from langchain_community.vectorstores import FAISS
from langchain.docstore.document import Document
from langchain_huggingface import HuggingFaceEmbeddings
import json

class RetrievalSystem:
    def __init__(self, document_path, embedder_model):
        self.embedder = HuggingFaceEmbeddings(model_name=embedder_model)
        self.vectorstore = self._build_vectorstore(document_path)

    def _build_vectorstore(self, document_path):
        with open(document_path, "r") as f:
            docs_data = json.load(f)
        documents = [Document(page_content=doc["content"], metadata={"category": doc["category"], "subkeyword": doc["subkeyword"]}) for doc in docs_data]
        return FAISS.from_documents(documents, embedding=self.embedder)

    def get_context(self, query, k=2):
        docs = self.vectorstore.similarity_search(query, k=k)
        return " ".join([doc.page_content for doc in docs])