chatbot.py

import os
from langchain_community.vectorstores import FAISS
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain.chains import RetrievalQA

class WatershedChatbot:
    def __init__(self):
        os.environ['HF_HOME'] = '/tmp/hf_cache'  # Set a writeable cache dir

        # Use valid HuggingFace model
        self.embedding = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")

        self.vectorstore = FAISS.load_local(
            "dpr_vector_store_hf",
            embeddings=self.embedding,
            allow_dangerous_deserialization=True
        )

        self.all_districts = sorted({doc.metadata.get("district", "") for doc in self.vectorstore.docstore._dict.values()})

        self.llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0.3)

    def get_districts(self):
        return self.all_districts

    def answer_query(self, query: str, district: str = None):
        retriever = self.vectorstore.as_retriever(
            search_kwargs={"filter": {"district": district}} if district and district != "All" else {}
        )
        qa_chain = RetrievalQA.from_chain_type(llm=self.llm, retriever=retriever, return_source_documents=True)
        result = qa_chain(query)
        return {
            "answer": result["result"],
            "sources": [
                {
                    "source_file": doc.metadata.get("source_file", "Unknown"),
                    "chunk_index": doc.metadata.get("chunk_index", "N/A")
                }
                for doc in result["source_documents"]
            ]
        }