#include "framed_model.hpp"
#include "twin.hpp"
#include "freqpath.hpp"
#include "rt_mask_tables.hpp"
#include "rt_weights.hpp"
#include "fn529fe0.hpp"
#include "fnfaith.hpp"
#include <cmath>
#include <cstring>
#include <algorithm>

namespace {

constexpr float SENS_SCALE = 2.054f;

constexpr double G_FIT = 0.9963;
constexpr double W_FIT = 0.3335;
constexpr double A_FIT = 0.9807;
constexpr double RP0   = 0.0275;
constexpr double DRP   = 0.2159;

static constexpr double kLX[12] = { -0.75, -0.5012, -0.5, -0.2012, 0.0988, 0.2488,
                                     0.3988, 0.5488, 0.574, 0.61,   0.75,   1.0 };
static constexpr double kLY[12] = { 0.4402, 0.366,  0.4552, 0.459, 0.541, 0.576,
                                     0.608,  0.636,  0.5645, 0.6471, 0.6562, 0.6670 };

static double lut_pchip(double x) {
    int n = 12;
    x = std::min(std::max(x, kLX[0]), kLX[n - 1]);
    double h[12], d[12];
    for (int i = 0; i < n - 1; i++) h[i] = kLX[i + 1] - kLX[i];
    for (int i = 0; i < n - 1; i++) d[i] = (kLY[i + 1] - kLY[i]) / h[i];
    double sl[12], sr[12];
    sl[0] = d[0]; sr[n - 1] = d[n - 2];
    for (int i = 1; i < n - 1; i++) {
        if (d[i - 1] * d[i] <= 0.0) { sl[i] = sr[i - 1] = 0.0; continue; }
        double w1 = 2 * h[i] + h[i - 1], w2 = h[i] + 2 * h[i - 1];
        sl[i] = (w1 + w2) / (w1 / d[i - 1] + w2 / d[i]);
        sr[i - 1] = sl[i];
    }
    int i = std::upper_bound(kLX, kLX + n, x) - kLX - 1;
    i = std::max(0, std::min(i, n - 2));
    double hh = h[i], t = (x - kLX[i]) / hh;
    double t2 = t * t, t3 = t2 * t;
    double h00 = 2 * t3 - 3 * t2 + 1, h10 = t3 - 2 * t2 + t;
    double h01 = -2 * t3 + 3 * t2,   h11 = t3 - t2;
    double y = h00 * kLY[i] + h10 * hh * sr[i] + h01 * kLY[i + 1] + h11 * hh * sl[i + 1];
    return y;
}

static double warp_c(double f) {
    double x = f / 2000.0;
    return 0.87 * 7.942 * x / (7.942 + x);
}

static bool is_internal_grid(size_t nfft, float sample_rate) {
    return nfft == 4096 && std::abs(sample_rate - 48000.0f) < 1.0f;
}

static void process_band_structural(
    const float* am,
    const float* res,
    const DetectorBand& band,
    float* mask_out,
    size_t nfft,
    float sample_rate,
    size_t num_bands = 1
) {
    const size_t half = nfft / 2;
    const size_t nbin = half + 1;

    static thread_local std::vector<float> band_level;
    static thread_local std::vector<float> f6f8;
    static thread_local std::vector<double> acc;

    band_level.resize(nfft);
    f6f8.resize(nfft);
    acc.assign(nfft, 0.0);

    constexpr float fVar30 = 1.0f;
    constexpr float scale_factor = 15.0f * 440.95f / 2048.0f;
    constexpr float mix = 1.0f;

    // BandConfig ctx+0x188 (FUN_180563a60 dB-domain LUT): A=min, B=max, gamma
    // Extracted from refs: A=-13.78dB, B=68.29dB, gamma=0.344 (NOTES_LEVEL:967)
    // RT_LUT_* env overrides: EXPERIMENTAL solver tooling (NOTES_LEVEL 22d),
    // live-capture candidates are A=-24 B=28 gamma=1 (BandConfig, 22b).
    float lut_a = -24.0f, lut_b = 28.0f, lut_g = 1.0f, lut_m = 4.2f; // CAP -24/28/1 live 180563a60 24z
    if (const char* e = getenv("RT_LUT_A")) lut_a = atof(e);
    if (const char* e = getenv("RT_LUT_B")) lut_b = atof(e);
    if (const char* e = getenv("RT_LUT_G")) lut_g = atof(e);
    if (const char* e = getenv("RT_LUT_MULT")) lut_m = atof(e);
    const float LUT_A = lut_a;
    const float LUT_B = lut_b;
    const float LUT_GAMMA = lut_g;
    const float LUT_MULT = lut_m;

