src/streamlit_app.py

import os
import streamlit as st
import torch
from transformers import AutoTokenizer
from predict_utils import predict_injury

# 🔹 Load tokenizer from Hugging Face
@st.cache_resource
def load_tokenizer():
    tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")
    return tokenizer

# 🔹 Load model from file
@st.cache_resource
def load_model():
    model_path = "model/final_injury_model.pt"
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

    if not os.path.exists(model_path):
        st.error(f"Model file not found at: {model_path}")
        return None

    model = torch.load(model_path, map_location=device)
    model.eval()
    return model

# 💡 Main app
def main():
    st.set_page_config(page_title="NBA Injury Type & Duration Classifier", page_icon="🏀")
    st.title("NBA Injury Type & Duration Classifier 🏀")

    model = load_model()
    tokenizer = load_tokenizer()

    if model is None or tokenizer is None:
        st.stop()

    st.markdown("""
    Enter an injury description and player details to get predicted injury type and expected recovery duration.
    """)

    # 🔹 User Inputs
    text = st.text_area("Injury description", "player has a sprained ankle")
    prior_injuries = st.number_input("Number of Prior Injuries", min_value=0, value=1)
    injury_type_id = st.selectbox("General Injury Type", {"bone": 0, "muscle": 1, "joint": 2, "ligament": 3})
    position_id = st.selectbox("Player Position", {"PG": 1, "SG": 2, "SF": 3, "PF": 4, "C": 5})

    # 🔹 Prediction button
    if st.button("Predict"):
        label_map_type = ["bone", "muscle", "joint", "ligament"]
        label_map_duration = ["short", "medium", "long"]

        try:
            type_label, type_conf, duration_label, duration_conf = predict_injury(
                model=model,
                tokenizer=tokenizer,
                text=text,
                prior_injuries=prior_injuries,
                injury_type_id=injury_type_id,
                position_id=position_id,
                label_map_type=label_map_type,
                label_map_duration=label_map_duration
            )

            st.success(f"**Predicted Injury Type:** {type_label} ({type_conf:.1%} confidence)")
            st.success(f"**Expected Duration:** {duration_label} ({duration_conf:.1%} confidence)")

        except Exception as e:
            st.error(f"Prediction failed: {e}")

if __name__ == "__main__":
    main()