type=int, default=4, help="Number of parallel workers for LLM inference")
     return p
 
 def main():
     dbg = dev_defaults()
-    opts = dbg if True else build_arg_parser().parse_args()
+    # opts = dbg if True else build_arg_parser().parse_args()
+    opts = build_arg_parser().parse_args()
     asyncio.run(run(opts))
 
 if __name__ == "__main__":

docs/examples/docker/demo_docker_polling.py

@@ -0,0 +1,149 @@
+
+#!/usr/bin/env python3
+"""
+demo_docker_polling.py
+Quick sanity-check for the asynchronous crawl job endpoints:
+
+  • POST  /crawl/job          – enqueue work, get task_id
+  • GET   /crawl/job/{id}     – poll status / fetch result
+
+The style matches demo_docker_api.py (console.rule banners, helper
+functions, coloured status lines).  Adjust BASE_URL as needed.
+
+Run:  python demo_docker_polling.py
+"""
+
+import asyncio, json, os, time, urllib.parse
+from typing import Dict, List
+
+import httpx
+from rich.console import Console
+from rich.panel   import Panel
+from rich.syntax  import Syntax
+
+console   = Console()
+BASE_URL  = os.getenv("BASE_URL", "http://localhost:11234")
+SIMPLE_URL = "https://example.org"
+LINKS_URL  = "https://httpbin.org/links/10/1"
+
+# --- helpers --------------------------------------------------------------
+
+
+def print_payload(payload: Dict):
+    console.print(Panel(Syntax(json.dumps(payload, indent=2),
+                               "json", theme="monokai", line_numbers=False),
+                        title="Payload", border_style="cyan", expand=False))
+
+
+async def check_server_health(client: httpx.AsyncClient) -> bool:
+    try:
+        resp = await client.get("/health")
+        if resp.is_success:
+            console.print("[green]Server healthy[/]")
+            return True
+    except Exception:
+        pass
+    console.print("[bold red]Server is not responding on /health[/]")
+    return False
+
+
+async def poll_for_result(client: httpx.AsyncClient, task_id: str,
+                          poll_interval: float = 1.5, timeout: float = 90.0):
+    """Hit /crawl/job/{id} until COMPLETED/FAILED or timeout."""
+    start = time.time()
+    while True:
+        resp = await client.get(f"/crawl/job/{task_id}")
+        resp.raise_for_status()
+        data = resp.json()
+        status = data.get("status")
+        if status.upper() in ("COMPLETED", "FAILED"):
+            return data
+        if time.time() - start > timeout:
+            raise TimeoutError(f"Task {task_id} did not finish in {timeout}s")
+        await asyncio.sleep(poll_interval)
+
+
+# --- demo functions -------------------------------------------------------
+
+
+async def demo_poll_single_url(client: httpx.AsyncClient):
+    payload = {
+        "urls": [SIMPLE_URL],
+        "browser_config": {"type": "BrowserConfig",
+                           "params": {"headless": True}},
+        "crawler_config": {"type": "CrawlerRunConfig",
+                           "params": {"cache_mode": "BYPASS"}}
+    }
+
+    console.rule("[bold blue]Demo A: /crawl/job Single URL[/]", style="blue")
+    print_payload(payload)
+
+    # enqueue
+    resp = await client.post("/crawl/job", json=payload)
+    console.print(f"Enqueue status: [bold]{resp.status_code}[/]")
+    resp.raise_for_status()
+    task_id = resp.json()["task_id"]
+    console.print(f"Task ID: [yellow]{task_id}[/]")
+
+    # poll
+    console.print("Polling…")
+    result = await poll_for_result(client, task_id)
+    console.print(Panel(Syntax(json.dumps(result, indent=2),
+                               "json", theme="fruity"),
+                        title="Final result", border_style="green"))
+    if result["status"] == "COMPLETED":
+        console.print("[green]✅ Crawl succeeded[/]")
+    else:
+        console.print("[red]❌ Crawl failed[/]")
+
+
+async def demo_poll_multi_url(client: httpx.AsyncClient):
+    payload = {
+        "urls": [SIMPLE_URL, LINKS_URL],
+        "browser_config": {"type": "BrowserConfig",
+                           "params": {"headless": True}},
+        "crawler_config": {"type": "CrawlerRunConfig",
+                           "params": {"cache_mode": "BYPASS"}}
+    }
+
+    console.rule("[bold magenta]Demo B: /crawl/job Multi-URL[/]",
+                 style="magenta")
+    print_payload(payload)
+
+    resp = await client.post("/crawl/job", json=payload)
+    console.print(f"Enqueue status: [bold]{resp.status_code}[/]")
+    resp.raise_for_status()
+    task_id = resp.json()["task_id"]
+    console.print(f"Task ID: [yellow]{task_id}[/]")
+
+    console.print("Polling…")
+    result = await poll_for_result(client, task_id)
+    console.print(Panel(Syntax(json.dumps(result, indent=2),
+                               "json", theme="fruity"),
+                        title="Final result", border_style="green"))
+    if result["status"] == "COMPLETED":
+        console.print(
+            f"[green]✅ {len(json.loads(result['result'])['results'])} URLs crawled[/]")
+    else:
+        console.print("[red]❌ Crawl failed[/]")
+
+
+# --- main runner ----------------------------------------------------------
+
+
+async def main_demo():
+    async with httpx.AsyncClient(base_url=BASE_URL, timeout=300.0) as client:
+        if not await check_server_health(client):
+            return
+        await demo_poll_single_url(client)
+        await demo_poll_multi_url(client)
+        console.rule("[bold green]Polling demos complete[/]", style="green")
+
+
+if __name__ == "__main__":
+    try:
+        asyncio.run(main_demo())
+    except KeyboardInterrupt:
+        console.print("\n[yellow]Interrupted by user[/]")
+    except Exception:
+        console.print_exception(show_locals=False)

docs/examples/hello_world.py

@@ -3,45 +3,24 @@ from crawl4ai import (
     AsyncWebCrawler,
     BrowserConfig,
     CrawlerRunConfig,
-    CacheMode,
     DefaultMarkdownGenerator,
     PruningContentFilter,
     CrawlResult
 )
 
