app.py

import gradio as gr
from huggingface_hub import InferenceClient
import string
import numpy as np
from transformers import AutoTokenizer
import onnxruntime as ort
import os

# Initialize client and models
client = InferenceClient(api_key=os.environ.get('HF_TOKEN'))

# Constants for EOU calculation
PUNCS = string.punctuation.replace("'", "")
MAX_HISTORY = 4
MAX_HISTORY_TOKENS = 1024
EOU_THRESHOLD = 0.5

# Initialize tokenizer and ONNX session
HG_MODEL = "livekit/turn-detector"
ONNX_FILENAME = "model_quantized.onnx"
tokenizer = AutoTokenizer.from_pretrained(HG_MODEL)
onnx_session = ort.InferenceSession(ONNX_FILENAME, providers=["CPUExecutionProvider"])

# Helper functions for EOU
def softmax(logits):
    exp_logits = np.exp(logits - np.max(logits))
    return exp_logits / np.sum(exp_logits)

def normalize_text(text):
    def strip_puncs(text):
        return text.translate(str.maketrans("", "", PUNCS))
    return " ".join(strip_puncs(text).lower().split())

def format_chat_ctx(chat_ctx):
    new_chat_ctx = []
    for msg in chat_ctx:
        if msg["role"] in ("user", "assistant"):
            content = normalize_text(msg["content"])
            if content:
                msg["content"] = content
                new_chat_ctx.append(msg)
    convo_text = tokenizer.apply_chat_template(
        new_chat_ctx, add_generation_prompt=False, add_special_tokens=False, tokenize=False
    )
    ix = convo_text.rfind("<|im_end|>")
    return convo_text[:ix]

def calculate_eou(chat_ctx, session):
    formatted_text = format_chat_ctx(chat_ctx[-MAX_HISTORY:])
    inputs = tokenizer(
        formatted_text,
        return_tensors="np",
        truncation=True,
        max_length=MAX_HISTORY_TOKENS,
    )
    input_ids = np.array(inputs["input_ids"], dtype=np.int64)
    outputs = session.run(["logits"], {"input_ids": input_ids})
    logits = outputs[0][0, -1, :]
    probs = softmax(logits)
    eou_token_id = tokenizer.encode("<|im_end|>")[-1]
    return probs[eou_token_id]



messages = []


def chatbot(user_input):
    global messages
    
    # Exit condition
    if user_input.lower() == "exit":
        messages = []  # Reset conversation history
        return "Chat ended. Refresh the page to start again."

    
    # Add user message to conversation history
    messages.append({"role": "user", "content": user_input})

    # Calculate EOU to determine if user has finished typing
    eou_prob = calculate_eou(messages, onnx_session)
    if eou_prob < EOU_THRESHOLD:
        yield "[I'm waiting for you to complete the sentence...]"
        return

    # Stream the chatbot's response
    stream = client.chat.completions.create(
        model="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
        messages=messages,
        temperature=0.6,
        max_tokens=1024,
        top_p=0.95,
        stream=True
    )

    bot_response = ""
    for chunk in stream:
        bot_response += chunk.choices[0].delta.content
        yield bot_response

    # Add final bot response to conversation history
    messages.append({"role": "assistant", "content": bot_response})

# Create Gradio interface
with gr.Blocks(theme='darkdefault') as demo:
    gr.Markdown("""# Chat with DeepSeek-R1
    Type your message below to interact with the chatbot. Type "exit" to end the conversation.
    """)

    with gr.Row():
        with gr.Column():
            user_input = gr.Textbox(label="Your Message", placeholder="Type your message here...")
            submit_button = gr.Button("Send")
        with gr.Column():
            chat_output = gr.Textbox(label="Chatbot Response", interactive=False)

    # Define interactions
    submit_button.click(chatbot, inputs=[user_input], outputs=[chat_output])

# Launch the app
demo.launch()