import gradio as gr
from transformers import pipeline

def text_tab(model_cache, unload_all_models):
    def load_text_model():
        unload_all_models()
        model_cache["text"] = pipeline("text-generation", model="gpt2", device=-1)
        return "Text model loaded!"

    def unload_text_model():
        if "text" in model_cache:
            del model_cache["text"]
        return "Text model unloaded!"

    def run_text(prompt):
        if "text" not in model_cache:
            return "Text model not loaded!"
        return model_cache["text"](prompt)[0]["generated_text"]

    with gr.Tab("Text Generation"):
        status = gr.Markdown("Model not loaded.")
        load_btn = gr.Button("Load Text Model")
        unload_btn = gr.Button("Unload Model")
        prompt = gr.Textbox(label="Prompt", value="Hello world")
        run_btn = gr.Button("Generate")
        output = gr.Textbox(label="Output")
        load_btn.click(load_text_model, None, status)
        unload_btn.click(unload_text_model, None, status)
        run_btn.click(run_text, prompt, output)