app/app.py

import os
from dotenv import load_dotenv

from llama_index.core import StorageContext, load_index_from_storage, VectorStoreIndex, SimpleDirectoryReader, ChatPromptTemplate
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.core import Settings


# Load environment variables
load_dotenv()


# Configure the Llama index settings
Settings.llm = HuggingFaceInferenceAPI(
    model_name="meta-llama/Meta-Llama-3-8B-Instruct",
    tokenizer_name="meta-llama/Meta-Llama-3-8B-Instruct",
    context_window=3900,
    token=os.getenv("HF_TOKEN"),
    max_new_tokens=1000,
    generate_kwargs={"temperature": 0.5},
)
Settings.embed_model = HuggingFaceEmbedding(
    model_name="BAAI/bge-small-en-v1.5"
)

# Define the directory for persistent storage and data
PERSIST_DIR = "./db"
DATA_DIR = "data"

# Ensure data directory exists
os.makedirs(DATA_DIR, exist_ok=True)
os.makedirs(PERSIST_DIR, exist_ok=True)

def data_ingestion():
    documents = SimpleDirectoryReader(DATA_DIR).load_data()
    storage_context = StorageContext.from_defaults()
    index = VectorStoreIndex.from_documents(documents)
    index.storage_context.persist(persist_dir=PERSIST_DIR)

def handle_query(query):
    storage_context = StorageContext.from_defaults(persist_dir=PERSIST_DIR)
    index = load_index_from_storage(storage_context)
    chat_text_qa_msgs = [
        (
            "user",
            """You are Q&A assistant named CHAT-DOC. Your main goal is to provide answers as accurately as possible, based on the instructions and context you have been given. If a question does not match the provided context or is outside the scope of the document, kindly advise the user to ask questions within the context of the document.
            Context:
            {context_str}
            Question:
            {query_str}
            """
        )
    ]
    text_qa_template = ChatPromptTemplate.from_messages(chat_text_qa_msgs)
    query_engine = index.as_query_engine(text_qa_template=text_qa_template)
    answer = query_engine.query(query)
    
    if hasattr(answer, 'response'):
        return answer.response
    elif isinstance(answer, dict) and 'response' in answer:
        return answer['response']
    else:
        return "Sorry, I couldn't find an answer."