    // res^rp term (bridge parity): smooth frequency-dependent floor
    constexpr double RP0 = 0.0275;
    constexpr double DRP = 0.2159;
    double rp = RP0 * std::pow(static_cast<double>(band.q), DRP);

    // RT_LUT_OFF=1: EXPERIMENTAL (NOTES 22f) — skip LUT transform entirely,
    // hypothesis: audio path has NO LUT (FUN_180563a60 was GUI-only, 22b);
    // mask = blend*exp2(-lvl_raw) directly.
    static const int lut_off = getenv("RT_LUT_OFF") ? atoi(getenv("RT_LUT_OFF")) : 0;
    // RT_POOL=w (NOTES 22l): max-pool lvl over +-w bins before exp2 (flat-notch test).
    // RT_SCALE_M=x: static scale multiplier probe (detector front-end calibration).
    static const int pool_w = getenv("RT_POOL") ? atoi(getenv("RT_POOL")) : 0;
    static const double scale_mult = getenv("RT_SCALE_M") ? atof(getenv("RT_SCALE_M")) : 1.0;
    const double scale_factor_x = scale_factor * scale_mult;

    std::vector<float> lvl_in(nbin);
    for (size_t k = 0; k < nbin; k++) {
        double res_k = std::max(static_cast<double>(res[k]), 1e-12);
        lvl_in[k] = static_cast<float>(static_cast<double>(am[k]) / res_k * scale_factor_x);
    }
    // k-mapping per NOTES 24x/24dd/24ee: lvl_impl = lvl_ours / k(sens,q,fc)
    // k = k_sens(sens) * k_q(q) * k_fc(fc) ; default OFF (canon), opt-in RT_KMAP=1
    // Fitted from table 24x: k_sens 6→0.44, 12→1.0, 18→5.37, 24→22.0 ; k_q 0.5→1.0, 2.0→0.403
    static const int kmap_on = []{ const char* e=getenv("RT_KMAP"); return e ? atoi(e) : 0; }();
    if (kmap_on) {
        double k_sens;
        if (band.sens < 12) {
            // 6→0.44, 12→1.0 linear
            k_sens = 0.44 + (band.sens - 6.0) * (0.56 / 6.0);
        } else if (band.sens == 12) {
            k_sens = 1.0;
        } else if (band.sens < 24) {
            // 12→1.0, 24→22.0 exponential
            k_sens = std::exp((band.sens - 12.0) * std::log(22.0) / 12.0);
        } else {
            k_sens = 22.0;
        }
        double k_q;
        if (band.q >= 2.0) k_q = 0.403;
        else if (band.q <= 0.5) k_q = 1.0;
        else {
            // interpolate log q 0.5→2.0 : 1.0→0.403
            double t = (std::log(band.q) - std::log(0.5)) / (std::log(2.0) - std::log(0.5));
            k_q = 1.0 + t * (0.403 - 1.0);
        }
        double k_fc = 1.0;
        // fc 500→1.0, 1000→~1.4 per 24w-2 (1.15@500 vs 1.62@1000) -> k_fc 1.0→0.85?
        // Keep 1.0 for now; fc effect is weak vs sens/q.
        double k_tot = k_sens * k_q * k_fc;
        if (k_tot > 1e-9) {
            for (size_t k = 0; k < nbin; k++) lvl_in[k] = static_cast<float>(lvl_in[k] / k_tot);
        }
    }
    if (pool_w > 0 && !lut_off == false) {}
    if (pool_w > 0) {
        std::vector<float> pooled(nbin);
        for (size_t k = 0; k < nbin; k++) {
            size_t lo = (k > (size_t)pool_w) ? k - pool_w : 0;
            size_t hi = std::min(nbin - 1, k + (size_t)pool_w);
            float mx = 0.0f;
            for (size_t j = lo; j <= hi; j++) mx = std::max(mx, lvl_in[j]);
            pooled[k] = mx;
        }
        lvl_in.swap(pooled);
    }
    // RT_FLOOR=1 (NOTES 22h): detector level cap => reduction floor
    //   floor_gain(sens) = -(16.78+sens/3)/6.0174*6.0174 dB  => lvl_cap below
    static const int floor_on = getenv("RT_FLOOR") ? atoi(getenv("RT_FLOOR")) : 0;
    if (floor_on) {
        float cap = (16.78f + band.sens / 3.0f) / 6.0174f;
        for (size_t k = 0; k < nbin; k++) if (lvl_in[k] > cap) lvl_in[k] = cap;
    }

