app.py

import gradio as gr
from langchain_community.llms import HuggingFacePipeline
from langchain import LLMChain, PromptTemplate
from langchain.memory import ConversationBufferMemory
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

# Load a free model from Hugging Face
model_name = "microsoft/DialoGPT-medium"  # Or try "tiiuae/falcon-rw-1b" or "gpt2"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Create pipeline
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=1000, do_sample=True)

# Wrap with HuggingFacePipeline
llm = HuggingFacePipeline(pipeline=pipe)

template = """You are a helpful assistant to answer user queries.
{chat_history}
User: {user_message}
Chatbot:"""

prompt = PromptTemplate(
    input_variables=["chat_history", "user_message"], template=template
)

memory = ConversationBufferMemory(memory_key="chat_history")

llm_chain = LLMChain(
    llm=llm,
    prompt=prompt,
    verbose=True,
    memory=memory,
)

def get_text_response(user_message, history):
    response = llm_chain.predict(user_message=user_message)
    return response

demo = gr.ChatInterface(
    get_text_response,
    examples=["How are you doing?", "What are your interests?", "Which places do you like to visit?"]
)


if __name__ == "__main__":
    demo.queue().launch(share=True, debug=True)