app.py

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

# Load model and tokenizer
model_name = "microsoft/DialoGPT-medium"  # You can change this to another HF model if needed
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Create text-generation pipeline
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_length=1000,
    do_sample=True,
    truncation=True,  # Explicit truncation to avoid HF warnings
    pad_token_id=tokenizer.eos_token_id  # Prevents warning for open-end generation
)

# Wrap with LangChain LLM wrapper
llm = HuggingFacePipeline(pipeline=pipe)

# Prompt Template
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
)

# Conversation memory (stores past messages)
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

# LangChain LLM Chain
llm_chain = LLMChain(
    llm=llm,
    prompt=prompt,
    memory=memory,
    verbose=True
)

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

# Gradio UI
demo = gr.ChatInterface(
    fn=get_text_response,
    examples=["How are you doing?", "What are your interests?", "Which places do you like to visit?"],
    title="AI Chatbot",
    description="A simple chatbot using LangChain + HuggingFace + Gradio",
    theme="default",
    type="chat"  # Uses newer format to avoid Gradio tuple warnings
)

# Launch
if __name__ == "__main__":
    demo.queue().launch(share=True)