src/Analytics/Data_exploration.py

"""
DATA_EXPLORATION.PY  Unite 2 : Analyses Exploratoires
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Fonctions publiques :
  enrich_ex(sheets)                -> 27 scores EX_* silencieux
  render_filter_bar(sheets)        -> filtre global, retourne sheets filtre
  render_exploration(sheets, sec)  -> visuels U2 (6+ par feuille)
"""

import re
import streamlit as st
import pandas as pd
import numpy as np
import plotly.graph_objects as go
import plotly.express as px
from datetime import datetime
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA

_C = {
    "bg":       "#0a0e12",
    "card":     "#0f141a",
    "border":   "rgba(80,100,120,0.25)",
    "accent":   "#58a6ff",
    "ex":       "#f39c12",
    "success":  "#2ecc71",
    "warning":  "#f39c12",
    "critical": "#c0392b",
    "neutral":  "#5a6a7a",
    "text":     "#a8b8c8",
    "subtext":  "#5a6a7a",
    "grid":     "rgba(80,100,120,0.12)",
}
_PAL  = ["#58a6ff","#2ecc71","#f39c12","#c0392b","#a78bfa","#38bdf8","#fb7185","#34d399"]
_FONT = "JetBrains Mono, Courier New, monospace"
_BASE = dict(
    paper_bgcolor=_C["card"], plot_bgcolor=_C["card"],
    font=dict(color=_C["text"], family=_FONT, size=11),
    title_font=dict(size=12, color=_C["text"], family=_FONT),
    margin=dict(t=44, b=36, l=48, r=24),
    legend=dict(bgcolor=_C["bg"], bordercolor=_C["border"], borderwidth=1,
                font=dict(size=10, family=_FONT)),
    xaxis=dict(gridcolor=_C["grid"], zerolinecolor=_C["grid"],
               tickfont=dict(family=_FONT, size=10)),
    yaxis=dict(gridcolor=_C["grid"], zerolinecolor=_C["grid"],
               tickfont=dict(family=_FONT, size=10)),
    colorway=_PAL,
)


# ── Helpers ───────────────────────────────────────────────────────────────────

def _L(**kw):
    m = {**_BASE}
    for ax in ("xaxis", "yaxis"):
        if ax in kw:
            m[ax] = {**m[ax], **kw.pop(ax)}
    m.update(kw)
    return m

def _ef(msg, h=220):
    fig = go.Figure()
    fig.add_annotation(text=msg, xref="paper", yref="paper", x=0.5, y=0.5,
                       showarrow=False, font=dict(size=12, color=_C["subtext"], family=_FONT))
    fig.update_layout(**_L(height=h))
    return fig

def _age_calc(series):
    today = datetime.today()
    def _a(v):
        try:
            d = pd.to_datetime(str(v), dayfirst=True, errors="coerce")
            return int((today - d).days / 365.25) if not pd.isna(d) else np.nan
        except Exception:
            return np.nan
    return series.apply(_a)

def _safe(df, col, default=0):
    return df[col].fillna(default) if col in df.columns else pd.Series(default, index=df.index)

def _ctx(text: str):
    """Bandeau de contexte orange, signature Exploration."""
    st.markdown(
        f'<div style="border-left:3px solid {_C["ex"]};padding:5px 12px;'
        f'margin-bottom:14px;font-family:{_FONT};font-size:0.62rem;'
        f'color:{_C["subtext"]};letter-spacing:0.8px;">{text}</div>',
        unsafe_allow_html=True,
    )

def _age_bin(series_age: pd.Series) -> pd.Series:
    """Convertit une serie d'ages en tranches textuelles. Pas de .cat.categories."""
    bins  = [0, 25, 35, 45, 55, 120]
    labs  = ["<25", "25-34", "35-44", "45-54", "55+"]
    # pd.cut retourne une Categorical; on la convertit en str pour eviter .cat
    result = pd.cut(series_age, bins=bins, labels=labs, right=False)
    return result.astype(str).replace("nan", np.nan)


# ══════════════════════════════════════════════════════════════════════════════
# ENRICHISSEMENT EX_*  (silencieux - 27 scores)
# ══════════════════════════════════════════════════════════════════════════════


def _pca_ex(df: pd.DataFrame, feat_cols: list, label_col: str,
            color_col: str, color_title: str, title: str,
            colorscale=None) -> go.Figure:
    """PCA 2D generique pour l'exploration — signature Exploration (orange)."""
    if colorscale is None:
        colorscale = [[0,"#2ecc71"],[0.5,"#f39c12"],[1,"#c0392b"]]
    avail = [c for c in feat_cols if c in df.columns and
             pd.api.types.is_numeric_dtype(df[c])]
    if len(df) < 3 or len(avail) < 2:
        return _ef(f"PCA indisponible ({len(df)} lignes, {len(avail)} features numeriques).", 340)
    X = df[avail].fillna(df[avail].median()).values
    try:
        X_sc  = StandardScaler().fit_transform(X)
        n_c   = min(2, X_sc.shape[0], X_sc.shape[1])
        pca   = PCA(n_components=n_c, random_state=42)
        comp  = pca.fit_transform(X_sc)
        var1  = pca.explained_variance_ratio_[0] * 100
        var2  = pca.explained_variance_ratio_[1] * 100 if n_c > 1 else 0.0
    except Exception as e:
        return _ef(f"PCA erreur : {e}", 340)
    pc2    = comp[:,1] if n_c > 1 else [0]*len(comp)
    labels = df[label_col].astype(str).str[:12] if label_col in df.columns else df.index.astype(str)
    color_vals = df[color_col].fillna(df[color_col].median()) if color_col in df.columns else pd.Series(0, index=df.index)
    fig = go.Figure(go.Scatter(
        x=comp[:,0], y=pc2, mode="markers+text",
        text=labels, textposition="top center",
        textfont=dict(size=8, family=_FONT, color=_C["subtext"]),
        marker=dict(size=12, color=color_vals, colorscale=colorscale,
                    showscale=True,
                    colorbar=dict(title=color_title, tickfont=dict(size=9, family=_FONT), len=0.65),
                    line=dict(width=1.5, color=_C["bg"])),
        hovertemplate=(
            "<b>%{text}</b><br>PC1 : %{x:.2f}<br>PC2 : %{y:.2f}<br>"
            + color_title + " : %{marker.color:.3f}<extra></extra>"
        ),
    ))
    fig.update_layout(**_L(
        title=f"{title}  ({var1:.0f}%+{var2:.0f}% var.)",
        height=340, showlegend=False,
        xaxis_title=f"PC1 ({var1:.1f}%)", yaxis_title=f"PC2 ({var2:.1f}%)",
    ))
    return fig


def enrich_ex(sheets: dict) -> dict:
    s   = {k: df.copy() for k, df in sheets.items() if isinstance(df, pd.DataFrame)}
    cli = s.get("Clients_KYC",           pd.DataFrame())
    gar = s.get("Garants_KYC",           pd.DataFrame())
    prt = s.get("Prets_Master",          pd.DataFrame())
    upd = s.get("Prets_Update",          pd.DataFrame())
    rem = s.get("Remboursements",        pd.DataFrame())
    adj = s.get("Ajustements_Echeances", pd.DataFrame())

    # ── Clients (5) ───────────────────────────────────────────────────────────
    if not cli.empty:
        # EX_Income_Volatility_Matrix
        if "Secteur_Activite" in cli.columns:
            rev   = _safe(cli, "Revenus_Mensuels").astype(float)
            autre = _safe(cli, "Autres_Revenus").astype(float)
            ratio = autre / rev.replace(0, np.nan)
            cli["EX_Income_Volatility_Matrix"] = (
                ratio.groupby(cli["Secteur_Activite"])
                     .transform(lambda x: x.std(ddof=0) if len(x) > 1 else 0.0)
            ).round(3)
        else:
            cli["EX_Income_Volatility_Matrix"] = 0.0

        # EX_AML_Risk_Correlation
        risky = {"ESPECES","INCONNU","TIERS","NON_DECLARE"}
        def _aml(row):
            aml = str(row.get("Verification_AML","")).upper()
            org = str(row.get("Origine_Fonds","")).upper()
            return 1 if aml != "VERIFIE" and any(r in org for r in risky) else 0
        cli["EX_AML_Risk_Correlation"] = cli.apply(_aml, axis=1)

        # EX_Contact_Reliability_Index
        def _contact(row):
            sc = 0
            ph = str(row.get("Telephone","")).strip()
            em = str(row.get("Email","")).strip()
            sc += 2 if re.match(r"^\+[\d\s\-]{9,15}$", ph) else (1 if re.match(r"[\d\+][\d\s\-]{7,}", ph) else 0)
            sc += 2 if re.match(r"^[^@\s]{2,}@[^@\s]{2,}\.[a-zA-Z]{2,6}$", em) else 0
            return sc
        cli["EX_Contact_Reliability_Index"] = cli.apply(_contact, axis=1)

        # EX_Neighborhood_Income_Benchmark
        if "Quartier" in cli.columns and "Revenus_Mensuels" in cli.columns:
            rev_q = cli.groupby("Quartier")["Revenus_Mensuels"].transform("mean")
            cli["EX_Neighborhood_Income_Benchmark"] = (
                _safe(cli,"Revenus_Mensuels") / rev_q.replace(0, np.nan)
            ).round(3)
        else:
            cli["EX_Neighborhood_Income_Benchmark"] = 1.0