-async def example_cdp():
-    browser_conf = BrowserConfig(
-        headless=False,
-        cdp_url="http://localhost:9223"
-    )
-    crawler_config = CrawlerRunConfig(
-        session_id="test",
-        js_code = """(() => { return {"result": "Hello World!"} })()""",
-        js_only=True
-    )
-    async with AsyncWebCrawler(
-        config=browser_conf,
-        verbose=True,
-    ) as crawler:
-        result : CrawlResult = await crawler.arun(
-            url="https://www.helloworld.org",
-            config=crawler_config,
-        )
-        print(result.js_execution_result)
-                   
 
 async def main():
-    browser_config = BrowserConfig(headless=False, verbose=True)
+    browser_config = BrowserConfig(
+        headless=False,
+        verbose=True,
+    )
     async with AsyncWebCrawler(config=browser_config) as crawler:
         crawler_config = CrawlerRunConfig(
-            cache_mode=CacheMode.BYPASS,
             markdown_generator=DefaultMarkdownGenerator(
-                content_filter=PruningContentFilter(
-                     threshold=0.48, threshold_type="fixed", min_word_threshold=0
-                )
+                content_filter=PruningContentFilter()
             ),
         )
-        result : CrawlResult = await crawler.arun(
+        result: CrawlResult = await crawler.arun(
             url="https://www.helloworld.org", config=crawler_config
         )
         print(result.markdown.raw_markdown[:500])

docs/examples/regex_extraction_quickstart.py

@@ -0,0 +1,143 @@
+# == File: regex_extraction_quickstart.py ==
+"""
+Mini–quick-start for RegexExtractionStrategy
+────────────────────────────────────────────
+3 bite-sized demos that parallel the style of *quickstart_examples_set_1.py*:
+
+1.  **Default catalog** – scrape a page and pull out e-mails / phones / URLs, etc.
+2.  **Custom pattern**  – add your own regex at instantiation time.
+3.  **LLM-assisted schema** – ask the model to write a pattern, cache it, then
+    run extraction _without_ further LLM calls.
+
+Run the whole thing with::
+
+    python regex_extraction_quickstart.py
+"""
+
+import os, json, asyncio
+from pathlib import Path
+from typing import List
+
+from crawl4ai import (
+    AsyncWebCrawler,
+    CrawlerRunConfig,
+    CrawlResult,
+    RegexExtractionStrategy,
+    LLMConfig,
+)
+
+# ────────────────────────────────────────────────────────────────────────────
+# 1. Default-catalog extraction
+# ────────────────────────────────────────────────────────────────────────────
+async def demo_regex_default() -> None:
+    print("\n=== 1. Regex extraction – default patterns ===")
+
+    url = "https://www.iana.org/domains/example"      # has e-mail + URLs
+    strategy = RegexExtractionStrategy(
+        pattern = RegexExtractionStrategy.Url | RegexExtractionStrategy.Currency
+    )               
+    config   = CrawlerRunConfig(extraction_strategy=strategy)
+
+    async with AsyncWebCrawler() as crawler:
+        result: CrawlResult = await crawler.arun(url, config=config)
+
+    print(f"Fetched {url} - success={result.success}")
+    if result.success:
+        data = json.loads(result.extracted_content)
+        for d in data[:10]:
+            print(f"  {d['label']:<12} {d['value']}")
+        print(f"... total matches: {len(data)}")
+    else:
+        print("  !!! crawl failed")
+
+
+# ────────────────────────────────────────────────────────────────────────────
+# 2. Custom pattern override / extension
+# ────────────────────────────────────────────────────────────────────────────
+async def demo_regex_custom() -> None:
+    print("\n=== 2. Regex extraction – custom price pattern ===")
+
+    url = "https://www.apple.com/shop/buy-mac/macbook-pro"
+    price_pattern = {"usd_price": r"\$\s?\d{1,3}(?:,\d{3})*(?:\.\d{2})?"}
+
+    strategy = RegexExtractionStrategy(custom = price_pattern)
+    config   = CrawlerRunConfig(extraction_strategy=strategy)
+
+    async with AsyncWebCrawler() as crawler:
+        result: CrawlResult = await crawler.arun(url, config=config)
+
+    if result.success:
+        data = json.loads(result.extracted_content)
+        for d in data:
+            print(f"  {d['value']}")
+        if not data:
+            print("  (No prices found - page layout may have changed)")
+    else:
+        print("  !!! crawl failed")
+
+
+# ────────────────────────────────────────────────────────────────────────────
+# 3. One-shot LLM pattern generation, then fast extraction
+# ────────────────────────────────────────────────────────────────────────────
+async def demo_regex_generate_pattern() -> None:
+    print("\n=== 3. generate_pattern → regex extraction ===")
+
+    cache_dir   = Path(__file__).parent / "tmp"
+    cache_dir.mkdir(exist_ok=True)
+    pattern_file = cache_dir / "price_pattern.json"
+
+    url = "https://www.lazada.sg/tag/smartphone/"
+
+    # ── 3-A. build or load the cached pattern
+    if pattern_file.exists():
+        pattern = json.load(pattern_file.open(encoding="utf-8"))
+        print("Loaded cached pattern:", pattern)
+    else:
+        print("Generating pattern via LLM…")
+
+        llm_cfg = LLMConfig(
+            provider="openai/gpt-4o-mini",
+            api_token="env:OPENAI_API_KEY",
+        )
+
+        # pull one sample page as HTML context
+        async with AsyncWebCrawler() as crawler:
+            html = (await crawler.arun(url)).fit_html 
+
+        pattern = RegexExtractionStrategy.generate_pattern(
+            label="price",
+            html=html,
+            query="Prices in Malaysian Ringgit (e.g. RM1,299.00 or RM200)",
+            llm_config=llm_cfg,
+        )
+
+        json.dump(pattern, pattern_file.open("w", encoding="utf-8"), indent=2)
+        print("Saved pattern:", pattern_file)
+
+    # ── 3-B. extraction pass – zero LLM calls
+    strategy = RegexExtractionStrategy(custom=pattern)
+    config   = CrawlerRunConfig(extraction_strategy=strategy, delay_before_return_html=3)
+
+    async with AsyncWebCrawler() as crawler:
+        result: CrawlResult = await crawler.arun(url, config=config)
+
+    if result.success:
+        data = json.loads(result.extracted_content)
+        for d in data[:15]:
+            print(f"  {d['value']}")
+        print(f"... total matches: {len(data)}")
+    else:
+        print("  !!! crawl failed")
+
+
+# ────────────────────────────────────────────────────────────────────────────
+# Entrypoint
+# ────────────────────────────────────────────────────────────────────────────
+async def main() -> None:
+    # await demo_regex_default()
+    # await demo_regex_custom()
+    await demo_regex_generate_pattern()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

