# -*- coding: utf-8 -*-
"""TinyLlama 1.1B LLM RAG Research Chatbot.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1gKNj3wQw1pUbUXLJ4TcQCW16ezvL8pPo
"""

!pip install pypdf
!pip install python-dotenv
!pip install -q transformers
!CMAKE_ARGS="-DLLAMA_CUBLAS=on" FORCE_CMAKE=1 pip install  llama-cpp-python --no-cache-dir
!pip install -q llama-index
!pip install -q transformers einops accelerate langchain bitsandbytes
!pip install sentence_transformers
!pip install llama-index-llms-huggingface
!pip install -q gradio
!pip install einops
!pip install accelerate

import logging
import sys

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

from llama_index.core import SimpleDirectoryReader, VectorStoreIndex
from llama_index.llms.huggingface import HuggingFaceLLM
from llama_index.core import Settings

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext

documents = SimpleDirectoryReader("/content/Data/").load_data()

len(documents)

documents[10]

from llama_index.core import PromptTemplate

system_prompt = "You are a Q&A assistant. Your goal is to answer questions as accurately as possible based on the instructions and context provided."



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

from llama_index.llms.huggingface import HuggingFaceLLM

import torch
llm = HuggingFaceLLM(
    context_window=2048,
    max_new_tokens=256,
    generate_kwargs={"temperature": 0.0, "do_sample": False},
    system_prompt=system_prompt,
    query_wrapper_prompt=query_wrapper_prompt,
    tokenizer_name="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
    model_name="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
    device_map="cuda",
    # uncomment this if using CUDA to reduce memory usage
    model_kwargs={"torch_dtype": torch.bfloat16},
)

from langchain.embeddings import HuggingFaceEmbeddings
from llama_index.embeddings.langchain import LangchainEmbedding

lc_embed_model = HuggingFaceEmbeddings(
    model_name="BAAI/bge-small-en-v1.5"
)
embed_model = LangchainEmbedding(lc_embed_model)

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

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

query_engine = index.as_query_engine()

def predict(input, history):
  response = query_engine.query(input)
  return str(response)

import gradio as gr
gr.ChatInterface(predict).launch(share=True)