        # EX_Professional_Seniority_Efficiency
        rn = (_safe(cli,"Revenus_Mensuels") - _safe(cli,"Charges_Estimees")).replace(0, np.nan)
        cli["EX_Professional_Seniority_Efficiency"] = (
            rn / _safe(cli,"Anciennete_Emploi",1).replace(0, np.nan)
        ).round(2)
        s["Clients_KYC"] = cli

    # ── Garants (5) ──────────────────────────────────────────────────────────
    if not gar.empty:
        gar["EX_Guarantor_Disposable_Income"] = (
            _safe(gar,"Revenus_Mensuels") - _safe(gar,"Charges_Estimees")
        ).round(0)
        cf = _safe(gar,"Checking_frequency",1).replace(0, np.nan)
        gar["EX_Wealth_to_Risk_Ratio"] = (_safe(gar,"Patrimoine_Declare") / cf).round(2)
        if "Ville" in gar.columns and "Secteur_Activite" in gar.columns:
            gar["EX_Guarantee_Diversity_Score"] = gar.groupby("Ville")["Secteur_Activite"].transform("nunique")
        else:
            gar["EX_Guarantee_Diversity_Score"] = 1
        if not cli.empty and "ID_Client" in cli.columns and "Moyen_Transfert" in cli.columns:
            cli_mt = set(cli["Moyen_Transfert"].str.upper().dropna().tolist())
            gar["EX_Bank_Transfer_Interoperability"] = gar.apply(
                lambda row: 1 if str(row.get("Moyen_Transfert","")).upper() in cli_mt else 0, axis=1)
        else:
            gar["EX_Bank_Transfer_Interoperability"] = 0
        rev_tot = (_safe(gar,"Revenus_Mensuels") + _safe(gar,"Autres_Revenus")).replace(0, np.nan)
        gar["EX_Garant_Liability_Pressure"] = (_safe(gar,"Charges_Estimees") / rev_tot).round(3)
        s["Garants_KYC"] = gar

    # ── Prets (5) ─────────────────────────────────────────────────────────────
    if not prt.empty:
        if not rem.empty and "ID_Pret" in rem.columns:
            int_enc = rem.groupby("ID_Pret")["Montant_Interets"].sum().to_dict()
            prt["EX_Effective_Yield_Realization"] = (
                _safe(prt,"ID_Pret").map(int_enc).fillna(0)
                / _safe(prt,"Cout_Credit",1).replace(0, np.nan)
            ).round(3)
        else:
            prt["EX_Effective_Yield_Realization"] = np.nan

        if "Motif" in prt.columns and "Statut" in prt.columns:
            def_mask = prt["Statut"].str.upper().isin(["DEFAUT","LITIGE"])
            def _def_by_motif(motif_val):
                sub = prt[prt["Motif"] == motif_val]
                if sub.empty: return 0.0
                return round(sub["Statut"].str.upper().isin(["DEFAUT","LITIGE"]).mean(), 3)
            prt["EX_Default_Probability_By_Motif"] = prt["Motif"].apply(_def_by_motif)
        else:
            prt["EX_Default_Probability_By_Motif"] = 0.0

        if not cli.empty and "ID_Client" in cli.columns and "Pers_Charge" in cli.columns:
            pch_map = cli.set_index("ID_Client")["Pers_Charge"].to_dict()
            prt["_pch"] = _safe(prt,"ID_Client").map(pch_map).fillna(1).replace(0,1)
            prt["EX_Loan_Size_vs_Family_Size"] = (_safe(prt,"Montant_Capital") / prt["_pch"]).round(0)
            prt.drop(columns=["_pch"], inplace=True, errors="ignore")
        else:
            prt["EX_Loan_Size_vs_Family_Size"] = np.nan

        if "Type_Pret" in prt.columns:
            prt["EX_Duration_Anomaly_Detection"] = (
                prt.groupby("Type_Pret")["Duree_Semaines"]
                   .transform(lambda x: (x - x.mean()) / x.std() if x.std() > 0 else 0.0)
            ).round(2)
        else:
            prt["EX_Duration_Anomaly_Detection"] = 0.0

        prt["EX_Endettement_Critical_Threshold"] = (
            _safe(prt,"Taux_Endettement").astype(float) > 40.0
        ).astype(int)
        s["Prets_Master"] = prt

    # ── Prets_Update (2) ─────────────────────────────────────────────────────
    if not upd.empty:
        modif_cols = ["Montant_Capital","Taux_Hebdo","Duree_Semaines","Motif","Moyen_Transfert"]
        if not prt.empty and "ID_Pret" in prt.columns:
            def _pattern(row):
                base = prt[prt["ID_Pret"] == row.get("ID_Pret","")]
                if base.empty: return 0
                base = base.iloc[0]
                return sum(1 for col in modif_cols if col in row and col in base and str(row[col]) != str(base[col]))
            upd["EX_Modification_Pattern_Clustering"] = upd.apply(_pattern, axis=1)
        else:
            upd["EX_Modification_Pattern_Clustering"] = 0

        if not prt.empty and "Montant_Total" in prt.columns and "ID_Pret" in upd.columns:
            mt_orig = prt.set_index("ID_Pret")["Montant_Total"].to_dict()
            upd["EX_Version_Drift_Analysis"] = (
                _safe(upd,"Montant_Total") - _safe(upd,"ID_Pret").map(mt_orig).fillna(0)
            ).round(0)
        else:
            upd["EX_Version_Drift_Analysis"] = np.nan
        s["Prets_Update"] = upd

    # ── Remboursements (5) ────────────────────────────────────────────────────
    if not rem.empty:
        rem["EX_Early_Payment_Propensity"] = (_safe(rem,"Jours_Retard").astype(float) < 0).astype(int)
        pen = _safe(rem,"Penalites_Retard").astype(float)
        mv  = _safe(rem,"Montant_Verse").astype(float)
        rem["EX_Collection_Cost_Efficiency"] = (mv / (mv + pen).replace(0, np.nan)).round(4)
        def _anomaly_hour(ts):
            try:
                h = pd.to_datetime(str(ts), errors="coerce").hour
                return 1 if pd.notna(h) and (h >= 23 or h < 5) else 0
            except Exception:
                return 0
        rem["EX_Timestamp_Anomaly_Log"] = rem.get("Timestamp", pd.Series(dtype=str)).apply(_anomaly_hour)
        rem["EX_Payment_Consistency_Score"] = (
            rem.groupby("ID_Client")["Montant_Verse"].transform(lambda x: x.std(ddof=0)).fillna(0).round(2)
        ) if "ID_Client" in rem.columns else 0.0
        if "Moyen_Paiement" in rem.columns and "Statut_Paiement" in rem.columns:
            good = rem["Statut_Paiement"].str.upper().isin(["PAYE","PONCTUEL","ANTICIPE"])
            rate = rem.groupby("Moyen_Paiement").apply(lambda g: good.reindex(g.index).mean())
            rem["EX_Recovery_Channel_Success"] = _safe(rem,"Moyen_Paiement").map(rate).round(3)
        else:
            rem["EX_Recovery_Channel_Success"] = np.nan
        s["Remboursements"] = rem

    # ── Ajustements (5) ──────────────────────────────────────────────────────
    if not adj.empty:
        if "ID_Pret" in adj.columns and not prt.empty and "ID_Client" in prt.columns:
            cli_map = prt.set_index("ID_Pret")["ID_Client"].to_dict()
            adj["_cli"] = _safe(adj,"ID_Pret").map(cli_map)
            adj["EX_Adjustment_Frequency_Per_Client"] = adj.groupby("_cli")["_cli"].transform("count").fillna(0)
            adj.drop(columns=["_cli"], inplace=True, errors="ignore")
        else:
            adj["EX_Adjustment_Frequency_Per_Client"] = 0
        if not prt.empty and "Statut" in prt.columns and "Raison" in adj.columns:
            def _r2d(raison):
                ids = adj[adj["Raison"] == raison].get("ID_Pret", pd.Series(dtype=str))
                sub = prt[prt["ID_Pret"].isin(ids)]
                return 0.0 if sub.empty else round(sub["Statut"].str.upper().isin(["DEFAUT","LITIGE"]).mean(), 3)
            adj["EX_Reason_to_Default_Bridge"] = adj.get("Raison", pd.Series(dtype=str)).apply(_r2d)
        else:
            adj["EX_Reason_to_Default_Bridge"] = np.nan
        if not prt.empty and "ID_Pret" in adj.columns and "Montant_Capital" in prt.columns:
            cap_map  = prt.set_index("ID_Pret")["Montant_Capital"].to_dict()
            cap_init = _safe(adj,"ID_Pret").map(cap_map).replace(0, np.nan)
            adj["EX_Amount_Expansion_Factor"] = (_safe(adj,"Montant_Additionnel") / cap_init).round(4)
        else:
            adj["EX_Amount_Expansion_Factor"] = np.nan
        if not prt.empty and "Date_Fin" in prt.columns:
            fm = prt.set_index("ID_Pret")["Date_Fin"].to_dict()
            df_ = pd.to_datetime(_safe(adj,"ID_Pret").map(fm), errors="coerce")
            da  = pd.to_datetime(adj.get("Date_Creation"), errors="coerce")
            adj["EX_Timing_of_Adjustment"] = (df_ - da).dt.days
        else:
            adj["EX_Timing_of_Adjustment"] = np.nan
        if not prt.empty and "Date_Creation" in prt.columns:
            dc_map = prt.set_index("ID_Pret")["Date_Creation"].to_dict()
            dp = pd.to_datetime(_safe(adj,"ID_Pret").map(dc_map), errors="coerce")
            da = pd.to_datetime(adj.get("Date_Creation"), errors="coerce")
            adj["EX_Adjustment_Latency"] = (da - dp).dt.days
        else:
            adj["EX_Adjustment_Latency"] = np.nan
        s["Ajustements_Echeances"] = adj