docs/examples/session_id_example.py

@@ -0,0 +1,38 @@
+import asyncio
+from crawl4ai import (
+    AsyncWebCrawler,
+    BrowserConfig,
+    CrawlerRunConfig,
+    DefaultMarkdownGenerator,
+    PruningContentFilter,
+    CrawlResult
+)
+
+                   
+
+async def main():    
+    browser_config = BrowserConfig(
+        headless=False, 
+        verbose=True,
+    )
+    async with AsyncWebCrawler(config=browser_config) as crawler:
+        crawler_config = CrawlerRunConfig(
+            session_id= "hello_world", # This help us to use the same page 
+        )
+        result : CrawlResult = await crawler.arun(
+            url="https://www.helloworld.org", config=crawler_config
+        )
+        # Add a breakpoint here, then you will the page is open and browser is not closed
+        print(result.markdown.raw_markdown[:500])
+        
+        new_config = crawler_config.clone(js_code=["(() => ({'data':'hello'}))()"], js_only=True)
+        result : CrawlResult = await crawler.arun( # This time there is no fetch and this only executes JS in the same opened page
+            url="https://www.helloworld.org", config= new_config
+        )
+        print(result.js_execution_result) # You should see {'data':'hello'} in the console
+        
+        # Get direct access to Playwright paege object. This works only if you use the same session_id and pass same config 
+        page, context = crawler.crawler_strategy.get_page(new_config)
+
+if __name__ == "__main__":
+    asyncio.run(main())

docs/md_v2/api/crawl-result.md

@@ -10,6 +10,7 @@ class CrawlResult(BaseModel):
     html: str
     success: bool
     cleaned_html: Optional[str] = None
+    fit_html: Optional[str] = None  # Preprocessed HTML optimized for extraction
     media: Dict[str, List[Dict]] = {}
     links: Dict[str, List[Dict]] = {}
     downloaded_files: Optional[List[str]] = None
@@ -50,7 +51,7 @@ if not result.success:
 ```
 
 ### 1.3 **`status_code`** *(Optional[int])*  
-**What**: The page’s HTTP status code (e.g., 200, 404).  
+**What**: The page's HTTP status code (e.g., 200, 404).  
 **Usage**:
 ```python
 if result.status_code == 404:
@@ -82,7 +83,7 @@ if result.response_headers:
 ```
 
 ### 1.7 **`ssl_certificate`** *(Optional[SSLCertificate])*  
-**What**: If `fetch_ssl_certificate=True` in your CrawlerRunConfig, **`result.ssl_certificate`** contains a  [**`SSLCertificate`**](../advanced/ssl-certificate.md) object describing the site’s certificate. You can export the cert in multiple formats (PEM/DER/JSON) or access its properties like `issuer`, 
+**What**: If `fetch_ssl_certificate=True` in your CrawlerRunConfig, **`result.ssl_certificate`** contains a  [**`SSLCertificate`**](../advanced/ssl-certificate.md) object describing the site's certificate. You can export the cert in multiple formats (PEM/DER/JSON) or access its properties like `issuer`, 
  `subject`, `valid_from`, `valid_until`, etc. 
 **Usage**:
 ```python
@@ -109,14 +110,6 @@ print(len(result.html))
 print(result.cleaned_html[:500])  # Show a snippet
 ```
 
-### 2.3 **`fit_html`** *(Optional[str])*  
-**What**: If a **content filter** or heuristic (e.g., Pruning/BM25) modifies the HTML, the “fit” or post-filter version.  
-**When**: This is **only** present if your `markdown_generator` or `content_filter` produces it.  
-**Usage**:
-```python
-if result.markdown.fit_html:
-    print("High-value HTML content:", result.markdown.fit_html[:300])
-```
 
 ---
 
@@ -135,7 +128,7 @@ Crawl4AI can convert HTML→Markdown, optionally including:
 - **`raw_markdown`** *(str)*: The full HTML→Markdown conversion.  
 - **`markdown_with_citations`** *(str)*: Same markdown, but with link references as academic-style citations.  
 - **`references_markdown`** *(str)*: The reference list or footnotes at the end.  
-- **`fit_markdown`** *(Optional[str])*: If content filtering (Pruning/BM25) was applied, the filtered “fit” text.  
+- **`fit_markdown`** *(Optional[str])*: If content filtering (Pruning/BM25) was applied, the filtered "fit" text.  
 - **`fit_html`** *(Optional[str])*: The HTML that led to `fit_markdown`.
 
 **Usage**:
@@ -157,7 +150,7 @@ print(result.markdown.raw_markdown[:200])
 print(result.markdown.fit_markdown)
 print(result.markdown.fit_html)
 ```
-**Important**: “Fit” content (in `fit_markdown`/`fit_html`) exists in result.markdown, only if you used a **filter** (like **PruningContentFilter** or **BM25ContentFilter**) within a `MarkdownGenerationStrategy`.
+**Important**: "Fit" content (in `fit_markdown`/`fit_html`) exists in result.markdown, only if you used a **filter** (like **PruningContentFilter** or **BM25ContentFilter**) within a `MarkdownGenerationStrategy`.
 
