app.py

import gradio as gr
import spaces
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from langchain.llms.base import LLM
from langchain.memory import ConversationBufferMemory
from langchain.chains import LLMChain, ConversationChain
from langchain_community.llms.huggingface_pipeline import HuggingFacePipeline
from langchain.prompts import PromptTemplate, ChatPromptTemplate

@spaces.GPU
def initialize_model_and_tokenizer(model_name="KvrParaskevi/Llama-2-7b-Hotel-Booking-Model"):
    model = AutoModelForCausalLM.from_pretrained(model_name)
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    return model, tokenizer

@spaces.GPU
def load_pipeline():
    model, tokenizer = initialize_model_and_tokenizer()
    pipe = pipeline("text-generation", 
                    model= model, 
                    tokenizer = tokenizer,
                    max_new_tokens = 20,
                    top_k = 30, 
                    early_stopping=True,
                    num_beams = 2,
                    temperature = 0.1,
                    repetition_penalty = 1.03)

    llm = HuggingFacePipeline(pipeline = pipe)
    return llm

@spaces.GPU
def chat_interface(inputs):
    question = inputs["input"]
    chat_history = inputs["history"]
    # Assuming `chain` is your instance of ConversationalRetrievalChain
    result = chain.run({"input": question, "history": chat_history})
    return result["response"]

llm = load_pipeline()

template = """<<SYS>>
You are an AI having conversation with a human. Below is an instruction that describes a task. 
Write a response that appropriately completes the request.
Reply with the most helpful and logic answer. During the conversation you need to ask the user 
the following questions to complete the hotel booking task.
1) Where would you like to stay and when?
2) How many people are staying in the room?
3) Do you prefer any ammenities like breakfast included or gym?
4) What is your name, your email address and phone number? 
Make sure you receive a logical answer from the user from every question to complete the hotel 
booking process.
<</SYS>>
Previous conversation:
{history}
Human: {input}
AI:"""
prompt = PromptTemplate(template=template, input_variables=["history", "input"])

memory = ConversationBufferMemory(memory_key="history", llm = llm, prompt = prompt)
llm_chain = ConversationChain(prompt=prompt, llm=llm, memory = memory)

with gr.Blocks() as demo:
    gr.Markdown("Hotel Booking Assistant Chat 🤗")
    chatbot = gr.Chatbot(label="Chat history")
    message = gr.Textbox(label="Ask me a question!")
    clear = gr.Button("Clear")
    #llm_chain, llm = init_chain(model, tokenizer)

    # Convert chat history to list of tuples
    chat_history_tuples = []
    for message in chat_history:
        chat_history_tuples.append((message[0], message[1]))
    
    result = llm_chain({"input": query, "history": chat_history_tuples})
    
    iface = gr.Interface(
        fn=chat_interface, 
        inputs=[
            gr.inputs.Textbox(lines=1, label="Question"),
            gr.inputs.Textbox(lines=5, label="Chat History"),
        ], 
        outputs="text"
    )
iface.launch()