import gradio as gr
import pickle
import numpy as np

# Load trained model and encoders
model = pickle.load(open("model.pkl", "rb"))
encoders = pickle.load(open("encoders.pkl", "rb"))

def recommend_stack(project_type, team_size, perf_need, experience):
    pt = encoders["project_type"].transform([project_type])[0]
    pn = encoders["perf_need"].transform([perf_need])[0]
    ex = encoders["experience"].transform([experience])[0]
    input_data = np.array([[pt, team_size, pn, ex]])
    pred_encoded = model.predict(input_data)[0]
    return f"🔧 Recommended Tech Stack: {encoders['stack'].inverse_transform([pred_encoded])[0]}"

demo = gr.Interface(
    fn=recommend_stack,
    inputs=[
        gr.Radio(["Web App", "API", "ML App", "Real-time App"], label="Project Type"),
        gr.Slider(1, 10, step=1, label="Team Size"),
        gr.Radio(["Low", "Medium", "High"], label="Performance Need"),
        gr.Radio(["Beginner", "Intermediate", "Expert"], label="Experience Level")
    ],
    outputs="text",
    title="Tech Stack Advisor",
    description="Get a recommended tech stack based on your project and team!"
)

demo.launch(server_name="0.0.0.0", server_port=5550)