langchain 0.4.0.dev0__py3-none-any.whl → 1.0.0a2__py3-none-any.whl

langchain/chains/natbot/crawler.py

@@ -1,472 +0,0 @@
-import time
-from sys import platform
-from typing import (
-    TYPE_CHECKING,
-    Any,
-    Optional,
-    TypedDict,
-    Union,
-)
-
-if TYPE_CHECKING:
-    from playwright.sync_api import Browser, CDPSession, Page
-
-black_listed_elements: set[str] = {
-    "html",
-    "head",
-    "title",
-    "meta",
-    "iframe",
-    "body",
-    "script",
-    "style",
-    "path",
-    "svg",
-    "br",
-    "::marker",
-}
-
-
-class ElementInViewPort(TypedDict):
-    """A typed dictionary containing information about elements in the viewport."""
-
-    node_index: str
-    backend_node_id: int
-    node_name: Optional[str]
-    node_value: Optional[str]
-    node_meta: list[str]
-    is_clickable: bool
-    origin_x: int
-    origin_y: int
-    center_x: int
-    center_y: int
-
-
-class Crawler:
-    """A crawler for web pages.
-
-    **Security Note**: This is an implementation of a crawler that uses a browser via
-        Playwright.
-
-        This crawler can be used to load arbitrary webpages INCLUDING content
-        from the local file system.
-
-        Control access to who can submit crawling requests and what network access
-        the crawler has.
-
-        Make sure to scope permissions to the minimal permissions necessary for
-        the application.
-
-        See https://python.langchain.com/docs/security for more information.
-    """
-
-    def __init__(self) -> None:
-        """Initialize the crawler."""
-        try:
-            from playwright.sync_api import sync_playwright
-        except ImportError as e:
-            msg = (
-                "Could not import playwright python package. "
-                "Please install it with `pip install playwright`."
-            )
-            raise ImportError(msg) from e
-        self.browser: Browser = (
-            sync_playwright().start().chromium.launch(headless=False)
-        )
-        self.page: Page = self.browser.new_page()
-        self.page.set_viewport_size({"width": 1280, "height": 1080})
-        self.page_element_buffer: dict[int, ElementInViewPort]
-        self.client: CDPSession
-
-    def go_to_page(self, url: str) -> None:
-        """Navigate to the given URL.
-
-        Args:
-            url: The URL to navigate to. If it does not contain a scheme, it will be
-                prefixed with "http://".
-        """
-        self.page.goto(url=url if "://" in url else "http://" + url)
-        self.client = self.page.context.new_cdp_session(self.page)
-        self.page_element_buffer = {}
-
-    def scroll(self, direction: str) -> None:
-        """Scroll the page in the given direction.
-
-        Args:
-            direction: The direction to scroll in, either "up" or "down".
-        """
-        if direction == "up":
-            self.page.evaluate(
-                "(document.scrollingElement || document.body).scrollTop = (document.scrollingElement || document.body).scrollTop - window.innerHeight;"  # noqa: E501
-            )
-        elif direction == "down":
-            self.page.evaluate(
-                "(document.scrollingElement || document.body).scrollTop = (document.scrollingElement || document.body).scrollTop + window.innerHeight;"  # noqa: E501
-            )
-
-    def click(self, id_: Union[str, int]) -> None:
-        """Click on an element with the given id.
-
-        Args:
-            id_: The id of the element to click on.
-        """
-        # Inject javascript into the page which removes the target= attribute from links
-        js = """
-		links = document.getElementsByTagName("a");
-		for (var i = 0; i < links.length; i++) {
-			links[i].removeAttribute("target");
-		}
-		"""
-        self.page.evaluate(js)
-
-        element = self.page_element_buffer.get(int(id_))
-        if element:
-            x: float = element["center_x"]
-            y: float = element["center_y"]
-
-            self.page.mouse.click(x, y)
-        else:
-            print("Could not find element")  # noqa: T201
-
-    def type(self, id_: Union[str, int], text: str) -> None:
-        """Type text into an element with the given id.
-
-        Args:
-            id_: The id of the element to type into.
-            text: The text to type into the element.
-        """
-        self.click(id_)
-        self.page.keyboard.type(text)
-
-    def enter(self) -> None:
-        """Press the Enter key."""
-        self.page.keyboard.press("Enter")
-
-    def crawl(self) -> list[str]:
-        """Crawl the current page.
-
-        Returns:
-            A list of the elements in the viewport.