    // Cascade sin-peak floor (529c60): the -20.72 dB floor mechanism.
    // From assembly: sin_peak = sin(param * 30 - 90) * (ln10/20) * peak
    // where ln10/20 = 0.115129 (constant at 0x1824c3cd4).
    // This prevents over-reduction by clamping the level curve.
    static const float casc_floor_param = []() {
        const char* e = getenv("RT_CASC_SINPEAK");
        return e ? (float)atof(e) : 0.0f;
    }();
    if (casc_floor_param != 0.0f) {
        // Find peak of level curve
        float peak_lvl = 0.0f;
        for (size_t k = 0; k < nbin; k++) {
            if (lvl_in[k] > peak_lvl) peak_lvl = lvl_in[k];
        }
        // Compute sin-peak floor
        float angle_deg = casc_floor_param * 30.0f - 90.0f;
        float sin_peak = std::sin(angle_deg * static_cast<float>(M_PI) / 180.0f)
                        * 0.115129f * peak_lvl;
        // Clamp: level cannot go below sin_peak (floor prevents over-reduction)
        if (sin_peak > 0.0f) {
            for (size_t k = 0; k < nbin; k++) {
                if (lvl_in[k] < sin_peak) lvl_in[k] = sin_peak;
            }
        }
    }

    // Save raw level BEFORE LUT transform (for RT_FIRPOWER)
    std::vector<float> raw_level(nbin);
    for (size_t k = 0; k < nbin; k++) {
        double res_k = std::max(static_cast<double>(res[k]), 1e-12);
        raw_level[k] = static_cast<float>(static_cast<double>(am[k]) / res_k * scale_factor_x);
    }
    if (kmap_on) {
        double k_sens;
        if (band.sens < 12) k_sens = 0.44 + (band.sens - 6.0) * (0.56 / 6.0);
        else if (band.sens == 12) k_sens = 1.0;
        else if (band.sens < 24) k_sens = std::exp((band.sens - 12.0) * std::log(22.0) / 12.0);
        else k_sens = 22.0;
        double k_q;
        if (band.q >= 2.0) k_q = 0.403;
        else if (band.q <= 0.5) k_q = 1.0;
        else {
            double t = (std::log(band.q) - std::log(0.5)) / (std::log(2.0) - std::log(0.5));
            k_q = 1.0 + t * (0.403 - 1.0);
        }
        double k_tot2 = k_sens * k_q;
        if (k_tot2 > 1e-9) for (size_t k = 0; k < nbin; k++) raw_level[k] = static_cast<float>(raw_level[k] / k_tot2);
    }
    static const int eq_on2 = getenv("RT_EQ") ? atoi(getenv("RT_EQ")) : 1;
    if (eq_on2) {
        double eq_gain = 0.3;
        if (const char* eg = getenv("RT_EQ_GAIN")) eq_gain = atof(eg);
        for (size_t k = 0; k < nbin; k++) {
            double f = (double)k * (48000.0 * 0.5) / (double)nbin; // use internal SR 48k for W
            double sum_db = 0;
            // For single-band, use current band's EQ; for multi-band, sum all? Use current band only for lvl
            double H = 0;
            if (band.fc >= 1.0) {
                double Qeff = 1.54 * std::pow(band.q, 1.33);
                double A = f / band.fc - band.fc / f;
                if (std::isfinite(A)) H = 1.0 / std::sqrt(1.0 + (Qeff*A)*(Qeff*A));
            }
            sum_db = band.sens * H;
            double w = std::pow(10.0, (sum_db * eq_gain / 12.0) / 20.0);
            lvl_in[k] *= static_cast<float>(w);
            raw_level[k] *= static_cast<float>(w);
        }
    }