 ---
 
@@ -169,7 +162,7 @@ print(result.markdown.fit_html)
 
 - `src` *(str)*: Media URL  
 - `alt` or `title` *(str)*: Descriptive text  
-- `score` *(float)*: Relevance score if the crawler’s heuristic found it “important”  
+- `score` *(float)*: Relevance score if the crawler's heuristic found it "important"  
 - `desc` or `description` *(Optional[str])*: Additional context extracted from surrounding text  
 
 **Usage**:
@@ -263,7 +256,7 @@ A `DispatchResult` object providing additional concurrency and resource usage in
 
 - **`task_id`**: A unique identifier for the parallel task.
 - **`memory_usage`** (float): The memory (in MB) used at the time of completion.
-- **`peak_memory`** (float): The peak memory usage (in MB) recorded during the task’s execution.
+- **`peak_memory`** (float): The peak memory usage (in MB) recorded during the task's execution.
 - **`start_time`** / **`end_time`** (datetime): Time range for this crawling task.
 - **`error_message`** (str): Any dispatcher- or concurrency-related error encountered.
 
@@ -358,7 +351,7 @@ async def handle_result(result: CrawlResult):
     # HTML
     print("Original HTML size:", len(result.html))
     print("Cleaned HTML size:", len(result.cleaned_html or ""))
-
+    
     # Markdown output
     if result.markdown:
         print("Raw Markdown:", result.markdown.raw_markdown[:300])

docs/md_v2/api/strategies.md

@@ -36,6 +36,45 @@ LLMExtractionStrategy(
 )
 ```
 
+### RegexExtractionStrategy
+
+Used for fast pattern-based extraction of common entities using regular expressions.
+
+```python
+RegexExtractionStrategy(
+    # Pattern Configuration
+    pattern: IntFlag = RegexExtractionStrategy.Nothing,  # Bit flags of built-in patterns to use
+    custom: Optional[Dict[str, str]] = None,           # Custom pattern dictionary {label: regex}
+    
+    # Input Format
+    input_format: str = "fit_html",                    # "html", "markdown", "text" or "fit_html"
+)
+
+# Built-in Patterns as Bit Flags
+RegexExtractionStrategy.Email           # Email addresses
+RegexExtractionStrategy.PhoneIntl       # International phone numbers 
+RegexExtractionStrategy.PhoneUS         # US-format phone numbers
+RegexExtractionStrategy.Url             # HTTP/HTTPS URLs
+RegexExtractionStrategy.IPv4            # IPv4 addresses
+RegexExtractionStrategy.IPv6            # IPv6 addresses
+RegexExtractionStrategy.Uuid            # UUIDs
+RegexExtractionStrategy.Currency        # Currency values (USD, EUR, etc)
+RegexExtractionStrategy.Percentage      # Percentage values
+RegexExtractionStrategy.Number          # Numeric values
+RegexExtractionStrategy.DateIso         # ISO format dates
+RegexExtractionStrategy.DateUS          # US format dates
+RegexExtractionStrategy.Time24h         # 24-hour format times
+RegexExtractionStrategy.PostalUS        # US postal codes
+RegexExtractionStrategy.PostalUK        # UK postal codes
+RegexExtractionStrategy.HexColor        # HTML hex color codes
+RegexExtractionStrategy.TwitterHandle   # Twitter handles
+RegexExtractionStrategy.Hashtag         # Hashtags
+RegexExtractionStrategy.MacAddr         # MAC addresses
+RegexExtractionStrategy.Iban            # International bank account numbers
+RegexExtractionStrategy.CreditCard      # Credit card numbers
+RegexExtractionStrategy.All             # All available patterns
+```
+
 ### CosineStrategy
 
 Used for content similarity-based extraction and clustering.
@@ -156,6 +195,55 @@ result = await crawler.arun(
 data = json.loads(result.extracted_content)
 ```
 