    return s


# ══════════════════════════════════════════════════════════════════════════════
# FILTRE GLOBAL
# ══════════════════════════════════════════════════════════════════════════════

def render_filter_bar(sheets: dict) -> dict:
    """Affiche les filtres, retourne un nouveau dict de DataFrames filtrés."""
    st.markdown(
        f'<div style="font-family:{_FONT};font-size:0.55rem;letter-spacing:2px;'
        f'text-transform:uppercase;color:{_C["subtext"]};margin-bottom:6px;">'
        f'FILTRE GLOBAL  ·  EXPLORATION</div>',
        unsafe_allow_html=True,
    )

    prt = sheets.get("Prets_Master", pd.DataFrame())
    rem = sheets.get("Remboursements", pd.DataFrame())
    cli = sheets.get("Clients_KYC", pd.DataFrame())

    fc1, fc2, fc3, fc4 = st.columns(4)

    with fc1:
        statuts = ["Tous"] + (sorted(prt["Statut"].dropna().unique().tolist()) if not prt.empty and "Statut" in prt.columns else [])
        f_statut = st.selectbox("Statut pret", statuts, key="ex_f_statut")
    with fc2:
        offres = ["Tous"] + (sorted(prt["Offre"].dropna().unique().tolist()) if not prt.empty and "Offre" in prt.columns else [])
        f_offre = st.selectbox("Offre", offres, key="ex_f_offre")
    with fc3:
        sp_vals = ["Tous"] + (sorted(rem["Statut_Paiement"].dropna().unique().tolist()) if not rem.empty and "Statut_Paiement" in rem.columns else [])
        f_sp = st.selectbox("Statut paiement", sp_vals, key="ex_f_sp")
    with fc4:
        villes = ["Tous"] + (sorted(cli["Ville"].dropna().unique().tolist()) if not cli.empty and "Ville" in cli.columns else [])
        f_ville = st.selectbox("Ville", villes, key="ex_f_ville")

    st.divider()

    # Appliquer les filtres
    filtered = {k: df.copy() for k, df in sheets.items() if isinstance(df, pd.DataFrame)}

    if f_statut != "Tous" and not prt.empty and "Statut" in prt.columns:
        ids_ok = set(prt[prt["Statut"] == f_statut].get("ID_Pret", pd.Series(dtype=str)).tolist())
        for sh in ("Prets_Master","Prets_Update","Remboursements","Ajustements_Echeances"):
            df = filtered.get(sh, pd.DataFrame())
            if not df.empty and "ID_Pret" in df.columns:
                filtered[sh] = df[df["ID_Pret"].isin(ids_ok)]
        if not filtered.get("Prets_Master",pd.DataFrame()).empty:
            ids_cli = set(filtered["Prets_Master"].get("ID_Client", pd.Series(dtype=str)).dropna().tolist())
            if not filtered.get("Clients_KYC",pd.DataFrame()).empty and "ID_Client" in filtered["Clients_KYC"].columns:
                filtered["Clients_KYC"] = filtered["Clients_KYC"][filtered["Clients_KYC"]["ID_Client"].isin(ids_cli)]

    if f_offre != "Tous" and "Prets_Master" in filtered and "Offre" in filtered["Prets_Master"].columns:
        filtered["Prets_Master"] = filtered["Prets_Master"][filtered["Prets_Master"]["Offre"] == f_offre]

    if f_sp != "Tous" and "Remboursements" in filtered and "Statut_Paiement" in filtered["Remboursements"].columns:
        filtered["Remboursements"] = filtered["Remboursements"][filtered["Remboursements"]["Statut_Paiement"] == f_sp]

    if f_ville != "Tous" and "Clients_KYC" in filtered and "Ville" in filtered["Clients_KYC"].columns:
        filtered["Clients_KYC"] = filtered["Clients_KYC"][filtered["Clients_KYC"]["Ville"] == f_ville]

    return filtered


# ══════════════════════════════════════════════════════════════════════════════
# VISUELS U2  (6+ par feuille, axes causaux)
# ══════════════════════════════════════════════════════════════════════════════

def _explore_clients(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.info("Table Clients_KYC vide apres filtres."); return

    _ctx("SOURCE : Clients_KYC  |  AXEE : mecanismes de survie financiere, "
         "concentration geographique, profils demographiques")

    c1, c2 = st.columns(2)

    # 1. Matrice survie financiere
    with c1:
        rev  = _safe(df,"Revenus_Mensuels").astype(float)
        chg  = _safe(df,"Charges_Estimees").astype(float)
        ral  = rev - chg
        lgt_col = df.get("Statut_Logement", pd.Series("Autre", index=df.index)).fillna("Autre")
        lgt_map = {"Locataire":_C["warning"],"Proprietaire":_C["success"],
                   "Heberge":_C["accent"],"Hébergé":_C["accent"]}
        fig = go.Figure()
        for lval in lgt_col.unique():
            mask = lgt_col == lval; sub = df[mask]
            fig.add_trace(go.Scatter(
                x=rev[mask], y=chg[mask], mode="markers", name=str(lval),
                marker=dict(size=12, color=lgt_map.get(str(lval),_C["neutral"]),
                            line=dict(width=1.5, color=_C["bg"])),
                text=sub.get("Nom_Complet", sub.index).astype(str),
                customdata=ral[mask].round(0),
                hovertemplate="<b>%{text}</b><br>Rev : %{x:,}<br>Chg : %{y:,}<br>Reste : %{customdata:,}<extra></extra>"))
        mx = max(rev.max(), chg.max()) * 1.05
        fig.add_trace(go.Scatter(x=[0,mx], y=[0,mx], mode="lines",
            line=dict(color=_C["critical"],dash="dot",width=1),showlegend=False,hoverinfo="skip"))
        fig.update_layout(**_L(title="SURVIE FINANCIERE  (ligne rouge = seuil critique)",
                               height=310, xaxis_title="Revenus", yaxis_title="Charges"))
        st.plotly_chart(fig, use_container_width=True)

    # 2. Strip plot revenus par secteur
    with c2:
        if "Secteur_Activite" in df.columns:
            bench = df.get("EX_Neighborhood_Income_Benchmark", pd.Series(1.0, index=df.index)).fillna(1.0)
            fig   = go.Figure()
            for i, sec in enumerate(df["Secteur_Activite"].dropna().unique()):
                sub = df[df["Secteur_Activite"]==sec]; b = bench.reindex(sub.index).fillna(1.0)
                np.random.seed(i)
                jit = np.random.uniform(-0.15, 0.15, len(sub))
                fig.add_trace(go.Scatter(
                    x=[str(sec)]*len(sub), y=sub["Revenus_Mensuels"].tolist(), mode="markers",
                    name=str(sec),
                    marker=dict(size=10,
                        color=[_C["success"] if v>=1.0 else _C["critical"] for v in b],
                        line=dict(width=1, color=_C["bg"]), opacity=0.85),
                    text=sub.get("Nom_Complet", sub.index).astype(str), customdata=b.round(2),
                    hovertemplate="<b>%{text}</b><br>%{y:,} FCFA<br>Bench : %{customdata}x<extra></extra>"))
            fig.update_layout(**_L(title="DISPERSION REVENUS/SECTEUR  (vert = dessus moy. quartier)",
                                   height=310, showlegend=False, xaxis_title="Secteur", yaxis_title="Revenus"))
            st.plotly_chart(fig, use_container_width=True)

    c3, c4 = st.columns(2)

    # 3. Heatmap geographique
    with c3:
        if "Ville" in df.columns and "Quartier" in df.columns:
            pivot = df.groupby(["Ville","Quartier"]).size().reset_index(name="n")
            fig   = go.Figure(go.Heatmap(
                x=pivot["Quartier"], y=pivot["Ville"], z=pivot["n"],
                colorscale=[[0,_C["card"]],[0.4,"#1a3a5c"],[1,_C["accent"]]],
                text=pivot["n"], texttemplate="%{text}",
                hovertemplate="<b>%{y}  ·  %{x}</b><br>Clients : %{z}<extra></extra>",
                colorbar=dict(title="Clients",tickfont=dict(size=9,family=_FONT)),
                xgap=2, ygap=2))
            fig.update_layout(**_L(title="DENSITE GEOGRAPHIQUE  Ville x Quartier",
                                   height=300, xaxis=dict(tickangle=-30)))
            st.plotly_chart(fig, use_container_width=True)

    # 4. Parallel categories
    with c4:
        cols_par = ["Etat_Civil","Genre","Statut_Pro"]
        if all(c in df.columns for c in cols_par):
            df2 = df[cols_par].fillna("N/A")
            fig = go.Figure(go.Parcats(
                dimensions=[
                    dict(label="Etat Civil", values=df2["Etat_Civil"]),
                    dict(label="Genre",       values=df2["Genre"]),
                    dict(label="Statut Pro",  values=df2["Statut_Pro"]),
                ],
                line=dict(color=_PAL[0], colorscale=[[0,_PAL[0]],[1,_PAL[1]]]),
                labelfont=dict(family=_FONT, size=10, color=_C["text"]),
                tickfont=dict(family=_FONT, size=9, color=_C["subtext"]),
            ))
            fig.update_layout(**_L(title="FLUX DEMOGRAPHIQUE  Etat Civil > Genre > Statut Pro",height=300))
            st.plotly_chart(fig, use_container_width=True)

    c5, c6 = st.columns(2)

