app.py

import os
import gradio as gr
from huggingface_hub import InferenceClient
import time

client = InferenceClient("lambdaindie/lambda-1v-1B", token = os.getenv("HF_TOKEN"))

css = """
@import url('https://fonts.googleapis.com/css2?family=JetBrains+Mono&display=swap');

* {
    font-family: 'JetBrains Mono', monospace !important;
}

html, body, .gradio-container {
    font-family: 'JetBrains Mono', monospace !important;
    background-color: #111 !important;
    color: #e0e0e0 !important;
}

textarea, input, button, select{
    background-color: transparent !important;
    color: #e0e0e0 !important;
    border: 1px solid #444 !important;
}


.markdown-think {
    background-color: #1e1e1e;
    border-left: 4px solid #555;
    padding: 10px;
    margin-bottom: 8px;
    font-style: italic;
    white-space: pre-wrap;
    animation: pulse 1.5s infinite ease-in-out;
}

@keyframes pulse {
    0% { opacity: 0.6; }
    50% { opacity: 1.0; }
    100% { opacity: 0.6; }
}
"""

theme = gr.themes.Base(
    primary_hue="gray",
    font=[
        gr.themes.GoogleFont("JetBrains Mono"),
        "monospace"
    ]
).set(
    body_background_fill="#111",
    body_text_color="#e0e0e0",
    button_primary_background_fill="#333",
    button_primary_text_color="#e0e0e0",
    input_background_fill="#222",
    input_border_color="#444",
    block_title_text_color="#fff"
)

def respond(message, history, system_message, max_tokens, temperature, top_p):
    messages = [{"role": "system", "content": system_message}] if system_message else []

    for user, assistant in history:
        if user:
            messages.append({"role": "user", "content": user})
        if assistant:
            messages.append({"role": "assistant", "content": assistant})

    thinking_prompt = messages + [{
        "role": "user",
        "content": f"{message}\n\nThink a bit step-by-step before answering."
    }]

    reasoning = ""
    yield '<div class="markdown-think">Thinking...</div>'

    start = time.time()

    for chunk in client.chat_completion(
        thinking_prompt,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = chunk.choices[0].delta.content or ""
        reasoning += token
        styled_thought = f'<div class="markdown-think">{reasoning.strip()}</div>'
        yield styled_thought

    elapsed = time.time() - start

    yield f"""
    <div style="margin-top:12px;padding:8px 12px;background-color:#222;border-left:4px solid #888;
            font-family:'JetBrains Mono', monospace;color:#ccc;font-size:14px;">
        Pensou por {elapsed:.1f} segundos
    </div>
    """

    time.sleep(2)

    final_prompt = messages + [
        {"role": "user", "content": message},
        {"role": "assistant", "content": reasoning.strip()},
        {"role": "user", "content": "Now answer based on your reasoning above."}
    ]

    final_answer = ""
    for chunk in client.chat_completion(
        final_prompt,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = chunk.choices[0].delta.content or ""
        final_answer += token
        yield final_answer.strip()

demo = gr.ChatInterface(
    fn=respond,
    title="λambdAI",
    theme=theme,
    css=css,
    additional_inputs=[
        gr.Textbox(value="", label="System Message"),
        gr.Slider(64, 2048, value=512, step=1, label="Max Tokens"),
        gr.Slider(0.1, 2.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p")
    ]
)

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