+### Regex Extraction
+
+```python
+import json
+from crawl4ai import AsyncWebCrawler, CrawlerRunConfig, RegexExtractionStrategy
+
+# Method 1: Use built-in patterns
+strategy = RegexExtractionStrategy(
+    pattern = RegexExtractionStrategy.Email | RegexExtractionStrategy.Url
+)
+
+# Method 2: Use custom patterns
+price_pattern = {"usd_price": r"\$\s?\d{1,3}(?:,\d{3})*(?:\.\d{2})?"}
+strategy = RegexExtractionStrategy(custom=price_pattern)
+
+# Method 3: Generate pattern with LLM assistance (one-time)
+from crawl4ai import LLMConfig
+
+async with AsyncWebCrawler() as crawler:
+    # Get sample HTML first
+    sample_result = await crawler.arun("https://example.com/products")
+    html = sample_result.fit_html
+    
+    # Generate regex pattern once
+    pattern = RegexExtractionStrategy.generate_pattern(
+        label="price",
+        html=html,
+        query="Product prices in USD format",
+        llm_config=LLMConfig(provider="openai/gpt-4o-mini")
+    )
+    
+    # Save pattern for reuse
+    import json
+    with open("price_pattern.json", "w") as f:
+        json.dump(pattern, f)
+    
+    # Use pattern for extraction (no LLM calls)
+    strategy = RegexExtractionStrategy(custom=pattern)
+    result = await crawler.arun(
+        url="https://example.com/products",
+        config=CrawlerRunConfig(extraction_strategy=strategy)
+    )
+    
+    # Process results
+    data = json.loads(result.extracted_content)
+    for item in data:
+        print(f"{item['label']}: {item['value']}")
+```
+
 ### CSS Extraction
 
 ```python
@@ -220,12 +308,28 @@ result = await crawler.arun(
 
 ## Best Practices
 
-1. **Choose the Right Strategy**
-   - Use `LLMExtractionStrategy` for complex, unstructured content
-   - Use `JsonCssExtractionStrategy` for well-structured HTML
+1. **Choose the Right Strategy**
+   - Use `RegexExtractionStrategy` for common data types like emails, phones, URLs, dates
+   - Use `JsonCssExtractionStrategy` for well-structured HTML with consistent patterns
+   - Use `LLMExtractionStrategy` for complex, unstructured content requiring reasoning
    - Use `CosineStrategy` for content similarity and clustering
 
-2. **Optimize Chunking**
+2. **Strategy Selection Guide**
+   ```
+   Is the target data a common type (email/phone/date/URL)? 
+   → RegexExtractionStrategy
+   
+   Does the page have consistent HTML structure?
+   → JsonCssExtractionStrategy or JsonXPathExtractionStrategy
+   
+   Is the data semantically complex or unstructured?
+   → LLMExtractionStrategy
+   
+   Need to find content similar to a specific topic?
+   → CosineStrategy
+   ```
+
+3. **Optimize Chunking**
    ```python
    # For long documents
    strategy = LLMExtractionStrategy(
@@ -234,7 +338,26 @@ result = await crawler.arun(
    )
    ```
 
-3. **Handle Errors**
+4. **Combine Strategies for Best Performance**
+   ```python
+   # First pass: Extract structure with CSS
+   css_strategy = JsonCssExtractionStrategy(product_schema)
+   css_result = await crawler.arun(url, config=CrawlerRunConfig(extraction_strategy=css_strategy))
+   product_data = json.loads(css_result.extracted_content)
+   
+   # Second pass: Extract specific fields with regex
+   descriptions = [product["description"] for product in product_data]
+   regex_strategy = RegexExtractionStrategy(
+       pattern=RegexExtractionStrategy.Email | RegexExtractionStrategy.PhoneUS,
+       custom={"dimension": r"\d+x\d+x\d+ (?:cm|in)"}
+   )
+   
+   # Process descriptions with regex
+   for text in descriptions:
+       matches = regex_strategy.extract("", text)  # Direct extraction
+   ```
+
+5. **Handle Errors**
    ```python
    try:
        result = await crawler.arun(
@@ -247,11 +370,31 @@ result = await crawler.arun(
        print(f"Extraction failed: {e}")
    ```
 
-4. **Monitor Performance**
+6. **Monitor Performance**
    ```python
    strategy = CosineStrategy(
        verbose=True,  # Enable logging
        word_count_threshold=20,  # Filter short content
        top_k=5  # Limit results
    )
+   ```
+
+7. **Cache Generated Patterns**
+   ```python
+   # For RegexExtractionStrategy pattern generation
+   import json
+   from pathlib import Path
+   
+   cache_dir = Path("./pattern_cache")
+   cache_dir.mkdir(exist_ok=True)
+   pattern_file = cache_dir / "product_pattern.json"
+   
+   if pattern_file.exists():
+       with open(pattern_file) as f:
+           pattern = json.load(f)
+   else:
+       # Generate once with LLM
+       pattern = RegexExtractionStrategy.generate_pattern(...)
+       with open(pattern_file, "w") as f:
+           json.dump(pattern, f)
    ```

docs/md_v2/extraction/no-llm-strategies.md

@@ -1,15 +1,20 @@
 # Extracting JSON (No LLM)
 
-One of Crawl4AI’s **most powerful** features is extracting **structured JSON** from websites **without** relying on large language models. By defining a **schema** with CSS or XPath selectors, you can extract data instantly—even from complex or nested HTML structures—without the cost, latency, or environmental impact of an LLM.
+One of Crawl4AI's **most powerful** features is extracting **structured JSON** from websites **without** relying on large language models. Crawl4AI offers several strategies for LLM-free extraction:
+
+1. **Schema-based extraction** with CSS or XPath selectors via `JsonCssExtractionStrategy` and `JsonXPathExtractionStrategy`
+2. **Regular expression extraction** with `RegexExtractionStrategy` for fast pattern matching
+
+These approaches let you extract data instantly—even from complex or nested HTML structures—without the cost, latency, or environmental impact of an LLM.
 
 **Why avoid LLM for basic extractions?**
 
-1. **Faster & Cheaper**: No API calls or GPU overhead.  
-2. **Lower Carbon Footprint**: LLM inference can be energy-intensive. A well-defined schema is practically carbon-free.  
-3. **Precise & Repeatable**: CSS/XPath selectors do exactly what you specify. LLM outputs can vary or hallucinate.  
-4. **Scales Readily**: For thousands of pages, schema-based extraction runs quickly and in parallel.
+1. **Faster & Cheaper**: No API calls or GPU overhead.  
+2. **Lower Carbon Footprint**: LLM inference can be energy-intensive. Pattern-based extraction is practically carbon-free.  
+3. **Precise & Repeatable**: CSS/XPath selectors and regex patterns do exactly what you specify. LLM outputs can vary or hallucinate.  
+4. **Scales Readily**: For thousands of pages, pattern-based extraction runs quickly and in parallel.
 
-Below, we’ll explore how to craft these schemas and use them with **JsonCssExtractionStrategy** (or **JsonXPathExtractionStrategy** if you prefer XPath). We’ll also highlight advanced features like **nested fields** and **base element attributes**.
+Below, we'll explore how to craft these schemas and use them with **JsonCssExtractionStrategy** (or **JsonXPathExtractionStrategy** if you prefer XPath). We'll also highlight advanced features like **nested fields** and **base element attributes**.
 
 ---
 
@@ -17,17 +22,17 @@ Below, we’ll explore how to craft these schemas and use them with **JsonCssExt
 
 A schema defines:
 
-1. A **base selector** that identifies each “container” element on the page (e.g., a product row, a blog post card).  
-2. **Fields** describing which CSS/XPath selectors to use for each piece of data you want to capture (text, attribute, HTML block, etc.).  
-3. **Nested** or **list** types for repeated or hierarchical structures.  
+1. A **base selector** that identifies each "container" element on the page (e.g., a product row, a blog post card).  
+2. **Fields** describing which CSS/XPath selectors to use for each piece of data you want to capture (text, attribute, HTML block, etc.).  
+3. **Nested** or **list** types for repeated or hierarchical structures.  
 
-For example, if you have a list of products, each one might have a name, price, reviews, and “related products.” This approach is faster and more reliable than an LLM for consistent, structured pages.
+For example, if you have a list of products, each one might have a name, price, reviews, and "related products." This approach is faster and more reliable than an LLM for consistent, structured pages.
 
 ---
 
 ## 2. Simple Example: Crypto Prices
 
-Let’s begin with a **simple** schema-based extraction using the `JsonCssExtractionStrategy`. Below is a snippet that extracts cryptocurrency prices from a site (similar to the legacy Coinbase example). Notice we **don’t** call any LLM:
+Let's begin with a **simple** schema-based extraction using the `JsonCssExtractionStrategy`. Below is a snippet that extracts cryptocurrency prices from a site (similar to the legacy Coinbase example). Notice we **don't** call any LLM:
 
 ```python
 import json
@@ -87,7 +92,7 @@ asyncio.run(extract_crypto_prices())
 
 **Highlights**:
 
-- **`baseSelector`**: Tells us where each “item” (crypto row) is.  
+- **`baseSelector`**: Tells us where each "item" (crypto row) is.  
 - **`fields`**: Two fields (`coin_name`, `price`) using simple CSS selectors.  
 - Each field defines a **`type`** (e.g., `text`, `attribute`, `html`, `regex`, etc.).
 