    # 5. Scatter AML risk vs revenu
    with c5:
        aml_risk = df.get("EX_AML_Risk_Correlation", pd.Series(0, index=df.index)).fillna(0)
        fig = go.Figure()
        for val, color, label in [(1,_C["critical"],"Risque AML"),(0,_C["success"],"AML OK")]:
            mask = aml_risk == val; sub = df[mask]
            if not sub.empty:
                fig.add_trace(go.Scatter(x=sub["Revenus_Mensuels"], y=_safe(sub,"Patrimoine_Declare"),
                    mode="markers", name=label,
                    marker=dict(size=11, color=color, line=dict(width=1.5, color=_C["bg"])),
                    text=sub.get("Nom_Complet", sub.index).astype(str),
                    hovertemplate="<b>%{text}</b><br>Rev : %{x:,}<br>Patr : %{y:,}<extra></extra>"))
        fig.update_layout(**_L(title="PROFIL AML x PATRIMOINE",height=290,xaxis_title="Revenus",yaxis_title="Patrimoine"))
        st.plotly_chart(fig, use_container_width=True)

    # 6. Bar volatilite revenu par secteur
    with c6:
        if "Secteur_Activite" in df.columns and "EX_Income_Volatility_Matrix" in df.columns:
            vol = df.groupby("Secteur_Activite")["EX_Income_Volatility_Matrix"].mean().sort_values()
            fig = go.Figure(go.Bar(x=vol.values, y=vol.index.astype(str), orientation="h",
                marker_color=[_C["critical"] if v>0.4 else _C["warning"] if v>0.2 else _C["accent"] for v in vol.values],
                text=[f"{v:.2f}" for v in vol.values], textposition="outside"))
            fig.update_layout(**_L(title="VOLATILITE REVENUS PAR SECTEUR  (EX_Income_Volatility)",
                                   height=290, showlegend=False))
            st.plotly_chart(fig, use_container_width=True)

    # 7. Scatter efficacite seniorite
    if "EX_Professional_Seniority_Efficiency" in df.columns and "Anciennete_Emploi" in df.columns:
        eff = df["EX_Professional_Seniority_Efficiency"].fillna(0)
        fig = go.Figure()
        if "Statut_Pro" in df.columns:
            for i, sp in enumerate(df["Statut_Pro"].dropna().unique()):
                sub = df[df["Statut_Pro"]==sp]
                fig.add_trace(go.Scatter(
                    x=sub["Anciennete_Emploi"], y=sub["EX_Professional_Seniority_Efficiency"].fillna(0),
                    mode="markers", name=str(sp),
                    marker=dict(size=10, color=_PAL[i%len(_PAL)], line=dict(width=1,color=_C["bg"])),
                    text=sub.get("Nom_Complet",sub.index).astype(str),
                    hovertemplate="<b>%{text}</b><br>Anciennete : %{x} mois<br>Efficacite : %{y:.1f}<extra></extra>"))
        fig.update_layout(**_L(title="EFFICACITE PAR SENIORITE  (revenu net / anciennete)",
                               height=260, xaxis_title="Anciennete (mois)", yaxis_title="Efficacite"))
        st.plotly_chart(fig, use_container_width=True)

    # 8. PCA 2D Clients
    st.divider()
    _ctx("CLUSTER MAP  PCA 2D  —  Similarite multi-dimensionnelle des clients")
    cli_feats = [
        "Revenus_Mensuels","Autres_Revenus","Charges_Estimees","Anciennete_Emploi",
        "Pers_Charge","Patrimoine_Declare","Checking_frequency",
        "DX_Living_Cost_Ratio","DX_Disposable_Income_Stability",
        "DX_Digital_Accessibility_Grade","DX_Professional_Trajectory_Index",
        "EX_Income_Volatility_Matrix","EX_Contact_Reliability_Index",
        "EX_Neighborhood_Income_Benchmark","EX_Professional_Seniority_Efficiency",
    ]
    # Couleur prioritaire : FL_Risk_Target si present, sinon DX_Living_Cost_Ratio
    color_col = "FL_Risk_Target" if "FL_Risk_Target" in df.columns else "DX_Living_Cost_Ratio"
    color_lbl = "Risque" if color_col == "FL_Risk_Target" else "LCR"
    c_pca1, c_pca2 = st.columns(2)
    with c_pca1:
        fig = _pca_ex(df, cli_feats, "Nom_Complet", color_col, color_lbl,
                      "CLIENTS  (couleur = Risque)")
        st.plotly_chart(fig, use_container_width=True)
    with c_pca2:
        fig = _pca_ex(df, cli_feats, "Nom_Complet", "DX_Disposable_Income_Stability",
                      "Stabilite rev.", "CLIENTS  (couleur = Stabilite Revenus)",
                      colorscale=[[0,"#c0392b"],[0.5,"#f39c12"],[1,"#2ecc71"]])
        st.plotly_chart(fig, use_container_width=True)


def _explore_garants(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.info("Table Garants_KYC vide."); return

    _ctx("SOURCE : Garants_KYC  |  AXEE : solidite mecanique, "
         "interoperabilite bancaire, pression financiere comparee")

    # Fiche comparative
    st.markdown(
        f'<div style="font-family:{_FONT};font-size:0.58rem;letter-spacing:2px;'
        f'color:{_C["subtext"]};text-transform:uppercase;margin-bottom:8px;">'
        f'COMPARAISON PROFILS GARANTS</div>',
        unsafe_allow_html=True,
    )
    kpis = [
        ("Revenus Mensuels",       "Revenus_Mensuels",               "{:,.0f} FCFA"),
        ("Patrimoine Declare",     "Patrimoine_Declare",             "{:,.0f} FCFA"),
        ("Charges Estimees",       "Charges_Estimees",               "{:,.0f} FCFA"),
        ("Personnes a charge",     "Pers_Charge",                    "{:.0f}"),
        ("Revenu disponible",      "EX_Guarantor_Disposable_Income", "{:,.0f} FCFA"),
        ("Pression financiere",    "EX_Garant_Liability_Pressure",   "{:.1%}"),
        ("Leverage patrimoine",    "DX_Guarantee_Leverage_Score",    "{:.1f}x"),
        ("Disponibilite",          "DX_Garant_Availability_Score",   "{:.0f} / 2"),
    ]
    cols = st.columns(len(df)+1)
    with cols[0]:
        for label,_,_ in kpis:
            st.markdown(f'<div style="font-family:{_FONT};font-size:0.58rem;color:{_C["subtext"]};'
                        f'padding:5px 0;border-bottom:1px solid {_C["border"]};">{label}</div>',
                        unsafe_allow_html=True)
    for i,(_, row) in enumerate(df.iterrows()):
        with cols[i+1]:
            nom = str(row.get("Nom_Complet", f"GARANT {i+1}"))
            st.markdown(f'<div style="font-family:{_FONT};font-size:0.58rem;color:{_C["accent"]};'
                        f'font-weight:700;padding:5px 0;border-bottom:1px solid {_C["border"]};">'
                        f'{nom[:20]}</div>', unsafe_allow_html=True)
            for _,kcol,kfmt in kpis:
                val = row.get(kcol, None)
                if val is None or (isinstance(val, float) and np.isnan(val)):
                    display = "—"
                else:
                    try:
                        display = kfmt.format(float(val))
                    except Exception:
                        display = str(val)
                st.markdown(f'<div style="font-family:{_FONT};font-size:0.58rem;color:{_C["text"]};'
                            f'padding:5px 0;border-bottom:1px solid {_C["border"]};">{display}</div>',
                            unsafe_allow_html=True)

    st.markdown("<br>", unsafe_allow_html=True)
    c1, c2 = st.columns(2)

    # 1. Scatter wealth vs risk
    with c1:
        wtr = df.get("EX_Wealth_to_Risk_Ratio", pd.Series(1.0, index=df.index)).fillna(1.0)
        fig = go.Figure(go.Scatter(
            x=_safe(df,"Revenus_Mensuels"), y=_safe(df,"Patrimoine_Declare"),
            mode="markers+text",
            marker=dict(size=(wtr.clip(0,100)/wtr.clip(0,100).max()*18+8).tolist() if wtr.max()>0 else [12]*len(df),
                        color=_PAL[0], line=dict(width=1.5, color=_C["bg"])),
            text=df.get("Nom_Complet", df.index).astype(str), textposition="top center",
            customdata=wtr.round(1),
            hovertemplate="<b>%{text}</b><br>Rev : %{x:,}<br>Patr : %{y:,}<br>W/R : %{customdata}<extra></extra>"))
        fig.update_layout(**_L(title="WEALTH-TO-RISK  (taille = ratio patrimoine/risque)",height=280,showlegend=False))
        st.plotly_chart(fig, use_container_width=True)

    # 2. Bar logistique entite x canal
    with c2:
        if "Entite_Financiere" in df.columns and "Moyen_Transfert" in df.columns:
            cross = df.groupby(["Entite_Financiere","Moyen_Transfert"]).size().reset_index(name="n")
            fig   = go.Figure()
            for i,mt in enumerate(cross["Moyen_Transfert"].unique()):
                sub = cross[cross["Moyen_Transfert"]==mt]
                fig.add_trace(go.Bar(name=str(mt),x=sub["Entite_Financiere"],y=sub["n"],marker_color=_PAL[i%len(_PAL)]))
            fig.update_layout(**_L(title="EFFICACITE LOGISTIQUE  Entite x Canal",barmode="stack",height=280))
            st.plotly_chart(fig, use_container_width=True)

    c3, c4 = st.columns(2)