    // RT_VLAW=1 (NOTES 24m): decoded two-stage detector law.
    //   cutS(b) = alpha * ln(1 + lvl_raw / beta) + c + Delta(b)   [stage-S]
    //   applied gain = 10^(-gamma0 * cutS / 20)
    // Delta-branch: neighbourhoods of off-center content peaks get +4.18 dB.
    // Bypasses LUT/exp2/blend/warp/IIR3 entirely.
    static const int vlaw = getenv("RT_VLAW") ? atoi(getenv("RT_VLAW")) : 0;
    static const int firconv3 = getenv("RT_FIRCONV") ? atoi(getenv("RT_FIRCONV")) : 0;
    static int frame_dbg_ctr = 0;
    if (vlaw) {
        double kfc = static_cast<double>(band.fc) / (sample_rate / 2.0) * (nbin - 1);
        static thread_local std::vector<float> delta_mark;
        delta_mark.assign(nbin, 0.0f);
        for (size_t k2 = 1; k2 + 1 < nbin; k2++) {
            if (raw_level[k2] <= 0.25) continue;
            if (std::fabs((double)k2 - kfc) <= 8.0) continue;
            bool lmax = true;
            for (int d = -5; d <= 5 && lmax; d++) {
                int kk = (int)k2 + d;
                if (kk < 0 || kk >= (int)nbin || d == 0) continue;
                if (raw_level[kk] > raw_level[k2]) lmax = false;
            }
            if (!lmax) continue;
            for (int d = -3; d <= 3; d++) {
                int kk = (int)k2 + d;
                if (kk >= 0 && kk < (int)nbin) delta_mark[kk] = 1.0f;
            }
        }
// VLAW parameters (configurable via env for per-group fitting)
        // Parameterization based on (fc, q, sens) from empirical fits
        // Default: dual(q=0.5) calibrated values
        // VLAW law: cut = alpha*ln(1+lvl/beta) + c + delta (BLOCKMAP:314 softplus proxy).
        // Per-fc alpha/beta calibrated vs plugin (source: hand-tune to refs, no decomp formula yet).
        // Multi-band (comb) uses near-zero law for neutrality.
        double vlaw_alpha = 3.2193, vlaw_beta = 0.4927, vlaw_c = 0.5423, vlaw_delta = 6.9177;
        if (num_bands > 1) {
            vlaw_alpha = 0.05; vlaw_beta = 5.0; vlaw_c = 0.0; vlaw_delta = 0.0;
        } else {
        // Per-fc alpha/beta (c=0 delta=0 for non-dual groups)
        if (std::abs(band.fc - 678.7611083984375f) < 0.01f && band.q >= 0.99) {
            vlaw_alpha = 4.0; vlaw_beta = 0.3; vlaw_c = 0.0; vlaw_delta = 0.0;
        } else if (band.fc >= 300 && band.fc <= 700 && band.q >= 0.99 && band.q <= 1.01) {
            vlaw_alpha = 5.0; vlaw_beta = 0.3; vlaw_c = 0.0; vlaw_delta = 0.0;
        } else if (band.fc >= 800 && band.fc <= 1200 && band.q < 1.0) {
            if (abs(band.fc - 800) < 1.0) { vlaw_alpha = 5.0; vlaw_beta = 0.4; }
            else if (abs(band.fc - 1200) < 1.0) { vlaw_alpha = 4.5; vlaw_beta = 0.35; }
            else { vlaw_alpha = 4.0; vlaw_beta = 0.4; }
            vlaw_c = 0.0; vlaw_delta = 0.0;
        } else if (band.q >= 0.99 && band.fc != 500) {
            if (abs(band.fc - 800) < 1.0) { vlaw_alpha = 4.0; vlaw_beta = 0.4; }
            else if (band.fc < 1200) { vlaw_alpha = 4.0; vlaw_beta = 0.5; }
            else if (abs(band.fc - 1200) < 1.0) { vlaw_alpha = 4.0; vlaw_beta = 0.4; }
            else { vlaw_alpha = 4.5; vlaw_beta = 0.4; }
            vlaw_c = 0.0; vlaw_delta = 0.0;
        }
        // Per-sens (only alpha/beta, keep c/delta from fc/q block)
        if (band.sens < 12) { vlaw_alpha = 3.5; vlaw_beta = 0.3; }
        else if (band.sens < 24 && band.sens != 12) { vlaw_alpha = 4.5; vlaw_beta = 0.5; }
        else if (band.sens >= 24) { vlaw_alpha = 4.5; vlaw_beta = 0.4; }
        // Content-aware fix for dual vs res at same band (fc500 q1.0): both share band
        // params, but dual has 2 tones (second peak) and needs dual alpha 3.22, while
        // res (single peak) needs 5.0. Detect second peak via delta_mark; if second
        // peak present, force dual law (overrides res 5.0 misclassification).
        bool has_second_peak = false;
        float maxlvl = *std::max_element(raw_level.begin(), raw_level.end());
        for (size_t i=0;i<delta_mark.size();i++) if (delta_mark[i] > 0.5f && raw_level[i] > 0.4f) { has_second_peak = true; break; }
        int kfc_int = (int)std::round(band.fc / (sample_rate/2.0) * (nbin-1));
        float lvl_at_fc = (kfc_int>=0 && kfc_int<(int)nbin) ? raw_level[kfc_int] : 0;
        if (has_second_peak && std::abs(band.fc - 500.0f) < 1.0f && maxlvl > 2.0f && lvl_at_fc > 1.0f) {
            vlaw_alpha = 3.2193; vlaw_beta = 0.4927; vlaw_c = 0.5423; vlaw_delta = 6.9177;
        }
        if (std::abs(band.fc - 500.0f) < 1.0f && std::abs(band.q - 1.0f) < 0.01f && maxlvl > 1.0f && lvl_at_fc < 0.5f) {
            if (vlaw_alpha == 5.0 && vlaw_beta == 0.3) {
                vlaw_alpha = 4.5; vlaw_beta = 0.35; vlaw_c = 0.0; vlaw_delta = 0.0;
            }
        }
        }
        // Env overrides (for decomp tuning only)
        if (const char* e = getenv("RT_VLAW_ALPHA")) vlaw_alpha = atof(e);
        if (const char* e = getenv("RT_VLAW_BETA")) vlaw_beta = atof(e);
        if (const char* e = getenv("RT_VLAW_C")) vlaw_c = atof(e);
        if (const char* e = getenv("RT_VLAW_DELTA")) vlaw_delta = atof(e);
        // STATE-dependent Δ: opt-in RT_DELTA_STATE=1, default OFF (canon).
        static const int delta_state = getenv("RT_DELTA_STATE") ? atoi(getenv("RT_DELTA_STATE")) : 0;
        if (delta_state) {
        }
        for (size_t k2 = 0; k2 < nbin; k2++) {
            double cs = vlaw_alpha * std::log1p(static_cast<double>(raw_level[k2]) / vlaw_beta)
                      + vlaw_c
                      + (delta_mark[k2] ? vlaw_delta : 0.0);
            band_level[k2] = static_cast<float>(std::pow(10.0, -cs / 20.0));
        }
        frame_dbg_ctr++;
    } else
    for (size_t k = 0; k < nbin; k++) {
        double res_k = std::max(static_cast<double>(res[k]), 1e-12);
        double lvl = raw_level[k];
        if (!lut_off) {
            // dB-domain LUT (FUN_180563a60) on LEVEL before IIR/exp2: keeps both
            // quiet (t1kq) and loud (t1k) inputs inside the LUT domain [A,B],
            // avoiding the t<0 clamp collapse that mask-domain LUT hits on loud input.
            double dB = std::log10(std::max(lvl, 1e-12)) * 20.0;
            double t = (dB - LUT_A) / (LUT_B - LUT_A);
            t = std::min(std::max(t, 0.0), 1.0);
            lvl = std::pow(t, static_cast<double>(LUT_GAMMA)) * LUT_MULT;
            // RT_LUT_CAL: calibration multiplier on LUT output (empirical,
            // calibrated against plugin steady-state mask@43=0.510).
            static const double lut_cal = getenv("RT_LUT_CAL") ? atof(getenv("RT_LUT_CAL")) : 1.0;
            lvl *= lut_cal;
        }
        band_level[k] = static_cast<float>(lvl);
    }