@@ -97,7 +102,7 @@ No LLM is needed, and the performance is **near-instant** for hundreds or thousa
 
 ### **XPath Example with `raw://` HTML**
 
-Below is a short example demonstrating **XPath** extraction plus the **`raw://`** scheme. We’ll pass a **dummy HTML** directly (no network request) and define the extraction strategy in `CrawlerRunConfig`.
+Below is a short example demonstrating **XPath** extraction plus the **`raw://`** scheme. We'll pass a **dummy HTML** directly (no network request) and define the extraction strategy in `CrawlerRunConfig`.
 
 ```python
 import json
@@ -168,12 +173,12 @@ asyncio.run(extract_crypto_prices_xpath())
 
 **Key Points**:
 
-1. **`JsonXPathExtractionStrategy`** is used instead of `JsonCssExtractionStrategy`.  
-2. **`baseSelector`** and each field’s `"selector"` use **XPath** instead of CSS.  
-3. **`raw://`** lets us pass `dummy_html` with no real network request—handy for local testing.  
+1. **`JsonXPathExtractionStrategy`** is used instead of `JsonCssExtractionStrategy`.  
+2. **`baseSelector`** and each field's `"selector"` use **XPath** instead of CSS.  
+3. **`raw://`** lets us pass `dummy_html` with no real network request—handy for local testing.  
 4. Everything (including the extraction strategy) is in **`CrawlerRunConfig`**.  
 
-That’s how you keep the config self-contained, illustrate **XPath** usage, and demonstrate the **raw** scheme for direct HTML input—all while avoiding the old approach of passing `extraction_strategy` directly to `arun()`.
+That's how you keep the config self-contained, illustrate **XPath** usage, and demonstrate the **raw** scheme for direct HTML input—all while avoiding the old approach of passing `extraction_strategy` directly to `arun()`.
 
 ---
 
@@ -187,7 +192,7 @@ We have a **sample e-commerce** HTML file on GitHub (example):
 ```
 https://gist.githubusercontent.com/githubusercontent/2d7b8ba3cd8ab6cf3c8da771ddb36878/raw/1ae2f90c6861ce7dd84cc50d3df9920dee5e1fd2/sample_ecommerce.html
 ```
-This snippet includes categories, products, features, reviews, and related items. Let’s see how to define a schema that fully captures that structure **without LLM**.
+This snippet includes categories, products, features, reviews, and related items. Let's see how to define a schema that fully captures that structure **without LLM**.
 
