import gradio as gr
from huggingface_hub import InferenceClient
from transformers import AutoModelForCausalLM, AutoTokenizer

# Initialize the Zephyr-7B client
zephyr_client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

# Load your fine-tuned GPT-2 model from Hugging Face
MODEL_NAME = "hackergeek98/therapist01"  # Replace with your model name
gpt2_tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
gpt2_model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)

# Initialize conversation history for GPT-2
conversation_history = ""

# Function to generate responses using Zephyr-7B
def respond_with_zephyr(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    for message in zephyr_client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content

        response += token
        yield response

# Function to generate responses using GPT-2
def respond_with_gpt2(user_input):
    global conversation_history

    # Update conversation history with user input
    conversation_history += f"User: {user_input}\n"
    
    # Tokenize the conversation history
    inputs = gpt2_tokenizer(conversation_history, return_tensors="pt", truncation=True, max_length=1024)

    # Generate a response from the model
    outputs = gpt2_model.generate(inputs['input_ids'], max_length=1024, num_return_sequences=1, no_repeat_ngram_size=2)

    # Decode the model's output
    response = gpt2_tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Update conversation history with the model's response
    conversation_history += f"Therapist: {response}\n"

    # Return the therapist's response
    return response

# Function to handle the model selection and response generation
def respond(message, history, model_choice, system_message, max_tokens, temperature, top_p):
    if model_choice == "Zephyr-7B":
        return respond_with_zephyr(message, history, system_message, max_tokens, temperature, top_p)
    elif model_choice == "GPT-2 Therapist":
        return respond_with_gpt2(message)
    else:
        return "Invalid model selection."

# Create Gradio interface
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Dropdown(choices=["Zephyr-7B", "GPT-2 Therapist"], label="Model", value="Zephyr-7B"),
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
    ],
    title="Multi-Model Chat Interface",
    description="Choose between Zephyr-7B and a fine-tuned GPT-2 model to chat with."
)

# Launch the app
if __name__ == "__main__":
    demo.launch()