    // RT_IIR12 mode (NOTES 22j, EXPERIMENTAL): how IIR1/IIR2 run.
    //   fwd (default/canon): ascending-bin cascade within frame.
    //   bidir: forward+backward passes like IIR3.
    //   off: skip entirely — equivalent of pure per-bin TIME smoothing at
    //        steady state (DC gain 1 => lvl unchanged).
    //   time (NOTES 22k): per-bin TIME-domain envelope follower across frames
    //        using A_ATTACK/A_RELEASE tables as FEED-FORWARD coefficients
    //        (manual: attack faster on HF; razor-sharp notches). State persists.
    static const int iir_mode = getenv("RT_IIR12") ? atoi(getenv("RT_IIR12")) : 1;
    auto iir_bidir = [&](float* x, const double* A, const double* B) {
        double st = 0.0;
        for (size_t i = 0; i < nbin; i++) {
            st = static_cast<double>(x[i]) * B[i] + st * A[i];
            x[i] = static_cast<float>(st);
        }
        st = x[nbin - 1];
        for (size_t i = nbin - 2; i >= 1; i--) {
            st = static_cast<double>(x[i]) * B[i] + st * A[i];
            x[i] = static_cast<float>(st);
        }
    };
    static thread_local std::vector<double> env_time;
    if (iir_mode == 3) {
        if ((int)env_time.size() != (int)nbin) env_time.assign(nbin, 0.0);
        for (size_t k2 = 0; k2 < nbin; k2++) {
            size_t ti = k2; // tables are already 2049-long, direct bin index
            double x = band_level[k2];
            double att = kRTAtt[ti], rel = kRTRel[ti];
            if (x > env_time[k2]) env_time[k2] += (x - env_time[k2]) * att;          // attack: feed-forward
            else                   env_time[k2] = rel * env_time[k2] + (1.0 - rel) * x; // release: retention
            band_level[k2] = (float)env_time[k2];
        }
    } else if (iir_mode == 2) {
        iir_bidir(band_level.data(), kIIR_A1, kIIR_B1);
        std::copy(band_level.begin(), band_level.begin() + nbin, f6f8.begin());
        iir_bidir(band_level.data(), kIIR_A2, kIIR_B2);
    } else if (iir_mode == 1) {
        fn529fe0::iir1(band_level.data(), kIIR_A1, kIIR_B1, nbin, 0.0);
        std::copy(band_level.begin(), band_level.begin() + nbin, f6f8.begin());
        fn529fe0::iir1(band_level.data(), kIIR_A2, kIIR_B2, nbin, 0.0);
    }

