app.py

from huggingface_hub import InferenceClient
import gradio as gr

css = '''
.gradio-container{max-width: 1000px !important}
h1{text-align:center}
footer {
    visibility: hidden
}
'''
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")

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 format_prompt(message, history, system_prompt=None, mood=None):
    prompt = "<s>"
    if mood:
        mood_description = mood_prompts.get(mood, "")
        prompt += f"[SYS] {mood_description} [/SYS] "
    for user_prompt, bot_response in history:
        prompt += f"[INST] {user_prompt} [/INST]"
        prompt += f" {bot_response}</s> "
    if system_prompt:
        prompt += f"[SYS] {system_prompt} [/SYS]"
    prompt += f"[INST] {message} [/INST]"
    return prompt

def generate(
    prompt, history, system_prompt=None, mood=None, temperature=0.2, max_new_tokens=1024, top_p=0.95, repetition_penalty=1.0,
):
    temperature = float(temperature)
    if temperature < 1e-2:
        temperature = 1e-2
    top_p = float(top_p)

    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )

    formatted_prompt = format_prompt(prompt, history, system_prompt, mood)
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
        yield output
    return output

def gradio_interface():
    with gr.Blocks(css=css, theme="allenai/gradio-theme") as demo:
        with gr.Row():
            mood = gr.radio(choices=list(mood_prompts.keys()), label="Response Mood", value="Friendly")
            history = gr.State([])
            system_prompt = gr.Textbox(placeholder="System prompt (optional)", lines=1)
            prompt = gr.Textbox(placeholder="Enter your message", lines=2)
            generate_btn = gr.Button("Generate")
            output = gr.Chatbot()
        
        generate_btn.click(
            generate,
            inputs=[prompt, history, system_prompt, mood],
            outputs=[output]
        )
    demo.queue().launch(show_api=False)

gradio_interface()