    # 3. Bar pression liability
    with c3:
        lp = df.get("EX_Garant_Liability_Pressure", pd.Series(0, index=df.index)).fillna(0)
        noms = df.get("Nom_Complet", df.index.astype(str))
        fig = go.Figure(go.Bar(x=noms.tolist(), y=lp.tolist(),
            marker_color=[_C["critical"] if v>0.6 else _C["warning"] if v>0.35 else _C["success"] for v in lp],
            text=[f"{v:.1%}" for v in lp], textposition="outside"))
        fig.add_hline(y=0.6, line_dash="dot", line_color=_C["critical"],
            annotation_text="Seuil danger 60%", annotation_font_size=9)
        fig.update_layout(**_L(title="PRESSION CHARGES / REVENUS TOTAUX",height=280,showlegend=False,yaxis=dict(tickformat=".0%")))
        st.plotly_chart(fig, use_container_width=True)

    # 4. Interoperabilite bancaire
    with c4:
        interop = df.get("EX_Bank_Transfer_Interoperability", pd.Series(0, index=df.index)).fillna(0)
        cnt     = interop.value_counts()
        fig     = go.Figure(go.Pie(
            labels=["Compatible","Incompatible"], values=[cnt.get(1,0), cnt.get(0,0)],
            marker_colors=[_C["success"], _C["critical"]], hole=0.5))
        fig.update_layout(**_L(title="INTEROPERABILITE BANCAIRE  (meme canal que client)",height=280))
        st.plotly_chart(fig, use_container_width=True)

    c5, c6 = st.columns(2)

    # 5. Box revenus par entite financiere
    with c5:
        if "Entite_Financiere" in df.columns:
            fig = go.Figure()
            for i,ef in enumerate(df["Entite_Financiere"].dropna().unique()):
                sub = df[df["Entite_Financiere"]==ef]["Revenus_Mensuels"].dropna()
                fig.add_trace(go.Box(y=sub, name=str(ef), marker_color=_PAL[i%len(_PAL)], boxpoints="all", jitter=0.3))
            fig.update_layout(**_L(title="REVENUS PAR ENTITE FINANCIERE",height=280,showlegend=False))
            st.plotly_chart(fig, use_container_width=True)

    # 6. Scatter famille vs disponible
    with c6:
        disp = df.get("EX_Guarantor_Disposable_Income", pd.Series(0, index=df.index)).fillna(0)
        fig  = go.Figure(go.Scatter(
            x=_safe(df,"Pers_Charge"), y=disp,
            mode="markers+text",
            marker=dict(size=13, color=[_C["critical"] if v<0 else _C["success"] for v in disp],
                        line=dict(width=1.5, color=_C["bg"])),
            text=df.get("Nom_Complet", df.index).astype(str), textposition="top center",
            hovertemplate="<b>%{text}</b><br>Personnes a charge : %{x}<br>Disponible : %{y:,}<extra></extra>"))
        fig.add_hline(y=0, line_dash="dot", line_color=_C["neutral"], line_width=1)
        fig.update_layout(**_L(title="PRESSION FAMILIALE vs REVENU DISPONIBLE",height=280,
                               showlegend=False,xaxis_title="Pers. a charge",yaxis_title="Disponible"))
        st.plotly_chart(fig, use_container_width=True)

    # 7. PCA 2D Garants
    st.divider()
    _ctx("CLUSTER MAP  PCA 2D  —  Positionnement comparatif des garants")
    gar_feats = [
        "Revenus_Mensuels","Autres_Revenus","Charges_Estimees",
        "Patrimoine_Declare","Pers_Charge","Anciennete_Emploi","Checking_frequency",
        "DX_Guarantee_Leverage_Score","DX_Family_Pressure_Index",
        "DX_Garant_Availability_Score","DX_Net_Worth_Credibility_Check",
        "EX_Guarantor_Disposable_Income","EX_Wealth_to_Risk_Ratio","EX_Garant_Liability_Pressure",
    ]
    g_pca1, g_pca2 = st.columns(2)
    with g_pca1:
        fig = _pca_ex(df, gar_feats, "Nom_Complet", "EX_Garant_Liability_Pressure",
                      "Pression fin.", "GARANTS  (couleur = Pression Charges/Revenus)")
        st.plotly_chart(fig, use_container_width=True)
    with g_pca2:
        fig = _pca_ex(df, gar_feats, "Nom_Complet", "DX_Guarantee_Leverage_Score",
                      "Leverage", "GARANTS  (couleur = Leverage Patrimoine)",
                      colorscale=[[0,"#c0392b"],[0.5,"#f39c12"],[1,"#2ecc71"]])
        st.plotly_chart(fig, use_container_width=True)


def _explore_prets(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.info("Table Prets_Master vide."); return

    _ctx("SOURCE : Prets_Master + Prets_Update  |  AXEE : elasticite taux/duree, "
         "pression d'endettement par age, performance hierarchique")

    # Fusion avec Update (version la plus recente)
    upd = sheets.get("Prets_Update", pd.DataFrame())
    if not upd.empty and "ID_Pret" in upd.columns and "Version" in upd.columns:
        latest = upd.sort_values("Version", ascending=False).drop_duplicates("ID_Pret").set_index("ID_Pret")
        df_m = df.copy().set_index("ID_Pret")
        for col in ["Taux_Hebdo","Duree_Semaines","Montant_Capital","Statut","Motif"]:
            if col in latest.columns:
                df_m[col] = latest[col].combine_first(df_m.get(col, pd.Series(dtype=object)))
        df_m = df_m.reset_index()
    else:
        df_m = df.copy()

    # Calcul tranches d'age (STRING, pas Categorical)
    cli = sheets.get("Clients_KYC", pd.DataFrame())
    if not cli.empty and "ID_Client" in cli.columns and "Date_Naissance" in cli.columns:
        cli2 = cli.copy()
        cli2["Age"] = _age_calc(cli2["Date_Naissance"])
        cli2["Tranche_Age"] = _age_bin(cli2["Age"])   # retourne str, pas Categorical
        age_map = cli2.set_index("ID_Client")["Tranche_Age"].to_dict()
        df_m["Tranche_Age"] = _safe(df_m,"ID_Client").map(age_map)
    else:
        df_m["Tranche_Age"] = np.nan

    TRANCHES = ["<25","25-34","35-44","45-54","55+"]   # ordre fixe, pas .cat.categories

    c1, c2 = st.columns(2)

    # 1. Elasticite taux / duree
    with c1:
        fig = go.Figure()
        for i,of in enumerate(df_m.get("Offre",pd.Series()).dropna().unique()):
            sub  = df_m[df_m["Offre"]==of]
            over = sub.get("EX_Endettement_Critical_Threshold", pd.Series(0,index=sub.index)).fillna(0)
            eyr  = sub.get("EX_Effective_Yield_Realization", pd.Series(0,index=sub.index)).fillna(0).clip(0,1)
            fig.add_trace(go.Scatter(
                x=sub["Duree_Semaines"], y=sub["Taux_Hebdo"], mode="markers", name=str(of),
                marker=dict(
                    size=[(14 if v else 8) for v in over],
                    color=_PAL[i%len(_PAL)],
                    symbol=["diamond" if v else "circle" for v in over],
                    line=dict(width=1.5, color=[_C["critical"] if v else _C["bg"] for v in over])),
                text=sub["ID_Pret"].astype(str),
                customdata=sub.get("Taux_Endettement",pd.Series(0,index=sub.index)).fillna(0).round(1),
                hovertemplate="<b>%{text}</b><br>%{x} sem.  Taux %{y:.2%}<br>Endett. : %{customdata}%<extra></extra>"))
        fig.update_layout(**_L(title="ELASTICITE TAUX/DUREE  (losange = endettement > 40%)",
                               height=310, xaxis_title="Duree (semaines)",
                               yaxis=dict(tickformat=".1%"), yaxis_title="Taux Hebdo"))
        st.plotly_chart(fig, use_container_width=True)

    # 2. Boxplot pression par tranche d'age  ← BUG CORRIGE ICI
    with c2:
        df_bp = df_m.dropna(subset=["Tranche_Age","Taux_Endettement"])
        if not df_bp.empty:
            crit  = df_bp.get("EX_Endettement_Critical_Threshold", pd.Series(0,index=df_bp.index)).fillna(0)
            fig   = go.Figure()
            # On itere sur la liste de strings, PAS sur .cat.categories
            for i, tr in enumerate(TRANCHES):
                sub    = df_bp[df_bp["Tranche_Age"] == tr]
                if sub.empty: continue
                n_crit = int(crit.reindex(sub.index).fillna(0).sum())
                fig.add_trace(go.Box(
                    y=sub["Taux_Endettement"], name=str(tr),
                    marker_color=_C["critical"] if n_crit > 0 else _PAL[i%len(_PAL)],
                    boxpoints="all", jitter=0.3,
                    hovertemplate=f"<b>{tr}</b><br>Endett. : %{{y:.1f}}%<extra></extra>"))
            fig.add_hline(y=40, line_dash="dot", line_color=_C["critical"],
                annotation_text="Seuil critique 40%",
                annotation_font=dict(size=9, color=_C["critical"]))
            fig.update_layout(**_L(title="PRESSION D'ENDETTEMENT PAR TRANCHE D'AGE",
                                   height=310, showlegend=False, yaxis_title="Taux Endettement (%)"))
            st.plotly_chart(fig, use_container_width=True)
        else:
            st.plotly_chart(_ef("Date_Naissance absente ou jointure impossible.", 310), use_container_width=True)

