import gradio as gr def chat(message, history): thinking = """ Analyzing query... MoE routing (2/8 experts) Generating response """ response = f"**MiniMind Max2**: {message}\n\nProcessed with MoE (25% active params)." return response, thinking def calculate(expr): try: result = eval(expr, {"__builtins__": {}}, {}) return f"Result: {result}" except: return "Error: Invalid expression" with gr.Blocks(title="MiniMind Max2") as demo: gr.Markdown("# 🧠 MiniMind Max2 API\n*Efficient Edge AI with MoE Architecture*") with gr.Tab("💬 Chat"): chatbot = gr.Chatbot(type="messages", height=300) msg = gr.Textbox(label="Message", placeholder="Ask anything...") thinking_box = gr.Textbox(label="Thinking Trace", lines=5) def respond(message, chat_history): response, thinking = chat(message, chat_history) chat_history.append({"role": "user", "content": message}) chat_history.append({"role": "assistant", "content": response}) return chat_history, "", thinking msg.submit(respond, [msg, chatbot], [chatbot, msg, thinking_box]) gr.Button("Send", variant="primary").click(respond, [msg, chatbot], [chatbot, msg, thinking_box]) with gr.Tab("🔧 Tools"): expr = gr.Textbox(label="Expression", value="2 + 2 * 3") result = gr.Textbox(label="Result") gr.Button("Calculate", variant="primary").click(calculate, expr, result) with gr.Tab("ℹ️ Info"): gr.Markdown(""" ## MiniMind Max2 Architecture - **MoE**: 8 experts, top-2 routing (25% activation) - **GQA**: 16 Q-heads, 4 KV-heads - **Capabilities**: Reasoning, Vision, Coding, Tools ## Docker ```bash docker pull sultanafariabd/minimind-max2 docker run -p 8000:8000 sultanafariabd/minimind-max2 ``` """) gr.Markdown("---\n[Model](https://huggingface.co/fariasultana/MiniMind) | [Collection](https://huggingface.co/collections/fariasultana/minimind-max2-edge-ai-models-69321e758f98df18d4f4ec05)") demo.launch()