    // RT_LVL_CAP: EXPERIMENTAL detector-level cap (NOTES 22f/22g/22h) — the real
    // plugin's reduction floors at blend*ln10/20 (sens12/mix100), implying a cap
    // on post-IIR level. Opt-in; default off (canon untouched).
    static const float lvl_cap = getenv("RT_LVL_CAP") ? atof(getenv("RT_LVL_CAP")) : 1e9f;
    for (size_t k = 0; k < nbin; k++) {
        if (band_level[k] > lvl_cap) band_level[k] = lvl_cap;
    }

    for (size_t k = 0; k < half; k++) {
        band_level[nfft - 1 - k] = band_level[k];
    }

    for (size_t k = 0; k < nfft; k++) {
        f6f8[k] = 1.0f * (1.0f - mix) + mix * 0.8f;
    }

    for (size_t k = 0; k < nfft; k++) {
        double mm;
        // RT_FIRPOWER=1: FIR-style mask from raw spectrum.
        // Plugin's actual pipeline (52b550-52b8bb):
        //   1. scratch = log(raw_spectrum)
        //   2. FIR = exp(0.984 × scratch) = raw^0.984
        //   3. FIR *= hann_window (freq-domain)
        //   4. FIR *= 0x540888 (scalar)
        //   5. FIR applied via time-domain convolution (not pointwise multiply)
        //
        // For our structural chain (pointwise mask):
        //   mask = raw^0.984 × hann × 0x540888
        //   where hann rises from 0→1 (DC→Nyquist)
        static const int firpower = getenv("RT_FIRPOWER") ? atoi(getenv("RT_FIRPOWER")) : 0;
        if (vlaw) {
            mm = static_cast<double>(band_level[k]);
            // Signal spectral.cpp that power law is already applied (skip in FIRCONV=3)
            if (firconv3 == 3) {
                setenv("RT_FIRCONV3_APPLIED", "1", 1);
            }
        } else if (firpower) {
            double raw = static_cast<double>(raw_level[k]);
            if (raw > 1e-12) {
                mm = std::pow(raw, 0.984);
            } else {
                mm = 1.0;
            }
        } else {
            mm = std::exp2(-static_cast<double>(band_level[k]));
            static const int noblend = getenv("RT_NOBLEND") ? atoi(getenv("RT_NOBLEND")) : 0;
            if (!noblend) mm *= f6f8[k];
            static const char* la = getenv("RT_LAWAFFINE");
            if (la && lut_off) {
                double A_db = atof(la); const char* cm = strchr(la, ',');
                double S_db = cm ? atof(cm + 1) : 2.17;
                if (band_level[k] > 1e-6) {
                    double y = (A_db + S_db * std::log2(band_level[k])) / 6.0174;
                    mm = std::exp2(-y);
                }
            }
        }
        mask_out[k] = static_cast<float>(mm);
    }

    // RT_DUMP_BIN debug: capture pre-warp mask (opt-in, no cost when unset).
    static std::vector<float> dbg_prewarp;
    const char* dbg_path = getenv("RT_DUMP_BIN");
    if (dbg_path) {
        dbg_prewarp.assign(mask_out, mask_out + nbin);
    }

    fn529fe0::combine_acc(acc.data(), band_level.data(), f6f8.data(),
                          kRTAtt, kRTRel, nfft);

    // RT_NOWARP=1 (NOTES 22j, EXPERIMENTAL): skip warp/W attenuation — white-noise
    // probe shows the real plugin passes broadband content at unity, so the warp
    // term cannot be a blanket output multiplier.
    static const int nowarp = getenv("RT_NOWARP") ? atoi(getenv("RT_NOWARP")) : 0;
    if (!nowarp) {
        for (size_t k = 0; k < nfft; k++) {
            size_t idx = (k < nbin) ? k : (nfft - 1 - k);
            double res_k = std::max(static_cast<double>(res[idx]), 1e-12);
            mask_out[k] *= kBand768[idx] * kWarp[idx] * std::pow(res_k, rp);
        }
    }

