app.py

import streamlit as st
import pandas as pd
import requests
import time
import textwrap
import os
import glob

API_BASE_URL = os.environ.get("API_BASE_URL", "http://localhost:8000")
CLIENT_ID = os.environ.get("CLIENT_ID", "DEMO_CLIENT_001")
APP_PASSWORD = os.environ.get("APP_PASSWORD", "s3poke") # Imposta questa nei Segreti di HF Space

st.set_page_config(page_title="Rilevatore di Ingredienti", layout="wide")

# CSS moderno e professionale con il font Inter
st.markdown("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');

html, body, [class*="css"]  {
    font-family: 'Inter', sans-serif;
}

.title-container {
    padding-bottom: 1rem;
    margin-bottom: 2rem;
    border-bottom: 1px solid #eaeaea;
}

.prediction-box {
    padding: 16px; 
    margin-bottom: 12px; 
    border-radius: 6px; 
    border-left: 4px solid #2563eb; 
    background-color: #f8fafc; 
    color: #1e293b;
    animation: fadeIn 0.5s ease-out;
    box-shadow: 0 1px 3px rgba(0,0,0,0.05);
}

.prediction-title {
    font-size: 1.1em;
    font-weight: 600;
    color: #0f172a;
}

.prediction-score {
    float: right; 
    color: #2563eb; 
    font-weight: 600; 
    font-size: 1.1em;
}

.prediction-desc {
    font-size: 0.9em; 
    color: #64748b;
    margin-top: 4px;
    display: block;
}

@keyframes fadeIn {
    0% { opacity: 0; transform: translateY(5px); }
    100% { opacity: 1; transform: translateY(0); }
}
</style>
""", unsafe_allow_html=True)

# Logica di autenticazione
def check_password():
    if not APP_PASSWORD:
        return True
    
    if "password_correct" not in st.session_state:
        st.session_state["password_correct"] = False

    if not st.session_state["password_correct"]:
        st.markdown('<div class="title-container"><h1>Accesso Rilevatore di Ingredienti</h1></div>', unsafe_allow_html=True)
        st.info("Questo spazio è privato. Inserisci la password di accesso.")
        password = st.text_input("Password", type="password")
        if st.button("Accedi"):
            if password == APP_PASSWORD:
                st.session_state["password_correct"] = True
                st.rerun()
            else:
                st.error("Password errata.")
        return False
    return True

if not check_password():
    st.stop()


def _do_configure_api():
    config_dir = os.path.join(os.path.dirname(__file__), "csv_config")
    csv_files = glob.glob(os.path.join(config_dir, "*.csv"))
    
    if not csv_files:
        return {"status": "error", "message": "Nessun CSV di configurazione trovato "}
    
    try:
        df = pd.read_csv(csv_files[0])
        required_cols = {"PRODUCT_ID", "DESCRIPTION", "FAMILY", "COMPONENT", "EXTRA DESCRIPTION"}
        if not required_cols.issubset(set(df.columns)):
            return {"status": "error", "message": "Colonne richieste mancanti nel CSV."}
        
        mask = (df["COMPONENT"].str.lower() == "ingredient") | (df["COMPONENT"].str.lower().str.startswith("proteine"))
        df_filtered = df[mask].copy()
        
        df_filtered["id"] = df_filtered["PRODUCT_ID"].astype(str)
        df_filtered["description"] = df_filtered["DESCRIPTION"].astype(str)
        
        df_filtered["FAMILY"] = df_filtered["FAMILY"].fillna("")
        df_filtered["EXTRA DESCRIPTION"] = df_filtered["EXTRA DESCRIPTION"].fillna("")
        df_filtered["extra_description"] = df_filtered["COMPONENT"].astype(str) + " - " + df_filtered["EXTRA DESCRIPTION"].astype(str)
        
        products = df_filtered[["id", "description", "extra_description"]].to_dict(orient="records")
        
        payload = {
            "client_id": CLIENT_ID,
            "context": "Rilevamento degli ingredienti all'interno delle Poke Bowl in Italia.",
            "products": products
        }
        
        resp = requests.post(f"{API_BASE_URL}/configure", json=payload)
        if resp.status_code == 200:
            return {"status": "success", "message": "API configurata con successo."}
        else:
            return {"status": "error", "message": f"Errore API {resp.status_code}: {resp.text}"}
    except Exception as e:
        return {"status": "error", "message": f"Impossibile configurare l'API: {e}"}

@st.cache_resource(show_spinner=False)
def configure_api():
    return _do_configure_api()

# Inizializza la configurazione silenziosamente
with st.spinner("Inizializzazione configurazione API..."):
    config_status = configure_api()

st.markdown('<div class="title-container"><h1>Rilevatore di Ingredienti</h1></div>', unsafe_allow_html=True)

if config_status["status"] == "error":
    st.error(f"Errore di Inizializzazione: {config_status['message']}")
    st.info("Controlla i file di configurazione e la connessione API.")

st.markdown("### Analisi dell'Immagine")
st.markdown("Carica l'immagine di una bowl o seleziona un'immagine di test per identificarne automaticamente gli ingredienti.")