-        """
-        page = self.page
-        page_element_buffer = self.page_element_buffer
-        start = time.time()
-
-        page_state_as_text = []
-
-        device_pixel_ratio: float = page.evaluate("window.devicePixelRatio")
-        if platform == "darwin" and device_pixel_ratio == 1:  # lies
-            device_pixel_ratio = 2
-
-        win_upper_bound: float = page.evaluate("window.pageYOffset")
-        win_left_bound: float = page.evaluate("window.pageXOffset")
-        win_width: float = page.evaluate("window.screen.width")
-        win_height: float = page.evaluate("window.screen.height")
-        win_right_bound: float = win_left_bound + win_width
-        win_lower_bound: float = win_upper_bound + win_height
-
-        # 	percentage_progress_start = (win_upper_bound / document_scroll_height) * 100
-        # 	percentage_progress_end = (
-        # 		(win_height + win_upper_bound) / document_scroll_height
-        # 	) * 100
-        percentage_progress_start = 1
-        percentage_progress_end = 2
-
-        page_state_as_text.append(
-            {
-                "x": 0,
-                "y": 0,
-                "text": f"[scrollbar {percentage_progress_start:0.2f}-{percentage_progress_end:0.2f}%]",  # noqa: E501
-            }
-        )
-
-        tree = self.client.send(
-            "DOMSnapshot.captureSnapshot",
-            {"computedStyles": [], "includeDOMRects": True, "includePaintOrder": True},
-        )
-        strings: dict[int, str] = tree["strings"]
-        document: dict[str, Any] = tree["documents"][0]
-        nodes: dict[str, Any] = document["nodes"]
-        backend_node_id: dict[int, int] = nodes["backendNodeId"]
-        attributes: dict[int, dict[int, Any]] = nodes["attributes"]
-        node_value: dict[int, int] = nodes["nodeValue"]
-        parent: dict[int, int] = nodes["parentIndex"]
-        node_names: dict[int, int] = nodes["nodeName"]
-        is_clickable: set[int] = set(nodes["isClickable"]["index"])
-
-        input_value: dict[str, Any] = nodes["inputValue"]
-        input_value_index: list[int] = input_value["index"]
-        input_value_values: list[int] = input_value["value"]
-
-        layout: dict[str, Any] = document["layout"]
-        layout_node_index: list[int] = layout["nodeIndex"]
-        bounds: dict[int, list[float]] = layout["bounds"]
-
-        cursor: int = 0
-
-        child_nodes: dict[str, list[dict[str, Any]]] = {}
-        elements_in_view_port: list[ElementInViewPort] = []
-
-        anchor_ancestry: dict[str, tuple[bool, Optional[int]]] = {"-1": (False, None)}
-        button_ancestry: dict[str, tuple[bool, Optional[int]]] = {"-1": (False, None)}
-
-        def convert_name(
-            node_name: Optional[str],
-            has_click_handler: Optional[bool],  # noqa: FBT001
-        ) -> str:
-            if node_name == "a":
-                return "link"
-            if node_name == "input":
-                return "input"
-            if node_name == "img":
-                return "img"
-            if (
-                node_name == "button" or has_click_handler
-            ):  # found pages that needed this quirk
-                return "button"
-            return "text"
-
-        def find_attributes(
-            attributes: dict[int, Any], keys: list[str]
-        ) -> dict[str, str]:
-            values = {}
-
-            for [key_index, value_index] in zip(*(iter(attributes),) * 2):
-                if value_index < 0:
-                    continue
-                key = strings[key_index]
-                value = strings[value_index]
-
-                if key in keys:
-                    values[key] = value
-                    keys.remove(key)
-
-                    if not keys:
-                        return values
-
-            return values
-
-        def add_to_hash_tree(
-            hash_tree: dict[str, tuple[bool, Optional[int]]],
-            tag: str,
-            node_id: int,
-            node_name: Optional[str],
-            parent_id: int,
-        ) -> tuple[bool, Optional[int]]:
-            parent_id_str = str(parent_id)
-            if parent_id_str not in hash_tree:
-                parent_name = strings[node_names[parent_id]].lower()
-                grand_parent_id = parent[parent_id]
-
-                add_to_hash_tree(
-                    hash_tree, tag, parent_id, parent_name, grand_parent_id
-                )
-
-            is_parent_desc_anchor, anchor_id = hash_tree[parent_id_str]
-
-            # even if the anchor is nested in another anchor, we set the "root" for all
-            # descendants to be ::Self
-            if node_name == tag:
-                value: tuple[bool, Optional[int]] = (True, node_id)
-            elif (
-                is_parent_desc_anchor
-            ):  # reuse the parent's anchor_id (which could be much higher in the tree)