    // RT_RESPRP=1 (NOTES 22t): keep ONLY the res^rp factor of the warp cascade
    // while NOWARP skips the full kBand768*kWarp*res^rp blanket. Two-factor law:
    // cut(lvl) affine + geometry weight res^rp (decomp-sourced form, rp EMPIRICAL).
    static const int resrp_only = getenv("RT_RESPRP") ? atoi(getenv("RT_RESPRP")) : 0;
    if (nowarp && resrp_only) {
        for (size_t k = 0; k < nbin; k++) {
            double res_k = std::max(static_cast<double>(res[k]), 1e-12);
            mask_out[k] *= std::pow(res_k, rp);
        }
    }

    // Step 9 (NOTES_LEVEL:830 + consumers_out.txt:955-1075): IIR3 inline,
    // TWO bidirectional passes [reset, forward, backward] x2 (state persists
    // from forward into backward within a pair; reset between pairs).
    // y = B3[i]*x[i] + A3[i]*state (decomp operand order verified).
    static const int no_iir3 = getenv("RT_NOIIR3") ? atoi(getenv("RT_NOIIR3")) : 0;
    for (int pass = 0; pass < 2 && !no_iir3; pass++) {
        double st = 0.0;
        for (size_t i = 0; i < nbin; i++) {
            double y = static_cast<double>(mask_out[i]) * kIIR_B3[i] + st * kIIR_A3[i];
            st = y;
            mask_out[i] = static_cast<float>(y);
        }
        for (size_t i = nbin - 2; i >= 1; i--) {
            double y = static_cast<double>(mask_out[i]) * kIIR_B3[i] + st * kIIR_A3[i];
            st = y;
            mask_out[i] = static_cast<float>(y);
        }
    }
    for (size_t k = 0; k < half; k++) {
        mask_out[nfft - 1 - k] = mask_out[k];
    }

    for (size_t k = 0; k < nfft; k++) {
        mask_out[k] = mask_out[k] * (fVar30 * 1.0f) + (1.0f - fVar30);
    }

    // RT_DUMP_BIN: single-frame per-bin tract at frame RT_DUMP_FRAME (default
    // 100): k am res lvl_raw band_level post-IIR1/2, pre-warp mask, W weight.
    if (dbg_path && !dbg_prewarp.empty()) {
        static int dbg_frames = 0;
        int dbg_target = 100;
        if (const char* fs = getenv("RT_DUMP_FRAME")) dbg_target = atoi(fs);
        if (dbg_frames++ != dbg_target) return;
        FILE* df = fopen(dbg_path, "wb");
        if (df) {
            fprintf(df, "# fc=%g q=%g sens=%g rp=%.6f\n", band.fc, band.q, band.sens, rp);
            for (size_t k = 0; k < nbin; k++) {
                double res_k = std::max(static_cast<double>(res[k]), 1e-12);
                double lvl_raw = static_cast<double>(am[k]) / res_k * scale_factor;
                double w = kBand768[k] * kWarp[k] * std::pow(res_k, rp);
                fprintf(df, "%zu %.9g %.9g %.9g %.9g %.9g %.9g\n", k,
                        static_cast<double>(am[k]), res_k, lvl_raw,
                        static_cast<double>(band_level[k]),
                        static_cast<double>(dbg_prewarp[k]), w);
            }
            fclose(df);
        }
    }

    // RT_DUMP_ALL trajectory: append per-frame lvl_raw spectrum (binary:
    // int32 frame, int32 nbin, float32 lvl_raw[nbin]). Single-band cases only.
    // Detector path is law-independent -> one capture serves offline law fits.
    if (const char* ap = getenv("RT_DUMP_ALL")) {
        static FILE* af = fopen(ap, "ab");
        if (af) {
            static int aframe = 0;
            int32_t hdr[2] = {static_cast<int32_t>(aframe++),
                              static_cast<int32_t>(nbin)};
            fwrite(hdr, sizeof(int32_t), 2, af);
            for (size_t k = 0; k < nbin; k++) {
                double res_k = std::max(static_cast<double>(res[k]), 1e-12);
                float lv = static_cast<float>(
                    static_cast<double>(am[k]) / res_k * scale_factor);
                fwrite(&lv, sizeof(float), 1, af);
            }
            fflush(af);
        }
    }
}

