app.py

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext
from llama_index.llms import HuggingFaceLLM
from langchain.document_loaders import PyPDFLoader


**Mount Google Drive to access data (you may need to authenticate)**

import pandas as pd
from datasets import load_dataset, concatenate_datasets

dataset_names = [
    "medalpaca/medical_meadow_mediqa",
    "medalpaca/medical_meadow_medical_flashcards",
    "medalpaca/medical_meadow_wikidoc_patient_information",
    "medalpaca/medical_meadow_wikidoc",
    "medalpaca/medical_meadow_pubmed_casual",
    "medalpaca/medical_meadow_medqa",
    "medalpaca/medical_meadow_health_advice",
    "medalpaca/medical_meadow_cord19"

]

datasets = [load_dataset(name, split = "train") for name in dataset_names]
combined_dataset = concatenate_datasets(datasets)

#from google.colab import drive
#drive.mount('/content/drive')

**Reading the data from the saved path in google drive**

#documents = SimpleDirectoryReader("/content/drive/MyDrive/Data").load_data()

from langchain.text_splitter import CharacterTextSplitter
from langchain import OpenAI
from langchain.document_loaders import PyPDFLoader

**Define a system prompt for the Q&A assistant**


from llama_index.prompts.prompts import SimpleInputPrompt


system_prompt = "You are a medical AI chatbot. Your goal is to answer questions as accurately as possible based on the instructions and context provided.Use only information from the previous context information. Do not invent stuff or give false information"



# This will wrap the default prompts that are internal to llama-index
query_wrapper_prompt = SimpleInputPrompt("<|USER|>{query_str}<|ASSISTANT|>")

**Log in to Hugging Face**

#!huggingface-cli login

**Configure the HuggingFaceLLM (Language Model)**

import torch

llm = HuggingFaceLLM(
    context_window=4096,
    max_new_tokens=256,
    generate_kwargs={"temperature": 0.5, "do_sample": False},
    system_prompt=system_prompt,
    query_wrapper_prompt=query_wrapper_prompt,
    tokenizer_name="NousResearch/Llama-2-7b-chat-hf",
    model_name="NousResearch/Llama-2-7b-chat-hf",
    device_map="auto",
    # uncomment this if using CUDA to reduce memory usage
    model_kwargs={"torch_dtype": torch.float16 , "load_in_8bit":True})

**Configure embeddings using Hugging Face model**

from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from llama_index import LangchainEmbedding, ServiceContext

embed_model = LangchainEmbedding(
  HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
)


**Configure the service context**

service_context = ServiceContext.from_defaults(
    chunk_size=1024,
    llm=llm,
    embed_model=embed_model
)

**Create a vector store index from the loaded documents**

index = VectorStoreIndex.from_documents(combined_dataset, service_context=service_context)

**Create a query engine for the index**

query_engine = index.as_query_engine()
response = query_engine.query("What is gross profit?")

print(response)

**To interact with the chatbot**

import gradio as gr

# Define your chatbot function
def chatbot_interface(query):
    response = query_engine.query(query)
    return response

# Create a Gradio interface
iface = gr.Interface(
    fn=chatbot_interface,
    inputs=gr.Textbox(placeholder="Enter your question here..."),
    outputs=gr.Textbox(),
    live=False,
    title="Insurance Chatbot Demo",
    description="Ask questions, and the chatbot will provide answers based on the provided context.",
)

# Launch the Gradio interface
iface.launch(debug=True)