    # 3. Sunburst performance
    hier = df_m.groupby(["Offre","Type_Pret","Statut"]).size().reset_index(name="n")
    fig  = px.sunburst(hier, path=["Offre","Type_Pret","Statut"], values="n", color="Statut",
        color_discrete_map={"ACTIF":"#58a6ff","TERMINE":"#2ecc71",
                            "EN_RETARD":"#f39c12","UPDATED":"#5a6a7a",
                            "DEFAUT":"#c0392b","LITIGE":"#c0392b"})
    fig.update_traces(textfont=dict(family=_FONT,size=11),
        hovertemplate="<b>%{label}</b><br>Prets : %{value}<br>Part : %{percentParent:.1%}<extra></extra>")
    fig.update_layout(**_L(title="RADAR DE PERFORMANCE  Offre > Type > Statut", height=370))
    st.plotly_chart(fig, use_container_width=True)

    c3, c4 = st.columns(2)

    # 4. Scatter anomalie de duree
    with c3:
        if "EX_Duration_Anomaly_Detection" in df_m.columns and "Type_Pret" in df_m.columns:
            fig = go.Figure()
            for i,tp in enumerate(df_m["Type_Pret"].dropna().unique()):
                sub = df_m[df_m["Type_Pret"]==tp]
                anom = sub["EX_Duration_Anomaly_Detection"].fillna(0).abs()
                fig.add_trace(go.Scatter(
                    x=sub["Duree_Semaines"], y=sub["EX_Duration_Anomaly_Detection"].fillna(0),
                    mode="markers", name=str(tp),
                    marker=dict(size=(anom*6+8).clip(upper=20).tolist(),
                                color=_PAL[i%len(_PAL)], line=dict(width=1,color=_C["bg"])),
                    text=sub["ID_Pret"].astype(str),
                    hovertemplate="<b>%{text}</b><br>Duree : %{x} sem.<br>Z-score : %{y:.2f}<extra></extra>"))
            fig.add_hline(y=0, line_dash="dot", line_color=_C["neutral"], line_width=1)
            fig.update_layout(**_L(title="ANOMALIE DE DUREE PAR TYPE  (z-score)",height=290,
                                   xaxis_title="Duree (semaines)",yaxis_title="Z-score"))
            st.plotly_chart(fig, use_container_width=True)

    # 5. Bar rendement effectif
    with c4:
        if "EX_Effective_Yield_Realization" in df_m.columns:
            df_m2 = df_m.dropna(subset=["EX_Effective_Yield_Realization"])
            if not df_m2.empty:
                fig = go.Figure(go.Bar(
                    x=df_m2["ID_Pret"].astype(str),
                    y=df_m2["EX_Effective_Yield_Realization"],
                    marker_color=[_C["success"] if v>=0.9 else _C["warning"] if v>=0.5 else _C["critical"]
                                  for v in df_m2["EX_Effective_Yield_Realization"]],
                    text=[f"{v:.1%}" for v in df_m2["EX_Effective_Yield_Realization"]],
                    textposition="outside"))
                fig.add_hline(y=1.0, line_dash="dot", line_color=_C["neutral"])
                fig.update_layout(**_L(title="RENDEMENT EFFECTIF REALISE  (1.0 = 100% interets encaisses)",
                                       height=290, showlegend=False, yaxis=dict(tickformat=".0%")))
                st.plotly_chart(fig, use_container_width=True)

    # 6. Scatter capital vs pers a charge
    if "EX_Loan_Size_vs_Family_Size" in df_m.columns:
        fig = go.Figure()
        for i,of in enumerate(df_m.get("Offre",pd.Series()).dropna().unique()):
            sub = df_m[df_m["Offre"]==of]
            pch = _safe(cli,"Pers_Charge") if not cli.empty else pd.Series(0,index=sub.index)
            pch_vals = sub.get("_pch_tmp", sub["ID_Client"].map(
                cli.set_index("ID_Client")["Pers_Charge"].to_dict() if not cli.empty and "ID_Client" in cli.columns else {}
            ).fillna(0)) if not cli.empty and "ID_Client" in cli.columns and "Pers_Charge" in cli.columns else pd.Series(1,index=sub.index)
            fig.add_trace(go.Scatter(
                x=pch_vals, y=sub["Montant_Capital"],
                mode="markers", name=str(of),
                marker=dict(size=10, color=_PAL[i%len(_PAL)], line=dict(width=1,color=_C["bg"])),
                text=sub["ID_Pret"].astype(str),
                hovertemplate="<b>%{text}</b><br>Pers. : %{x}<br>Capital : %{y:,}<extra></extra>"))
        fig.update_layout(**_L(title="CAPITAL vs TAILLE MENAGE",height=260,
                               xaxis_title="Personnes a charge",yaxis_title="Montant Capital"))
        st.plotly_chart(fig, use_container_width=True)

    # PCA 2D Prets
    st.divider()
    _ctx("CLUSTER MAP  PCA 2D  —  Positionnement des prets dans l'espace contractuel")
    prt_feats_pca = [
        "Montant_Capital","Taux_Hebdo","Taux_Endettement","Duree_Semaines",
        "Montant_Versement","Montant_Total","Cout_Credit","Nb_Versements",
        "DX_Interest_Yield_Factor","DX_Debt_to_Income_Pressure",
        "DX_Motif_Risk_Weight","DX_Collateral_Coverage_Ratio",
        "EX_Endettement_Critical_Threshold","EX_Duration_Anomaly_Detection",
        "EX_Effective_Yield_Realization",
    ]
    df_pca_prt = df_m.copy()
    df_pca_prt["_prt_risk"] = (
        df_pca_prt.get("DX_Motif_Risk_Weight", pd.Series(0.5, index=df_pca_prt.index)).fillna(0.5) * 0.5
        + (df_pca_prt["Taux_Endettement"].fillna(0) / 100).clip(0, 1) * 0.5
    ).round(3)
    p_pca1, p_pca2 = st.columns(2)
    with p_pca1:
        fig = _pca_ex(df_pca_prt, prt_feats_pca, "ID_Pret", "_prt_risk",
                      "Risque contrat", "PRETS  (couleur = Motif Risk + Endettement)")
        st.plotly_chart(fig, use_container_width=True)
    with p_pca2:
        fig = _pca_ex(df_pca_prt, prt_feats_pca, "Nom_Complet", "Taux_Endettement",
                      "Endettement %", "PRETS  (couleur = Taux Endettement)",
                      colorscale=[[0,"#2ecc71"],[0.4,"#f39c12"],[1,"#c0392b"]])
        st.plotly_chart(fig, use_container_width=True)


def _explore_prets_update(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.info("Table Prets_Update vide."); return

    _ctx("SOURCE : Prets_Update  |  AXEE : amplitude des modifications, "
         "derive de montant par version, patterns de renegociation")

    prt = sheets.get("Prets_Master", pd.DataFrame())
    c1, c2 = st.columns(2)

    # 1. Version drift par pret
    with c1:
        if not prt.empty and "ID_Pret" in prt.columns and "Montant_Total" in prt.columns and "Montant_Total" in df.columns:
            mt_orig = prt.set_index("ID_Pret")["Montant_Total"].to_dict()
            drift   = df.copy()
            drift["MT_Orig"] = _safe(drift,"ID_Pret").map(mt_orig)
            drift["Drift"]   = _safe(drift,"Montant_Total") - drift["MT_Orig"].fillna(0)
            fig = go.Figure()
            for pid in drift["ID_Pret"].dropna().unique():
                sub = drift[drift["ID_Pret"]==pid].sort_values("Version")
                col = _C["critical"] if sub["Drift"].iloc[-1]>0 else _C["success"]
                fig.add_trace(go.Scatter(x=sub["Version"],y=sub["Drift"],mode="lines+markers",
                    name=str(pid),line=dict(color=col,width=1.5),marker=dict(size=7),
                    hovertemplate=f"<b>{pid}</b><br>V%{{x}}<br>Derive : %{{y:+,.0f}}<extra></extra>"))
            fig.add_hline(y=0,line_dash="dot",line_color=_C["neutral"],line_width=1)
            fig.update_layout(**_L(title="DERIVE MONTANT TOTAL PAR VERSION  (rouge = hausse)",
                                   height=290,xaxis_title="Version",yaxis_title="Ecart vs original"))
            st.plotly_chart(fig, use_container_width=True)

    # 2. Amplitude modifications
    with c2:
        if "EX_Modification_Pattern_Clustering" in df.columns:
            fig = go.Figure(go.Bar(
                x=df["ID_Pret_Updated"].astype(str) if "ID_Pret_Updated" in df.columns else df.index.astype(str),
                y=df["EX_Modification_Pattern_Clustering"],
                marker_color=[_C["critical"] if v>=3 else _C["warning"] if v>=2 else _C["accent"]
                              for v in df["EX_Modification_Pattern_Clustering"]],
                text=df["EX_Modification_Pattern_Clustering"],textposition="outside"))
            fig.update_layout(**_L(title="AMPLITUDE DES MODIFICATIONS  (rouge = 3+ champs)",
                                   height=290,showlegend=False,yaxis_title="Champs modifies"))
            st.plotly_chart(fig, use_container_width=True)

    c3, c4 = st.columns(2)

