tools.py

from langchain_core.tools import tool
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_community.document_loaders import PyPDFLoader
from langchain_openai import OpenAIEmbeddings
from langchain_community.tools import WikipediaQueryRun, ArxivQueryRun
from langchain_community.utilities import WikipediaAPIWrapper, ArxivAPIWrapper
from langchain_community.tools.tavily_search import TavilySearchResults
from dotenv import load_dotenv
import os

load_dotenv()

# ==============================
# CONFIG
# ==============================
VECTORSTORE_DIR = "data/vectorstore"
os.makedirs(VECTORSTORE_DIR, exist_ok=True)


# ==============================
# VECTOR STORE CREATION
# ==============================
def build_vectorstore(file_path: str):
    loader = PyPDFLoader(file_path)
    documents = loader.load()

    splitter = RecursiveCharacterTextSplitter(
        chunk_size=500,
        chunk_overlap=100
    )

    chunks = splitter.split_documents(documents)

    embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
    vectorstore = FAISS.from_documents(chunks, embeddings)

    vectorstore.save_local(VECTORSTORE_DIR)
    return vectorstore


def update_retriever(file_path: str):
    """Rebuild vectorstore when a new document is uploaded."""
    build_vectorstore(file_path)


# ==============================
# RAG TOOL (HF SAFE)
# ==============================
def create_rag_tool():

    @tool
    def rag_search(query: str) -> str:
        """
        Retrieve relevant information from uploaded documents.
        """

        if not os.path.exists(os.path.join(VECTORSTORE_DIR, "index.faiss")):
            return "No document has been uploaded yet."

        embeddings = OpenAIEmbeddings(model="text-embedding-3-small")

        vectorstore = FAISS.load_local(
            VECTORSTORE_DIR,
            embeddings,
            allow_dangerous_deserialization=True
        )

        retriever = vectorstore.as_retriever(search_kwargs={"k": 4})
        docs = retriever.invoke(query)

        if not docs:
            return "No relevant information found in the document."

        return "\n\n".join(d.page_content for d in docs)

    return rag_search


# ==============================
# EXTRA TOOLS
# ==============================

@tool
def wikipedia_search(query: str) -> dict:
    """Search Wikipedia."""
    try:
        return {"results": WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper()).run(query)}
    except Exception as e:
        return {"error": str(e)}


@tool
def arxiv_search(query: str) -> dict:
    """Search academic papers on arXiv."""
    try:
        return {"results": ArxivQueryRun(api_wrapper=ArxivAPIWrapper()).run(query)}
    except Exception as e:
        return {"error": str(e)}


@tool
def tavily_search(query: str) -> dict:
    """Search the web using Tavily."""
    try:
        return {"results": TavilySearchResults(max_results=5).run(query)}
    except Exception as e:
        return {"error": str(e)}