-                value = (True, anchor_id)
-            else:
-                value = (
-                    False,
-                    None,
-                )
-                # not a descendant of an anchor, most likely it will become text, an
-                # interactive element or discarded
-
-            hash_tree[str(node_id)] = value
-
-            return value
-
-        for index, node_name_index in enumerate(node_names):
-            node_parent = parent[index]
-            node_name: Optional[str] = strings[node_name_index].lower()
-
-            is_ancestor_of_anchor, anchor_id = add_to_hash_tree(
-                anchor_ancestry, "a", index, node_name, node_parent
-            )
-
-            is_ancestor_of_button, button_id = add_to_hash_tree(
-                button_ancestry, "button", index, node_name, node_parent
-            )
-
-            try:
-                cursor = layout_node_index.index(index)
-                # TODO replace this with proper cursoring, ignoring the fact this is
-                # O(n^2) for the moment
-            except ValueError:
-                continue
-
-            if node_name in black_listed_elements:
-                continue
-
-            [x, y, width, height] = bounds[cursor]
-            x /= device_pixel_ratio
-            y /= device_pixel_ratio
-            width /= device_pixel_ratio
-            height /= device_pixel_ratio
-
-            elem_left_bound = x
-            elem_top_bound = y
-            elem_right_bound = x + width
-            elem_lower_bound = y + height
-
-            partially_is_in_viewport = (
-                elem_left_bound < win_right_bound
-                and elem_right_bound >= win_left_bound
-                and elem_top_bound < win_lower_bound
-                and elem_lower_bound >= win_upper_bound
-            )
-
-            if not partially_is_in_viewport:
-                continue
-
-            meta_data: list[str] = []
-
-            # inefficient to grab the same set of keys for kinds of objects, but it's
-            # fine for now
-            element_attributes = find_attributes(
-                attributes[index], ["type", "placeholder", "aria-label", "title", "alt"]
-            )
-
-            ancestor_exception = is_ancestor_of_anchor or is_ancestor_of_button
-            ancestor_node_key = (
-                None
-                if not ancestor_exception
-                else str(anchor_id)
-                if is_ancestor_of_anchor
-                else str(button_id)
-            )
-            ancestor_node = (
-                None
-                if not ancestor_exception
-                else child_nodes.setdefault(str(ancestor_node_key), [])
-            )
-
-            if node_name == "#text" and ancestor_exception and ancestor_node:
-                text = strings[node_value[index]]
-                if text in {"|", "•"}:
-                    continue
-                ancestor_node.append({"type": "type", "value": text})
-            else:
-                if (
-                    node_name == "input" and element_attributes.get("type") == "submit"
-                ) or node_name == "button":
-                    node_name = "button"
-                    element_attributes.pop(
-                        "type", None
-                    )  # prevent [button ... (button)..]
-
-                for key in element_attributes:
-                    if ancestor_exception and ancestor_node:
-                        ancestor_node.append(
-                            {
-                                "type": "attribute",
-                                "key": key,
-                                "value": element_attributes[key],
-                            }
-                        )
-                    else:
-                        meta_data.append(element_attributes[key])
-
-            element_node_value = None
-
-            if node_value[index] >= 0:
-                element_node_value = strings[node_value[index]]
-                if (
-                    element_node_value == "|"
-                    # commonly used as a separator, does not add much context - lets
-                    # save ourselves some token space
-                ):
-                    continue
-            elif (
-                node_name == "input"
-                and index in input_value_index
-                and element_node_value is None
-            ):
-                node_input_text_index = input_value_index.index(index)
-                text_index = input_value_values[node_input_text_index]
-                if node_input_text_index >= 0 and text_index >= 0:
-                    element_node_value = strings[text_index]
-
-            # remove redundant elements
-            if ancestor_exception and (node_name not in {"a", "button"}):
