tools.py

from typing import Type, Optional

from langchain_community.document_loaders import AsyncChromiumLoader
from langchain_community.document_transformers import BeautifulSoupTransformer
from langchain_community.tools.wikipedia.tool import WikipediaQueryInput
from langchain_community.utilities import WikipediaAPIWrapper
from langchain_core.callbacks import CallbackManagerForToolRun
from langchain_core.tools import BaseTool
from langchain_tavily import TavilySearch
from pydantic import BaseModel, Field


def _get_web_page(url: str) -> str:
    """Fetches the content of a web page and transforms (beautify) it into a string."""

    loader = AsyncChromiumLoader([url])
    docs = loader.load()
    bs_transformer = BeautifulSoupTransformer()
    docs = bs_transformer.transform_documents(docs)
    return '\n'.join(['=' * 30 + '\n' + doc.page_content for doc in docs])

class WikipediaQueryLoad(BaseTool):
    """Tool that searches the Wikipedia API."""

    name: str = "wikipedia"
    description: str = (
        "A wrapper around Wikipedia. "
        "Useful for when you need to answer general questions about "
        "people, places, companies, facts, historical events, or other subjects. "
        "Input should be a search query."
    )
    api_wrapper: WikipediaAPIWrapper = WikipediaAPIWrapper(top_k_results=1, doc_content_chars_max=20000)

    args_schema: Type[BaseModel] = WikipediaQueryInput

    def _run(
        self,
        query: str,
        run_manager: Optional[CallbackManagerForToolRun] = None,
    ) -> str:
        """Use the Wikipedia tool."""
        page_titles = self.api_wrapper.wiki_client.search(
            query[:300], results=1
        )
        summary = self.api_wrapper.run(query)

        # Wikipedia python package doesn't properly support some wiki syntax (i.e. tables), so
        # the full wiki page is read separately
        full_page = _get_web_page(f"https://en.wikipedia.org/wiki/{page_titles[0]}")

        res = [
            '**Wiki page url**:',
            page_titles[0],
            '**Wiki page summary:**',
            summary,
            '**Full page content:**',
            full_page
        ]
        return '\n'.join(res)


class WebScrapTool(BaseTool):
    name: str = "webscraper"

    description: str = "Loads full content of the web page."

    # Load HTML
    def _run(
        self,
        url: str,
        run_manager: Optional[CallbackManagerForToolRun] = None,
    ) -> str:
        return _get_web_page(url)


class CalculatorTool(BaseTool):
    """Tool that performs basic calculations."""

    name: str = "calculator"
    description: str = (
        "A calculator. "
        "Useful for when you need to perform basic calculations."
    )

    def _run(
        self,
        expression: str,
        run_manager: Optional[CallbackManagerForToolRun] = None,
    ) -> float:
        """Use the calculator tool."""
        return eval(expression)

def get_all_tools() -> list[BaseTool]:
    """Get all tools."""
    return [
        WikipediaQueryLoad(api_wrapper=WikipediaAPIWrapper(top_k_results=1, doc_content_chars_max=20000)),
        WebScrapTool(),
        TavilySearch(max_results=5, topic="general"),
        CalculatorTool()
    ]