    # 3. Scatter delta vs velocity
    with c3:
        if "DX_Contract_Volatility_Delta" in df.columns and "DX_Update_Velocity" in df.columns:
            fig = go.Figure(go.Scatter(
                x=df["DX_Update_Velocity"],y=df["DX_Contract_Volatility_Delta"],mode="markers",
                marker=dict(size=12,color=[_C["critical"] if v>0 else _C["success"] for v in df["DX_Contract_Volatility_Delta"]],
                    line=dict(width=1.5,color=_C["bg"])),
                text=df["ID_Pret"].astype(str) if "ID_Pret" in df.columns else df.index.astype(str),
                hovertemplate="<b>%{text}</b><br>Velocity : %{x}j<br>Delta : %{y:+,.0f}<extra></extra>"))
            fig.add_hline(y=0,line_dash="dot",line_color=_C["neutral"],line_width=1)
            fig.update_layout(**_L(title="DELTA CONTRAT vs VELOCITY",height=280,
                xaxis_title="Jours depuis creation",yaxis_title="Delta Montant Total",showlegend=False))
            st.plotly_chart(fig, use_container_width=True)

    # 4. Gravite renegociation
    with c4:
        if "DX_Renegotiation_Gravity_Score" in df.columns:
            fig = go.Figure(go.Bar(
                x=df["ID_Pret_Updated"].astype(str) if "ID_Pret_Updated" in df.columns else df.index.astype(str),
                y=df["DX_Renegotiation_Gravity_Score"],
                marker_color=[_C["critical"] if v>0.10 else _C["accent"] for v in df["DX_Renegotiation_Gravity_Score"]],
                text=df["DX_Renegotiation_Gravity_Score"].round(3),textposition="outside"))
            fig.update_layout(**_L(title="GRAVITE RENEGOCIATION  (score keywords crise)",
                                   height=280,showlegend=False))
            st.plotly_chart(fig, use_container_width=True)

    c5, c6 = st.columns(2)

    # 5. Evolution taux par version
    with c5:
        if not prt.empty and "Taux_Hebdo" in prt.columns and "Taux_Hebdo" in df.columns:
            taux_orig = prt.set_index("ID_Pret")["Taux_Hebdo"].to_dict()
            fig = go.Figure()
            for pid in df["ID_Pret"].dropna().unique():
                sub  = df[df["ID_Pret"]==pid].sort_values("Version")
                orig = taux_orig.get(pid, np.nan)
                if not (isinstance(orig,float) and np.isnan(orig)):
                    vals = [orig] + sub["Taux_Hebdo"].tolist()
                    vers = ["V1"] + [f"V{v}" for v in sub["Version"]]
                    col  = _C["critical"] if vals[-1]>orig else _C["success"]
                    fig.add_trace(go.Scatter(x=vers,y=vals,mode="lines+markers",name=str(pid),
                        line=dict(color=col,width=1.5),marker=dict(size=6),
                        hovertemplate=f"<b>{pid}</b>  %{{x}}<br>Taux : %{{y:.2%}}<extra></extra>"))
            fig.update_layout(**_L(title="EVOLUTION TAUX HEBDO PAR VERSION",height=280,
                                   yaxis=dict(tickformat=".2%")))
            st.plotly_chart(fig, use_container_width=True)

    # 6. Fragilite structurelle
    with c6:
        if "DX_Structural_Fragility_Index" in df.columns:
            fig = go.Figure(go.Bar(
                x=df["ID_Pret"].astype(str) if "ID_Pret" in df.columns else df.index.astype(str),
                y=df["DX_Structural_Fragility_Index"],
                marker_color=[_C["critical"] if v>0.15 else _C["warning"] if v>0.08 else _C["accent"]
                              for v in df["DX_Structural_Fragility_Index"]],
                text=df["DX_Structural_Fragility_Index"].round(4),textposition="outside"))
            fig.update_layout(**_L(title="FRAGILITE STRUCTURELLE  (modifications/semaines duree)",
                                   height=280,showlegend=False))
            st.plotly_chart(fig, use_container_width=True)

    # 7. PCA 2D Prets Update
    st.divider()
    _ctx("CLUSTER MAP  PCA 2D  —  Similarite des modifications contractuelles")
    upd_feats_pca = [
        "Version","Montant_Capital","Taux_Hebdo","Taux_Endettement",
        "Duree_Semaines","Montant_Total","Cout_Credit",
        "DX_Contract_Volatility_Delta","DX_Update_Velocity",
        "DX_Renegotiation_Gravity_Score","DX_Structural_Fragility_Index",
        "EX_Modification_Pattern_Clustering","EX_Version_Drift_Analysis",
    ]
    df_upd_pca = df.copy()
    df_upd_pca["_grav"] = df_upd_pca.get("DX_Renegotiation_Gravity_Score",
        pd.Series(0, index=df_upd_pca.index)).fillna(0)
    raw_drift  = df_upd_pca.get("EX_Version_Drift_Analysis", pd.Series(0, index=df_upd_pca.index)).fillna(0)
    mx_drift   = raw_drift.abs().max()
    df_upd_pca["_drift_n"] = (raw_drift / mx_drift).round(3) if mx_drift > 0 else pd.Series(0.0, index=df_upd_pca.index)
    u_pca1, u_pca2 = st.columns(2)
    with u_pca1:
        fig = _pca_ex(df_upd_pca, upd_feats_pca, "ID_Pret", "_grav",
                      "Gravite reneg.", "UPDATES  (couleur = Gravite Renegociation)")
        st.plotly_chart(fig, use_container_width=True)
    with u_pca2:
        fig = _pca_ex(df_upd_pca, upd_feats_pca, "ID_Pret", "_drift_n",
                      "Derive montant", "UPDATES  (couleur = Derive Montant Total)",
                      colorscale=[[0,"#58a6ff"],[0.5,"#f39c12"],[1,"#c0392b"]])
        st.plotly_chart(fig, use_container_width=True)


def _explore_remboursements(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.info("Table Remboursements vide."); return

    _ctx("SOURCE : Remboursements  |  AXEE : comportement temporel, "
         "correlation montant/retard, fenetres horaires d'encaissement")

    c1, c2 = st.columns(2)

    # 1. Scatter delai vs montant
    with c1:
        sp_c = {"EN_RETARD":_C["critical"],"PONCTUEL":_C["success"],
                "ANTICIPE":_C["accent"],"PAYE":_C["success"],"PARTIEL":_C["warning"]}
        anomaly = df.get("EX_Timestamp_Anomaly_Log", pd.Series(0, index=df.index)).fillna(0)
        fig = go.Figure()
        for sp in df["Statut_Paiement"].dropna().unique():
            mask = df["Statut_Paiement"]==sp; sub = df[mask]; a = anomaly[mask]
            fig.add_trace(go.Scatter(x=sub["Jours_Retard"],y=sub["Montant_Verse"],mode="markers",name=str(sp),
                marker=dict(size=[14 if x else 9 for x in a],
                    color=sp_c.get(str(sp).upper(),_C["neutral"]),
                    symbol=["x" if x else "circle" for x in a],
                    line=dict(width=1.5,color=_C["bg"])),
                text=sub["ID_Pret"].astype(str),
                hovertemplate=f"<b>{sp}</b><br>Retard : %{{x}}j<br>Montant : %{{y:,}}<extra></extra>"))
        fig.add_vline(x=0,line_dash="dot",line_color=_C["neutral"],line_width=1)
        fig.update_layout(**_L(title="DELAI vs MONTANT  (X = heure atypique)",height=310,
                               xaxis_title="Jours Retard",yaxis_title="Montant Verse"))
        st.plotly_chart(fig, use_container_width=True)

    # 2. Heatmap temporelle
    with c2:
        ts = pd.to_datetime(df.get("Timestamp", pd.Series(dtype=str)), errors="coerce")
        df2 = df.copy(); df2["Heure"] = ts.dt.hour
        jmap = {"Monday":"Lundi","Tuesday":"Mardi","Wednesday":"Mercredi",
                "Thursday":"Jeudi","Friday":"Vendredi","Saturday":"Samedi","Sunday":"Dimanche"}
        df2["Jour"] = ts.dt.day_name().map(jmap)
        jord = ["Lundi","Mardi","Mercredi","Jeudi","Vendredi","Samedi","Dimanche"]
        df2v = df2.dropna(subset=["Heure","Jour"])
        if not df2v.empty:
            pivot = df2v.groupby(["Jour","Heure"]).size().reset_index(name="n")
            fig   = go.Figure(go.Heatmap(
                x=pivot["Heure"],y=pivot["Jour"],z=pivot["n"],
                colorscale=[[0,_C["card"]],[0.4,"#1a3a5c"],[1,_C["ex"]]],
                text=pivot["n"],texttemplate="%{text}",
                hovertemplate="<b>%{y}</b>  %{x}h00 : %{z} paiements<extra></extra>",
                colorbar=dict(title="Paiements",tickfont=dict(size=9,family=_FONT)),
                xgap=2,ygap=2))
            fig.update_layout(**_L(title="FENETRES D'ENCAISSEMENT  (jour x heure)",height=310,
                xaxis=dict(title="Heure",dtick=2),
                yaxis=dict(categoryorder="array",categoryarray=jord)))
            st.plotly_chart(fig, use_container_width=True)
        else:
            st.plotly_chart(_ef("Timestamps non parsables.",310), use_container_width=True)

    c3, c4 = st.columns(2)

