app.py

import gradio as gr
from openai import OpenAI
import os

css = '''
.gradio-container{max-width: 1000px !important}
h1{text-align:center}
footer {
    visibility: hidden
}
'''

ACCESS_TOKEN = os.getenv("HF_TOKEN")

client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1/",
    api_key=ACCESS_TOKEN,
)

# Mood prompts dictionary
mood_prompts = {
    "Fun": "Respond in a light-hearted, playful manner.",
    "Serious": "Respond in a thoughtful, serious tone.",
    "Professional": "Respond in a formal, professional manner.",
    "Upset": "Respond in a slightly irritated, upset tone.",
    "Empathetic": "Respond in a warm and understanding tone.",
    "Optimistic": "Respond in a positive, hopeful manner.",
    "Sarcastic": "Respond with a hint of sarcasm.",
    "Motivational": "Respond with encouragement and motivation.",
    "Curious": "Respond with a sense of wonder and curiosity.",
    "Humorous": "Respond with a touch of humor.",
    "Cautious": "Respond with careful consideration and caution.",
    "Assertive": "Respond with confidence and assertiveness.",
    "Friendly": "Respond in a warm and friendly manner.",
    "Romantic": "Respond with affection and romance.",
    "Nostalgic": "Respond with a sense of longing for the past.",
    "Grateful": "Respond with gratitude and appreciation.",
    "Inspirational": "Respond with inspiration and positivity.",
    "Casual": "Respond in a relaxed and informal tone.",
    "Formal": "Respond with a high level of formality.",
    "Pessimistic": "Respond with a focus on potential negatives.",
    "Excited": "Respond with enthusiasm and excitement.",
    "Melancholic": "Respond with a sense of sadness or longing.",
    "Confident": "Respond with self-assurance and confidence.",
    "Suspicious": "Respond with caution and doubt.",
    "Reflective": "Respond with deep thought and introspection.",
    "Joyful": "Respond with happiness and joy.",
    "Mysterious": "Respond with an air of mystery and intrigue.",
    "Aggressive": "Respond with force and intensity.",
    "Calm": "Respond with a sense of peace and tranquility.",
    "Gloomy": "Respond with a sense of sadness or pessimism.",
    "Encouraging": "Respond with words of support and encouragement.",
    "Sympathetic": "Respond with understanding and compassion.",
    "Disappointed": "Respond with a tone of disappointment.",
    "Proud": "Respond with a sense of pride and accomplishment.",
    "Playful": "Respond in a fun and playful manner.",
    "Inquisitive": "Respond with curiosity and interest.",
    "Supportive": "Respond with reassurance and support.",
    "Reluctant": "Respond with hesitation and reluctance.",
    "Confused": "Respond with uncertainty and confusion.",
    "Energetic": "Respond with high energy and enthusiasm.",
    "Relaxed": "Respond with a calm and laid-back tone.",
    "Grumpy": "Respond with a touch of irritation.",
    "Hopeful": "Respond with a sense of hope and optimism.",
    "Indifferent": "Respond with a lack of strong emotion.",
    "Surprised": "Respond with shock and astonishment.",
    "Tense": "Respond with a sense of urgency or anxiety.",
    "Enthusiastic": "Respond with eagerness and excitement.",
    "Worried": "Respond with concern and apprehension."
}

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
    mood
):
    # Update system message with mood prompt
    mood_prompt = mood_prompts.get(mood, "")
    full_system_message = f"{system_message} {mood_prompt}".strip()

    messages = [{"role": "system", "content": full_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 client.chat.completions.create(
        model="meta-llama/Meta-Llama-3.1-8B-Instruct",
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
        messages=messages,
    ):
        token = message.choices[0].delta.content
        
        response += token
        yield response

demo = gr.Interface(
    fn=respond,
    inputs=[
        gr.Textbox(label="Input message"),
        gr.Textbox(value="", label="System message", visible=False),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens", visible=False),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature", visible=False),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P", visible=False),
        gr.Dropdown(choices=list(mood_prompts.keys()), label="Mood", value="Casual"),
    ],
    outputs=gr.Textbox(label="Response"),
    css=css,
    theme="allenai/gradio-theme",
)

if __name__ == "__main__":
    demo.launch()