# Cerca la cartella delle immagini di test (demo/test_images o ../tests)
possible_test_dirs = [
    os.path.join(os.path.dirname(__file__), "test_images"),
    os.path.join(os.path.dirname(__file__), "..", "tests")
]
test_images_dir = next((d for d in possible_test_dirs if os.path.exists(d)), None)

image_to_analyze_path = None
uploaded_img = None

if test_images_dir:
    test_images = []
    for ext in ('*.png', '*.jpg', '*.jpeg', '*.webp'):
        test_images.extend(glob.glob(os.path.join(test_images_dir, ext)))
    test_images = sorted(test_images)

    if test_images:
        selection_mode = st.radio("Scegli la sorgente dell'immagine:", ["Carica la tua", "Seleziona un'immagine di test"], horizontal=True)
        if selection_mode == "Seleziona un'immagine di test":
            img_names = [os.path.basename(p) for p in test_images]
            selected_img_name = st.selectbox("Seleziona un'immagine di test:", img_names)
            image_to_analyze_path = os.path.join(test_images_dir, selected_img_name)
        else:
            uploaded_img = st.file_uploader("Carica Immagine", type=["png", "jpg", "jpeg", "webp"], label_visibility="collapsed")
else:
    uploaded_img = st.file_uploader("Carica Immagine", type=["png", "jpg", "jpeg", "webp"], label_visibility="collapsed")

if uploaded_img is not None or image_to_analyze_path is not None:
    col1, col2 = st.columns([1, 1], gap="large")
    
    with col1:
        if uploaded_img is not None:
            st.image(uploaded_img, width='stretch')
            img_name = uploaded_img.name
            img_bytes = uploaded_img.getvalue()
            img_type = uploaded_img.type
        else:
            with open(image_to_analyze_path, "rb") as f:
                img_bytes = f.read()
            st.image(img_bytes, width='stretch')
            img_name = os.path.basename(image_to_analyze_path)
            img_type = "image/jpeg" if img_name.lower().endswith(('.jpg', '.jpeg')) else "image/png"
        
    with col2:
        if st.button("Identifica Ingredienti", type="primary", width='stretch'):
            with st.spinner("Analisi dell'immagine in corso..."):
                files = {"image": (img_name, img_bytes, img_type)}
                data = {"client_id": CLIENT_ID}
                try:
                    resp = requests.post(f"{API_BASE_URL}/predict", files=files, data=data)
                    
                    if resp.status_code == 409:
                        st.info("Configurazione assente in questa istanza. Sincronizzazione in corso... riprova la predizione in automatico.")
                        conf_resp = _do_configure_api()
                        if conf_resp.get("status") == "success":
                            # Ripopolare i file consumati da request.post e riprovare
                            files = {"image": (img_name, img_bytes, img_type)}
                            resp = requests.post(f"{API_BASE_URL}/predict", files=files, data=data)
                        else:
                            st.error(f"Errore durante l'auto-configurazione: {conf_resp.get('message')}")
                    
                    if resp.status_code == 200:
                        predictions = resp.json().get("predictions", [])
                        if not predictions:
                            st.info("Nessun ingrediente rilevato nell'immagine.")
                        else:
                            st.success("Analisi completata")
                            
                            result_container = st.empty()
                            displayed_markdown = "<h4>Componenti Rilevati</h4>"
                            
                            for pred in predictions:
                                score_pct = pred.get("score", 0) * 100
                                desc = pred.get("description") or "Nessun dettaglio aggiuntivo disponibile"
                                p_id = pred.get("product_id")
                                
                                item_html = textwrap.dedent(f"""
                                    <div class="prediction-box">
                                        <span class="prediction-title">{p_id}</span> 
                                        <span class="prediction-score">{score_pct:.1f}%</span><br>
                                        <span class="prediction-desc">{desc}</span>
                                    </div>
                                """)
                                displayed_markdown += item_html

                                result_container.markdown(displayed_markdown, unsafe_allow_html=True)
                                time.sleep(0.08)  # Elegante ritardo d'animazione
                                
                    else:
                        st.error(f"Errore API {resp.status_code}: {resp.text}")
                except Exception as e:
                    st.error(f"Impossibile connettersi all'API: {e}")