    # 3. Bar propension paiement anticipe par client
    with c3:
        if "ID_Client" in df.columns and "EX_Early_Payment_Propensity" in df.columns:
            prop = df.groupby("ID_Client")["EX_Early_Payment_Propensity"].mean().sort_values(ascending=False)
            fig  = go.Figure(go.Bar(x=prop.index.astype(str),y=prop.values,
                marker_color=[_C["success"] if v>0.5 else _C["warning"] if v>0 else _C["critical"] for v in prop.values],
                text=[f"{v:.0%}" for v in prop.values],textposition="outside"))
            fig.update_layout(**_L(title="PROPENSION PAIEMENT ANTICIPE PAR CLIENT",
                                   height=280,showlegend=False,yaxis=dict(tickformat=".0%")))
            st.plotly_chart(fig, use_container_width=True)

    # 4. Scatter montant vs solde avant
    with c4:
        if "Solde_Avant" in df.columns:
            fig = go.Figure()
            sp_c2 = {"EN_RETARD":_C["critical"],"PONCTUEL":_C["success"],"ANTICIPE":_C["accent"]}
            for sp in df["Statut_Paiement"].dropna().unique():
                mask = df["Statut_Paiement"]==sp; sub = df[mask]
                fig.add_trace(go.Scatter(x=sub["Solde_Avant"],y=sub["Montant_Verse"],mode="markers",name=str(sp),
                    marker=dict(size=10,color=sp_c2.get(str(sp).upper(),_C["neutral"]),line=dict(width=1,color=_C["bg"])),
                    hovertemplate=f"<b>{sp}</b><br>Solde : %{{x:,}}<br>Verse : %{{y:,}}<extra></extra>"))
            mx = max(df["Solde_Avant"].max(),df["Montant_Verse"].max())*1.05
            fig.add_trace(go.Scatter(x=[0,mx],y=[0,mx],mode="lines",
                line=dict(color=_C["subtext"],dash="dot",width=1),showlegend=False,hoverinfo="skip"))
            fig.update_layout(**_L(title="MONTANT VERSE vs SOLDE INITIAL",height=280,
                                   xaxis_title="Solde Avant",yaxis_title="Montant Verse"))
            st.plotly_chart(fig, use_container_width=True)

    c5, c6 = st.columns(2)

    # 5. Bar coherence du paiement par client
    with c5:
        if "ID_Client" in df.columns and "EX_Payment_Consistency_Score" in df.columns:
            cons = df.groupby("ID_Client")["EX_Payment_Consistency_Score"].mean().sort_values()
            fig  = go.Figure(go.Bar(x=cons.index.astype(str),y=cons.values,
                marker_color=[_C["critical"] if v>5000 else _C["warning"] if v>2000 else _C["success"] for v in cons.values],
                text=cons.round(0).astype(int).tolist(),textposition="outside"))
            fig.update_layout(**_L(title="ECART-TYPE MONTANT VERSE PAR CLIENT  (coherence)",
                                   height=280,showlegend=False,yaxis_title="Ecart-type"))
            st.plotly_chart(fig, use_container_width=True)

    # 6. Timeline chronologique
    with c6:
        d2 = df.copy(); d2["Date_Paiement"] = pd.to_datetime(d2.get("Date_Paiement"),errors="coerce")
        d2 = d2.dropna(subset=["Date_Paiement"])
        if not d2.empty:
            sp_c3 = {"EN_RETARD":_C["critical"],"PONCTUEL":_C["success"],"ANTICIPE":_C["accent"]}
            fig = go.Figure()
            for sp in d2["Statut_Paiement"].dropna().unique():
                mask = d2["Statut_Paiement"]==sp; sub = d2[mask]
                fig.add_trace(go.Scatter(x=sub["Date_Paiement"],y=sub["Montant_Verse"],mode="markers",name=str(sp),
                    marker=dict(size=10,color=sp_c3.get(str(sp).upper(),_C["neutral"]),line=dict(width=1.5,color=_C["bg"])),
                    text=sub["ID_Pret"].astype(str),
                    hovertemplate="<b>%{text}</b><br>%{x|%d/%m/%Y}<br>%{y:,}<extra></extra>"))
            fig.update_layout(**_L(title="TIMELINE CHRONOLOGIQUE DES PAIEMENTS",height=280))
            st.plotly_chart(fig, use_container_width=True)

    # 7. PCA 2D Remboursements
    st.divider()
    _ctx("CLUSTER MAP  PCA 2D  —  Comportements de paiement regroupes")
    rem_feats_pca = [
        "Montant_Verse","Montant_Principal","Montant_Interets","Penalites_Retard",
        "Solde_Avant","Solde_Apres","Jours_Retard",
        "DX_Profitability_Realization_Rate","DX_Overpayment_Tendency",
        "DX_Channel_Risk_Correlation","DX_Temporal_Payment_Behavior",
        "EX_Early_Payment_Propensity","EX_Payment_Consistency_Score",
        "EX_Collection_Cost_Efficiency",
    ]
    df_rem_pca = df.copy()
    # Score risque remboursement : retard positif normalise
    jr = df_rem_pca["Jours_Retard"].clip(lower=0)
    jr_max = jr.max()
    df_rem_pca["_rem_risk"] = (jr / jr_max).round(3) if jr_max > 0 else pd.Series(0.0, index=df_rem_pca.index)
    r_pca1, r_pca2 = st.columns(2)
    with r_pca1:
        fig = _pca_ex(df_rem_pca, rem_feats_pca, "ID_Pret", "_rem_risk",
                      "Risque retard", "REMBOURSEMENTS  (couleur = Jours Retard norm.)")
        st.plotly_chart(fig, use_container_width=True)
    with r_pca2:
        fig = _pca_ex(df_rem_pca, rem_feats_pca, "ID_Pret", "Montant_Verse",
                      "Montant", "REMBOURSEMENTS  (couleur = Montant Verse)",
                      colorscale=[[0,"#c0392b"],[0.5,"#f39c12"],[1,"#2ecc71"]])
        st.plotly_chart(fig, use_container_width=True)


def _explore_ajustements(df: pd.DataFrame, sheets: dict):
    if df.empty:
        st.markdown(
            f'<div style="background:{_C["card"]};border:1px solid {_C["border"]};'
            f'border-left:3px solid {_C["neutral"]};padding:16px 20px;font-family:{_FONT};">'
            f'<div style="font-size:0.50rem;letter-spacing:2px;color:{_C["subtext"]};'
            f'text-transform:uppercase;margin-bottom:6px;">STATUS OPERATIONNEL</div>'
            f'<div style="font-size:1rem;color:{_C["subtext"]};">TABLE VIDE  EN ATTENTE D\'ALIMENTATION</div>'
            f'<div style="font-size:0.58rem;color:{_C["subtext"]};margin-top:4px;">'
            f'Analyses causales actives des le premier ajustement enregistre.</div>'
            f'</div>', unsafe_allow_html=True)
        return

    _ctx("SOURCE : Ajustements_Echeances + Prets_Master  |  AXEE : "
         "frequence, impact financier, correlation defauts futurs")

    c1, c2 = st.columns(2)
    with c1:
        cnt = df["Raison"].value_counts()
        r2d = df.groupby("Raison")["EX_Reason_to_Default_Bridge"].mean() if "EX_Reason_to_Default_Bridge" in df.columns else pd.Series(0,index=cnt.index)
        bc  = [_C["critical"] if r2d.get(r,0)>0.3 else _C["warning"] if r2d.get(r,0)>0.1 else _C["accent"] for r in cnt.index]
        fig = go.Figure(go.Bar(x=cnt.values,y=cnt.index.astype(str),orientation="h",
            marker_color=bc,text=cnt.values,textposition="outside"))
        fig.update_layout(**_L(title="RAISONS  (couleur = pont defaut)",height=300,showlegend=False))
        st.plotly_chart(fig, use_container_width=True)
    with c2:
        if "EX_Timing_of_Adjustment" in df.columns and "EX_Amount_Expansion_Factor" in df.columns:
            fig = go.Figure(go.Scatter(
                x=df["EX_Timing_of_Adjustment"].fillna(0),
                y=df["EX_Amount_Expansion_Factor"].fillna(0),
                mode="markers",
                marker=dict(size=12,color=_C["ex"],line=dict(width=1.5,color=_C["bg"])),
                text=df["ID_Pret"].astype(str) if "ID_Pret" in df.columns else df.index.astype(str),
                hovertemplate="<b>%{text}</b><br>Timing : %{x}j<br>Expansion : %{y:.2%}<extra></extra>"))
            fig.update_layout(**_L(title="TIMING vs EXPANSION MONTANT",height=300,showlegend=False,
                                   xaxis_title="Jours avant fin",yaxis_title="Facteur expansion"))
            st.plotly_chart(fig, use_container_width=True)


# ── Routeur public ────────────────────────────────────────────────────────────

_MAP = {
    "CLIENTS":        ("Clients_KYC",           _explore_clients),
    "GARANTS":        ("Garants_KYC",           _explore_garants),
    "PRETS":          ("Prets_Master",          _explore_prets),
    "PRETS_UPDATE":   ("Prets_Update",          _explore_prets_update),
    "PRETS UPDATE":   ("Prets_Update",          _explore_prets_update),
    "REMBOURSEMENTS": ("Remboursements",        _explore_remboursements),
    "AJUSTEMENTS":    ("Ajustements_Echeances", _explore_ajustements),
}

def render_exploration(sheets: dict, section: str):
    section_up = section.upper().replace("_"," ").strip()
    for key in [section, section_up, section.upper()]:
        if key in _MAP:
            tab_key, fn = _MAP[key]
            fn(sheets.get(tab_key, pd.DataFrame()), sheets)
            return
    st.warning(f"Section inconnue : {section}")