AgentWebSearch Package

Description

...

Quickstart

Create a AgentWebSearch Object

from agentwebsearch import AgentWebSearch
from agentwebsearch.websearch.request import WebSearchRequest
from agentwebsearch.search.client import DefaultSearchClient
from agentwebsearch.webscraper import DefaultWebScraper
from agentwebsearch.indexsearch import HNSWInMemoryIndexDB
from agentwebsearch.llm import OpenAIChatModel
from agentwebsearch.embedding import OpenAIEmbeddingModel


embedding_model = OpenAIEmbeddingModel(
    model="text-embedding-3-large",
    api_key="YOUR_OPENAI_API_KEY"
)

llm = OpenAIChatModel(
    model="gpt-4o",
    api_key="YOUR_OPENAI_API_KEY",
    temperature=0.7
)

index_db = HNSWInMemoryIndexDB(embedding_model=embedding_model)
search_client = DefaultSearchClient()
scraper = DefaultWebScraper()

websearch = AgentWebSearch(
    search_client=search_client,
    index_db=index_db,
    scraper=scraper,
    llm=llm,
    embedding_model=embedding_model
)

Execute AgentWebSearch

from agentwebsearch.websearch.request import RequestQuery, RequestQueryMessage

req = WebSearchRequest(
    query=RequestQuery(
        messages=[
            RequestQueryMessage(
                role="user",
                content="Wann wurde der Bundeskanzler 2025 gewählt?"
            )
        ]
    )
)

result = websearch.execute(req)
# or
for result in websearch.execute(req, stream=True):
    print(result)
    yield result

Deploy as MCP-Server

# server.py
from agentwebsearch.mcp import WebSearchFastMCP

mcp = WebSearchFastMCP("Demo 🚀")


@mcp.tool
def other_tool():
    return "Other tool"


if __name__ == "__main__":
    # available tools:
    #   - websearch
    #   - other_tool

    mcp.run(
        transport="streamable-http",
        host="127.0.0.1",
        port=8000,
        path="/mcp"
    )

# python run server.py