-                continue
-
-            elements_in_view_port.append(
-                {
-                    "node_index": str(index),
-                    "backend_node_id": backend_node_id[index],
-                    "node_name": node_name,
-                    "node_value": element_node_value,
-                    "node_meta": meta_data,
-                    "is_clickable": index in is_clickable,
-                    "origin_x": int(x),
-                    "origin_y": int(y),
-                    "center_x": int(x + (width / 2)),
-                    "center_y": int(y + (height / 2)),
-                }
-            )
-
-        # lets filter further to remove anything that does not hold any text nor has
-        # click handlers + merge text from leaf#text nodes with the parent
-        elements_of_interest = []
-        id_counter = 0
-
-        for element in elements_in_view_port:
-            node_index = element.get("node_index")
-            node_name = element.get("node_name")
-            element_node_value = element.get("node_value")
-            node_is_clickable = element.get("is_clickable")
-            node_meta_data: Optional[list[str]] = element.get("node_meta")
-
-            inner_text = f"{element_node_value} " if element_node_value else ""
-            meta = ""
-
-            if node_index in child_nodes:
-                for child in child_nodes[node_index]:
-                    entry_type = child.get("type")
-                    entry_value = child.get("value")
-
-                    if entry_type == "attribute" and node_meta_data:
-                        entry_key = child.get("key")
-                        node_meta_data.append(f'{entry_key}="{entry_value}"')
-                    else:
-                        inner_text += f"{entry_value} "
-
-            if node_meta_data:
-                meta_string = " ".join(node_meta_data)
-                meta = f" {meta_string}"
-
-            if inner_text != "":
-                inner_text = f"{inner_text.strip()}"
-
-            converted_node_name = convert_name(node_name, node_is_clickable)
-
-            # not very elegant, more like a placeholder
-            if (
-                (converted_node_name != "button" or meta == "")
-                and converted_node_name not in {"link", "input", "img", "textarea"}
-            ) and inner_text.strip() == "":
-                continue
-
-            page_element_buffer[id_counter] = element
-
-            if inner_text != "":
-                elements_of_interest.append(
-                    f"""<{converted_node_name} id={id_counter}{meta}>{inner_text}</{converted_node_name}>"""  # noqa: E501
-                )
-            else:
-                elements_of_interest.append(
-                    f"""<{converted_node_name} id={id_counter}{meta}/>"""
-                )
-            id_counter += 1
-
-        print(f"Parsing time: {time.time() - start:0.2f} seconds")  # noqa: T201
-        return elements_of_interest

langchain/chains/natbot/prompt.py

@@ -1,143 +0,0 @@
-from langchain_core.prompts.prompt import PromptTemplate
-
-_PROMPT_TEMPLATE = """
-You are an agents controlling a browser. You are given:
-
-	(1) an objective that you are trying to achieve
-	(2) the URL of your current web page
-	(3) a simplified text description of what's visible in the browser window (more on that below)
-
-You can issue these commands:
-	SCROLL UP - scroll up one page
-	SCROLL DOWN - scroll down one page
-	CLICK X - click on a given element. You can only click on links, buttons, and inputs!
-	TYPE X "TEXT" - type the specified text into the input with id X
-	TYPESUBMIT X "TEXT" - same as TYPE above, except then it presses ENTER to submit the form
-
-The format of the browser content is highly simplified; all formatting elements are stripped.
-Interactive elements such as links, inputs, buttons are represented like this:
-
-		<link id=1>text</link>
-		<button id=2>text</button>
-		<input id=3>text</input>
-
-Images are rendered as their alt text like this:
-
-		<img id=4 alt=""/>
-
-Based on your given objective, issue whatever command you believe will get you closest to achieving your goal.
-You always start on Google; you should submit a search query to Google that will take you to the best page for
-achieving your objective. And then interact with that page to achieve your objective.
-
-If you find yourself on Google and there are no search results displayed yet, you should probably issue a command
-like "TYPESUBMIT 7 "search query"" to get to a more useful page.
-
-Then, if you find yourself on a Google search results page, you might issue the command "CLICK 24" to click
-on the first link in the search results. (If your previous command was a TYPESUBMIT your next command should
-probably be a CLICK.)
-
-Don't try to interact with elements that you can't see.