 ```python
 schema = {
@@ -333,24 +338,253 @@ async def extract_ecommerce_data():
 asyncio.run(extract_ecommerce_data())
 ```
 
-If all goes well, you get a **structured** JSON array with each “category,” containing an array of `products`. Each product includes `details`, `features`, `reviews`, etc. All of that **without** an LLM.
+If all goes well, you get a **structured** JSON array with each "category," containing an array of `products`. Each product includes `details`, `features`, `reviews`, etc. All of that **without** an LLM.
 
 ---
 
-## 4. Why “No LLM” Is Often Better
+## 4. RegexExtractionStrategy - Fast Pattern-Based Extraction
 
-1. **Zero Hallucination**: Schema-based extraction doesn’t guess text. It either finds it or not.  
-2. **Guaranteed Structure**: The same schema yields consistent JSON across many pages, so your downstream pipeline can rely on stable keys.  
-3. **Speed**: LLM-based extraction can be 10–1000x slower for large-scale crawling.  
-4. **Scalable**: Adding or updating a field is a matter of adjusting the schema, not re-tuning a model.
+Crawl4AI now offers a powerful new zero-LLM extraction strategy: `RegexExtractionStrategy`. This strategy provides lightning-fast extraction of common data types like emails, phone numbers, URLs, dates, and more using pre-compiled regular expressions.
 
-**When might you consider an LLM?** Possibly if the site is extremely unstructured or you want AI summarization. But always try a schema approach first for repeated or consistent data patterns.
+### Key Features
+
+- **Zero LLM Dependency**: Extracts data without any AI model calls
+- **Blazing Fast**: Uses pre-compiled regex patterns for maximum performance
+- **Built-in Patterns**: Includes ready-to-use patterns for common data types
+- **Custom Patterns**: Add your own regex patterns for domain-specific extraction
+- **LLM-Assisted Pattern Generation**: Optionally use an LLM once to generate optimized patterns, then reuse them without further LLM calls
+
+### Simple Example: Extracting Common Entities
+
+The easiest way to start is by using the built-in pattern catalog:
+
+```python
+import json
+import asyncio
+from crawl4ai import (
+    AsyncWebCrawler,
+    CrawlerRunConfig,
+    RegexExtractionStrategy
+)
+
+async def extract_with_regex():
+    # Create a strategy using built-in patterns for URLs and currencies
+    strategy = RegexExtractionStrategy(
+        pattern = RegexExtractionStrategy.Url | RegexExtractionStrategy.Currency
+    )
+    
+    config = CrawlerRunConfig(extraction_strategy=strategy)
+    
+    async with AsyncWebCrawler() as crawler:
+        result = await crawler.arun(
+            url="https://example.com",
+            config=config
+        )
+        
+        if result.success:
+            data = json.loads(result.extracted_content)
+            for item in data[:5]:  # Show first 5 matches
+                print(f"{item['label']}: {item['value']}")
+            print(f"Total matches: {len(data)}")
+
+asyncio.run(extract_with_regex())
+```
+
+### Available Built-in Patterns
+
+`RegexExtractionStrategy` provides these common patterns as IntFlag attributes for easy combining:
+
+```python
+# Use individual patterns
+strategy = RegexExtractionStrategy(pattern=RegexExtractionStrategy.Email)
+
+# Combine multiple patterns
+strategy = RegexExtractionStrategy(
+    pattern = (
+        RegexExtractionStrategy.Email | 
+        RegexExtractionStrategy.PhoneUS | 
+        RegexExtractionStrategy.Url
+    )
+)
+
+# Use all available patterns
+strategy = RegexExtractionStrategy(pattern=RegexExtractionStrategy.All)
+```
+
+Available patterns include:
+- `Email` - Email addresses
+- `PhoneIntl` - International phone numbers
+- `PhoneUS` - US-format phone numbers
+- `Url` - HTTP/HTTPS URLs
+- `IPv4` - IPv4 addresses
+- `IPv6` - IPv6 addresses
+- `Uuid` - UUIDs
+- `Currency` - Currency values (USD, EUR, etc.)
+- `Percentage` - Percentage values
+- `Number` - Numeric values
+- `DateIso` - ISO format dates
+- `DateUS` - US format dates
+- `Time24h` - 24-hour format times
+- `PostalUS` - US postal codes
+- `PostalUK` - UK postal codes
+- `HexColor` - HTML hex color codes
+- `TwitterHandle` - Twitter handles
+- `Hashtag` - Hashtags
+- `MacAddr` - MAC addresses
+- `Iban` - International bank account numbers
+- `CreditCard` - Credit card numbers
+
+### Custom Pattern Example
+
+For more targeted extraction, you can provide custom patterns:
+
+```python
+import json
+import asyncio
+from crawl4ai import (
+    AsyncWebCrawler,
+    CrawlerRunConfig,
+    RegexExtractionStrategy
+)
+
+async def extract_prices():
+    # Define a custom pattern for US Dollar prices
+    price_pattern = {"usd_price": r"\$\s?\d{1,3}(?:,\d{3})*(?:\.\d{2})?"}
+    
+    # Create strategy with custom pattern
+    strategy = RegexExtractionStrategy(custom=price_pattern)
+    config = CrawlerRunConfig(extraction_strategy=strategy)
+    
+    async with AsyncWebCrawler() as crawler:
+        result = await crawler.arun(
+            url="https://www.example.com/products",
+            config=config
+        )
+        
+        if result.success:
+            data = json.loads(result.extracted_content)
+            for item in data:
+                print(f"Found price: {item['value']}")
+
+asyncio.run(extract_prices())
+```
+
+### LLM-Assisted Pattern Generation
+
+For complex or site-specific patterns, you can use an LLM once to generate an optimized pattern, then save and reuse it without further LLM calls:
+
+```python
+import json
+import asyncio
+from pathlib import Path
+from crawl4ai import (
+    AsyncWebCrawler,
+    CrawlerRunConfig,
+    RegexExtractionStrategy,
+    LLMConfig
+)
+
+async def extract_with_generated_pattern():
+    cache_dir = Path("./pattern_cache")
+    cache_dir.mkdir(exist_ok=True)
+    pattern_file = cache_dir / "price_pattern.json"
+    
+    # 1. Generate or load pattern
+    if pattern_file.exists():
+        pattern = json.load(pattern_file.open())
+        print(f"Using cached pattern: {pattern}")
+    else:
+        print("Generating pattern via LLM...")
+        
+        # Configure LLM
+        llm_config = LLMConfig(
+            provider="openai/gpt-4o-mini",
+            api_token="env:OPENAI_API_KEY",
+        )
+        
+        # Get sample HTML for context
+        async with AsyncWebCrawler() as crawler:
+            result = await crawler.arun("https://example.com/products")
+            html = result.fit_html
+        
+        # Generate pattern (one-time LLM usage)
+        pattern = RegexExtractionStrategy.generate_pattern(
+            label="price",
+            html=html,
+            query="Product prices in USD format",
+            llm_config=llm_config,
+        )
+        
+        # Cache pattern for future use
+        json.dump(pattern, pattern_file.open("w"), indent=2)
+    
+    # 2. Use pattern for extraction (no LLM calls)
+    strategy = RegexExtractionStrategy(custom=pattern)
+    config = CrawlerRunConfig(extraction_strategy=strategy)
+    
+    async with AsyncWebCrawler() as crawler:
+        result = await crawler.arun(
+            url="https://example.com/products",
+            config=config
+        )
+        
+        if result.success:
+            data = json.loads(result.extracted_content)
+            for item in data[:10]:
+                print(f"Extracted: {item['value']}")
+            print(f"Total matches: {len(data)}")
+
+asyncio.run(extract_with_generated_pattern())
+```
+
+This pattern allows you to:
+1. Use an LLM once to generate a highly optimized regex for your specific site
+2. Save the pattern to disk for reuse 
+3. Extract data using only regex (no further LLM calls) in production
+
+### Extraction Results Format
+
+The `RegexExtractionStrategy` returns results in a consistent format:
+
+```json
+[
+  {
+    "url": "https://example.com",
+    "label": "email",
+    "value": "contact@example.com",
+    "span": [145, 163]
+  },
+  {
+    "url": "https://example.com",
+    "label": "url",
+    "value": "https://support.example.com",
+    "span": [210, 235]
+  }
+]
+```
+
+Each match includes:
+- `url`: The source URL
+- `label`: The pattern name that matched (e.g., "email", "phone_us")
+- `value`: The extracted text
+- `span`: The start and end positions in the source content
 
 ---
 
-## 5. Base Element Attributes & Additional Fields
+## 5. Why "No LLM" Is Often Better
 
-It’s easy to **extract attributes** (like `href`, `src`, or `data-xxx`) from your base or nested elements using:
+1. **Zero Hallucination**: Pattern-based extraction doesn't guess text. It either finds it or not.  
+2. **Guaranteed Structure**: The same schema or regex yields consistent JSON across many pages, so your downstream pipeline can rely on stable keys.  
+3. **Speed**: LLM-based extraction can be 10–1000x slower for large-scale crawling.  
+4. **Scalable**: Adding or updating a field is a matter of adjusting the schema or regex, not re-tuning a model.
+
+**When might you consider an LLM?** Possibly if the site is extremely unstructured or you want AI summarization. But always try a schema or regex approach first for repeated or consistent data patterns.
+
+---
+
+## 6. Base Element Attributes & Additional Fields
+
+It's easy to **extract attributes** (like `href`, `src`, or `data-xxx`) from your base or nested elements using:
 
