query_data.py

import argparse
# from dataclasses import dataclass
from langchain_chroma import Chroma
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_huggingface import HuggingFaceEndpoint

from langchain.prompts import ChatPromptTemplate

from langchain.chains import LLMChain
from langchain_core.prompts import PromptTemplate
import os

CHROMA_PATH = "chroma"

PROMPT_TEMPLATE = """
Answer the question based only on the following context:

{context}

---

Answer the question based on the above context: {question}
"""


def query_data(query_text):

    # Prepare the DB.
    embedding_function = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
    db = Chroma(persist_directory=CHROMA_PATH, embedding_function=embedding_function)

    # Search the DB.
    results = db.similarity_search_with_relevance_scores(query_text, k=3)
    if len(results) == 0 or results[0][1] < 0.2:
        print(f"Unable to find matching results.")
        return

    context_text = "\n\n---\n\n".join([doc.page_content for doc, _score in results])
    prompt_template = ChatPromptTemplate.from_template(PROMPT_TEMPLATE)

    repo_id = "HuggingFaceH4/zephyr-7b-beta"

    llm = HuggingFaceEndpoint(
        repo_id=repo_id,
        max_length = 512,
        temperature=0.5,
        huggingfacehub_api_token=os.environ['HF_TOKEN'],
    )
    llm_chain = prompt_template | llm

    response_text = llm_chain.invoke({"question": query_text, "context":context_text})
    
    sources = [doc.metadata.get("source", None) for doc, _score in results]
    formatted_response = f"{response_text}\nSources: {sources}"
    return formatted_response