-
-Here are some examples:
-
-EXAMPLE 1:
-==================================================
-CURRENT BROWSER CONTENT:
-------------------
-<link id=1>About</link>
-<link id=2>Store</link>
-<link id=3>Gmail</link>
-<link id=4>Images</link>
-<link id=5>(Google apps)</link>
-<link id=6>Sign in</link>
-<img id=7 alt="(Google)"/>
-<input id=8 alt="Search"></input>
-<button id=9>(Search by voice)</button>
-<button id=10>(Google Search)</button>
-<button id=11>(I'm Feeling Lucky)</button>
-<link id=12>Advertising</link>
-<link id=13>Business</link>
-<link id=14>How Search works</link>
-<link id=15>Carbon neutral since 2007</link>
-<link id=16>Privacy</link>
-<link id=17>Terms</link>
-<text id=18>Settings</text>
-------------------
-OBJECTIVE: Find a 2 bedroom house for sale in Anchorage AK for under $750k
-CURRENT URL: https://www.google.com/
-YOUR COMMAND:
-TYPESUBMIT 8 "anchorage redfin"
-==================================================
-
-EXAMPLE 2:
-==================================================
-CURRENT BROWSER CONTENT:
-------------------
-<link id=1>About</link>
-<link id=2>Store</link>
-<link id=3>Gmail</link>
-<link id=4>Images</link>
-<link id=5>(Google apps)</link>
-<link id=6>Sign in</link>
-<img id=7 alt="(Google)"/>
-<input id=8 alt="Search"></input>
-<button id=9>(Search by voice)</button>
-<button id=10>(Google Search)</button>
-<button id=11>(I'm Feeling Lucky)</button>
-<link id=12>Advertising</link>
-<link id=13>Business</link>
-<link id=14>How Search works</link>
-<link id=15>Carbon neutral since 2007</link>
-<link id=16>Privacy</link>
-<link id=17>Terms</link>
-<text id=18>Settings</text>
-------------------
-OBJECTIVE: Make a reservation for 4 at Dorsia at 8pm
-CURRENT URL: https://www.google.com/
-YOUR COMMAND:
-TYPESUBMIT 8 "dorsia nyc opentable"
-==================================================
-
-EXAMPLE 3:
-==================================================
-CURRENT BROWSER CONTENT:
-------------------
-<button id=1>For Businesses</button>
-<button id=2>Mobile</button>
-<button id=3>Help</button>
-<button id=4 alt="Language Picker">EN</button>
-<link id=5>OpenTable logo</link>
-<button id=6 alt ="search">Search</button>
-<text id=7>Find your table for any occasion</text>
-<button id=8>(Date selector)</button>
-<text id=9>Sep 28, 2022</text>
-<text id=10>7:00 PM</text>
-<text id=11>2 people</text>
-<input id=12 alt="Location, Restaurant, or Cuisine"></input>
-<button id=13>Let's go</button>
-<text id=14>It looks like you're in Peninsula. Not correct?</text>
-<button id=15>Get current location</button>
-<button id=16>Next</button>
-------------------
-OBJECTIVE: Make a reservation for 4 for dinner at Dorsia in New York City at 8pm
-CURRENT URL: https://www.opentable.com/
-YOUR COMMAND:
-TYPESUBMIT 12 "dorsia new york city"
-==================================================
-
-The current browser content, objective, and current URL follow. Reply with your next command to the browser.
-
-CURRENT BROWSER CONTENT:
-------------------
-{browser_content}
-------------------
-
-OBJECTIVE: {objective}
-CURRENT URL: {url}
-PREVIOUS COMMAND: {previous_command}
-YOUR COMMAND:
-"""  # noqa: E501
-PROMPT = PromptTemplate(
-    input_variables=["browser_content", "url", "previous_command", "objective"],
-    template=_PROMPT_TEMPLATE,
-)

langchain/chains/openai_functions/__init__.py

@@ -1,44 +0,0 @@
-from langchain_core.utils.function_calling import convert_to_openai_function
-
-from langchain.chains.openai_functions.base import (
-    create_openai_fn_chain,
-    create_structured_output_chain,
-)
-from langchain.chains.openai_functions.citation_fuzzy_match import (
-    create_citation_fuzzy_match_chain,
-    create_citation_fuzzy_match_runnable,
-)
-from langchain.chains.openai_functions.extraction import (
-    create_extraction_chain,
-    create_extraction_chain_pydantic,
-)
-from langchain.chains.openai_functions.qa_with_structure import (
-    create_qa_with_sources_chain,
-    create_qa_with_structure_chain,
-)
-from langchain.chains.openai_functions.tagging import (
-    create_tagging_chain,
-    create_tagging_chain_pydantic,
-)
-from langchain.chains.structured_output.base import (
-    create_openai_fn_runnable,
-    create_structured_output_runnable,
-    get_openai_output_parser,
-)
-
-__all__ = [
-    "convert_to_openai_function",
-    "create_citation_fuzzy_match_chain",
-    "create_citation_fuzzy_match_runnable",
-    "create_extraction_chain",
-    "create_extraction_chain_pydantic",
-    "create_openai_fn_chain",
-    "create_openai_fn_runnable",  # backwards compatibility
-    "create_qa_with_sources_chain",
-    "create_qa_with_structure_chain",
-    "create_structured_output_chain",
-    "create_structured_output_runnable",  # backwards compatibility
-    "create_tagging_chain",
-    "create_tagging_chain_pydantic",
-    "get_openai_output_parser",  # backwards compatibility
-]