 ```json
 {
@@ -361,11 +595,11 @@ It’s easy to **extract attributes** (like `href`, `src`, or `data-xxx`) from y
 }
 ```
 
-You can define them in **`baseFields`** (extracted from the main container element) or in each field’s sub-lists. This is especially helpful if you need an item’s link or ID stored in the parent `<div>`.
+You can define them in **`baseFields`** (extracted from the main container element) or in each field's sub-lists. This is especially helpful if you need an item's link or ID stored in the parent `<div>`.
 
 ---
 
-## 6. Putting It All Together: Larger Example
+## 7. Putting It All Together: Larger Example
 
 Consider a blog site. We have a schema that extracts the **URL** from each post card (via `baseFields` with an `"attribute": "href"`), plus the title, date, summary, and author:
 
@@ -389,19 +623,20 @@ Then run with `JsonCssExtractionStrategy(schema)` to get an array of blog post o
 
 ---
 
-## 7. Tips & Best Practices
+## 8. Tips & Best Practices
 
-1. **Inspect the DOM** in Chrome DevTools or Firefox’s Inspector to find stable selectors.  
-2. **Start Simple**: Verify you can extract a single field. Then add complexity like nested objects or lists.  
-3. **Test** your schema on partial HTML or a test page before a big crawl.  
-4. **Combine with JS Execution** if the site loads content dynamically. You can pass `js_code` or `wait_for` in `CrawlerRunConfig`.  
-5. **Look at Logs** when `verbose=True`: if your selectors are off or your schema is malformed, it’ll often show warnings.  
-6. **Use baseFields** if you need attributes from the container element (e.g., `href`, `data-id`), especially for the “parent” item.  
-7. **Performance**: For large pages, make sure your selectors are as narrow as possible.
+1. **Inspect the DOM** in Chrome DevTools or Firefox's Inspector to find stable selectors.  
+2. **Start Simple**: Verify you can extract a single field. Then add complexity like nested objects or lists.  
+3. **Test** your schema on partial HTML or a test page before a big crawl.  
+4. **Combine with JS Execution** if the site loads content dynamically. You can pass `js_code` or `wait_for` in `CrawlerRunConfig`.  
+5. **Look at Logs** when `verbose=True`: if your selectors are off or your schema is malformed, it'll often show warnings.  
+6. **Use baseFields** if you need attributes from the container element (e.g., `href`, `data-id`), especially for the "parent" item.  
+7. **Performance**: For large pages, make sure your selectors are as narrow as possible.
+8. **Consider Using Regex First**: For simple data types like emails, URLs, and dates, `RegexExtractionStrategy` is often the fastest approach.
 
 ---
 
-## 8. Schema Generation Utility
+## 9. Schema Generation Utility
 
 While manually crafting schemas is powerful and precise, Crawl4AI now offers a convenient utility to **automatically generate** extraction schemas using LLM. This is particularly useful when:
 
@@ -481,27 +716,26 @@ strategy = JsonCssExtractionStrategy(css_schema)
    - Use OpenAI for production-quality schemas
    - Use Ollama for development, testing, or when you need a self-hosted solution
 
-That's it for **Extracting JSON (No LLM)**! You've seen how schema-based approaches (either CSS or XPath) can handle everything from simple lists to deeply nested product catalogs—instantly, with minimal overhead. Enjoy building robust scrapers that produce consistent, structured JSON for your data pipelines!
-
 ---
 
-## 9. Conclusion
+## 10. Conclusion
 
-With **JsonCssExtractionStrategy** (or **JsonXPathExtractionStrategy**), you can build powerful, **LLM-free** pipelines that:
+With Crawl4AI's LLM-free extraction strategies - `JsonCssExtractionStrategy`, `JsonXPathExtractionStrategy`, and now `RegexExtractionStrategy` - you can build powerful pipelines that:
 
 - Scrape any consistent site for structured data.  
-- Support nested objects, repeating lists, or advanced transformations.  
+- Support nested objects, repeating lists, or pattern-based extraction.  
 - Scale to thousands of pages quickly and reliably.
 
-**Next Steps**:
+**Choosing the Right Strategy**:
 
-- Combine your extracted JSON with advanced filtering or summarization in a second pass if needed.  
-- For dynamic pages, combine strategies with `js_code` or infinite scroll hooking to ensure all content is loaded.
+- Use **`RegexExtractionStrategy`** for fast extraction of common data types like emails, phones, URLs, dates, etc.
+- Use **`JsonCssExtractionStrategy`** or **`JsonXPathExtractionStrategy`** for structured data with clear HTML patterns
+- If you need both: first extract structured data with JSON strategies, then use regex on specific fields
 
-**Remember**: For repeated, structured data, you don’t need to pay for or wait on an LLM. A well-crafted schema plus CSS or XPath gets you the data faster, cleaner, and cheaper—**the real power** of Crawl4AI.
+**Remember**: For repeated, structured data, you don't need to pay for or wait on an LLM. Well-crafted schemas and regex patterns get you the data faster, cleaner, and cheaper—**the real power** of Crawl4AI.
 
-**Last Updated**: 2025-01-01
+**Last Updated**: 2025-05-02
 
 ---
 
-That’s it for **Extracting JSON (No LLM)**! You’ve seen how schema-based approaches (either CSS or XPath) can handle everything from simple lists to deeply nested product catalogs—instantly, with minimal overhead. Enjoy building robust scrapers that produce consistent, structured JSON for your data pipelines!
+That's it for **Extracting JSON (No LLM)**! You've seen how schema-based approaches (either CSS or XPath) and regex patterns can handle everything from simple lists to deeply nested product catalogs—instantly, with minimal overhead. Enjoy building robust scrapers that produce consistent, structured JSON for your data pipelines!