app.py

import os, random
import torch
import pandas as pd
from flask import Flask, render_template, request, jsonify
from sentence_transformers import SentenceTransformer, util
import gradio as gr

BASE_DIR = os.path.abspath(os.path.dirname(__file__))

app = Flask(
    __name__,
    template_folder=os.path.join(BASE_DIR, "templates"),
    static_folder=os.path.join(BASE_DIR, "static")
)

CSV_DATA = "dataset_2026.csv"
EMB_FILE = "embeddings_questions.pt"
TOP_K_RECOMMANDATIONS = 5

print("🔄 Chargement du modèle...")
try:
    model = SentenceTransformer(
        "OrdalieTech/Solon-embeddings-mini-beta-1.1",
        device="cpu",
        trust_remote_code=True
    )
    print("✓ Modèle principal (Solon) chargé")
except Exception as e:
    print(f"⚠️ Échec du modèle principal (probablement dû à la version Hugging Face Hub): {e}")
    print("🔄 Chargement du modèle de secours (paraphrase-multilingual)...")
    model = SentenceTransformer(
        "paraphrase-multilingual-MiniLM-L12-v2",
        device="cpu"
    )
    print("✓ Modèle de secours chargé avec succès !")

df_cache = None

def load_data():
    global df_cache
    if df_cache is None:
        try:
            # We enforce reading the Latin-1 CSV using ';' as sep.
            df_cache = pd.read_csv(CSV_DATA, sep=";", encoding="latin-1")
            df_cache.columns = df_cache.columns.str.strip()
        except:
            # Fallback if somehow it's utf-8 with comma
            df_cache = pd.read_csv(CSV_DATA, sep=None, engine='python', encoding="utf-8")
            df_cache.columns = df_cache.columns.str.strip()
    return df_cache

def load_or_create_embeddings(df):
    if os.path.exists(EMB_FILE):
        emb = torch.load(EMB_FILE, map_location="cpu")
        if emb.shape[0] == len(df) and emb.shape[1] == model.get_sentence_embedding_dimension():
            return emb
        print("⚠️ Dimensions embeddings incorrectes, recréation...")

    print("🔨 Création embeddings...")
    questions = df["Question"].astype(str).tolist()
    emb = model.encode(
        questions,
        convert_to_tensor=True,
        normalize_embeddings=True
    )
    torch.save(emb, EMB_FILE)
    return emb

def enrich_message(base):
    return random.choice([
        f"Bonne question 🙂\n\n{base}",
        f"Voici la réponse détaillée :\n\n{base}",
        f"Voici ce que j'ai trouvé :\n\n{base}",
        base
    ])

def process_question(question):
    df = load_data()
    emb_base = load_or_create_embeddings(df)

    emb_q = model.encode(question, convert_to_tensor=True, normalize_embeddings=True)
    scores = util.pytorch_cos_sim(emb_q, emb_base)[0]

    best_idx = torch.argmax(scores).item()
    confidence = int(scores[best_idx].item() * 100)

    if confidence < 40:
        recs = df["Question"].sample(min(TOP_K_RECOMMANDATIONS, len(df))).tolist()
        return {
            "response": "Je ne suis pas sûr",
            "confidence": confidence,
            "matched": "—",
            "intent": "Inconnu",
            "recs": recs,
            "service": None,
            "lat": None,
            "lon": None
        }

    row = df.iloc[best_idx]
    
    service_val = str(row["Service"]).strip() if pd.notna(row["Service"]) else None
    if service_val and service_val.lower() in ['nan', 'none', 'null', '']: service_val = None

    link_val = str(row["ServiceLink"]).strip() if pd.notna(row["ServiceLink"]) else None
    if link_val and link_val.lower() in ['nan', 'none', 'null', '']: link_val = None

    try:
        raw_lat = str(row["Latitude"]).replace(',', '.')
        raw_lon = str(row["Longitude"]).replace(',', '.')
        lat_val = float(raw_lat)
        lon_val = float(raw_lon)
    except Exception:
        lat_val = None
        lon_val = None

    return {
        "response": enrich_message(row["Response"]),
        "confidence": confidence,
        "matched": row["Question"],
        "intent": row["Intent"],
        "recs": [],
        "service": service_val,
        "link": link_val,
        "lat": lat_val,
        "lon": lon_val
    }

@app.route("/")
def index():
    return render_template("index.html")

@app.route("/ask", methods=["POST"])
def ask():
    return jsonify(process_question(request.json.get("question", "")))

@app.route("/api/services", methods=["GET"])
def get_services():
    df = load_data()
    valid_df = df.dropna(subset=["Service", "Latitude", "Longitude"])
    
    services = []
    seen = set()

    for _, row in valid_df.iterrows():
        try:
            name = str(row["Service"]).strip()
            lat = float(str(row["Latitude"]).replace(',', '.'))
            lon = float(str(row["Longitude"]).replace(',', '.'))

            if name and name not in seen:
                services.append({"service": name, "lat": lat, "lon": lon})
                seen.add(name)
        except Exception:
            pass
            
    return jsonify({"status": "success", "services": services})

# ==============================
# HUGGING FACE SPACES (GRADIO)
# ==============================
def gradio_chat(message, history):
    return process_question(message)["response"]

iface = gr.ChatInterface(
    fn=gradio_chat,
    title="AskLaQ Assistant"
)

if __name__ == "__main__":
    print("🚀 Serveur Flask lancé sur http://127.0.0.1:7860")
    app.run(host="0.0.0.0", port=7860, debug=True, use_reloader=False)