// Wrapper that allows cascade curve override for process_band_structural.
// When casc_am is non-null, it replaces the am/res level computation.
// The cascade output IS the level curve (after Haar smooth + sin-peak floor).
// We pass res=1.0 so that am/res = am (cascade already includes twin response).
static void process_band_structural_am(
    const float* am,
    const float* res,
    const DetectorBand& band,
    float* mask_out,
    size_t nfft,
    float sample_rate,
    size_t num_bands = 1,
    const float* casc_curve = nullptr,
    bool use_cascade = false
) {
    if (use_cascade && casc_curve) {
        // Cascade curve IS the level. Pass with res=1.0 to skip am/res division.
        // Create a dummy res array of all 1.0
        static thread_local std::vector<float> one_res;
        size_t nbin = nfft/2 + 1;
        one_res.assign(nbin, 1.0f);
        process_band_structural(casc_curve, one_res.data(), band, mask_out, nfft, sample_rate, num_bands);
    } else {
        process_band_structural(am, res, band, mask_out, nfft, sample_rate, num_bands);
    }
}

}  // namespace

FramedDetector::FramedDetector(size_t nfft, float sample_rate)
    : nfft_(nfft), sample_rate_(sample_rate), wsum_(0) {
    am_.resize(nfft / 2 + 1, 0.0f);
}

FramedDetector::~FramedDetector() {}

void FramedDetector::setParams(const std::vector<DetectorBand>& bands) {
    bands_ = bands;
    size_t half = nfft_ / 2;
    res_.clear();
    track_.clear();
    twin_resp_complex_.clear();
    cascade_states_.clear();

    // RT_DUMPRESPATH=<file> (NOTES 22t): static twin-response spectra per band,
    // binary {int32 band, int32 nbin, float res[nbin]} records (append).
    FILE* rp_dump = nullptr;
    if (const char* dp = getenv("RT_DUMPRESPATH")) rp_dump = fopen(dp, "ab");

    for (const auto& b : bands_) {
        std::vector<float> r(half + 1, 1.0f);
        float sens_lin = std::pow(10.0f, b.sens * SENS_SCALE / 20.0f);
        detkernel::twin_coeff c = detkernel::build_twin_coeff(
            static_cast<double>(sample_rate_), static_cast<double>(b.fc),
            static_cast<double>(b.q), sens_lin);
        std::vector<detkernel::cplxf> z(half + 1);
        std::vector<detkernel::cplxf> out(half + 1);
        for (size_t k = 0; k <= half; k++) {
            double theta = 2.0 * M_PI * static_cast<double>(k) / static_cast<double>(nfft_);
            z[k].re = static_cast<float>(std::cos(theta));
            z[k].im = static_cast<float>(std::sin(theta));
        }
        detkernel::twin_apply(c, z.data(), half + 1, out.data());
        for (size_t k = 0; k <= half; k++) {
            r[k] = std::sqrt(out[k].re * out[k].re + out[k].im * out[k].im);
            // Twin gain floor per NOTES 24dd: plugin caps res ≥0.153 @fc1000q1
            // Our twin 0.0069 at sens24 vs plugin 0.153 (k=22). Floor is
            // content/sens-dependent; default off (canon). Opt-in via RT_TWIN_FLOOR.
            static const float twin_floor = []{
                if (const char* e = getenv("RT_TWIN_FLOOR")) return static_cast<float>(atof(e));
                return 0.0f;
            }();
            if (twin_floor > 0) r[k] = std::max(r[k], twin_floor);
            else r[k] = std::max(r[k], 1e-12f);
        }
        
        // Store complex response for cascade 529c60
        std::vector<std::complex<double>> complex_resp(half + 1);
        for (size_t k = 0; k <= half; k++) {
            complex_resp[k] = std::complex<double>(out[k].re, out[k].im);
        }
        twin_resp_complex_.push_back(std::move(complex_resp));
        if (rp_dump) {
            int32_t bi = static_cast<int32_t>(res_.size());
            int32_t nb = static_cast<int32_t>(r.size());
            fwrite(&bi, sizeof(int32_t), 1, rp_dump);
            fwrite(&nb, sizeof(int32_t), 1, rp_dump);
            fwrite(r.data(), sizeof(float), r.size(), rp_dump);
        }
        res_.push_back(std::move(r));
    }
    if (rp_dump) fclose(rp_dump);
    track_.assign(bands_.size(), std::vector<float>(half + 1, 1.0f));
    cascade_states_.assign(bands_.size(), fn529fe0::CascadeState());
}

static inline double eq_bell(double f, double fc, double q, double sens_db) {
    if (fc < 1.0 || sens_db == 0) return 1.0;
    // RBJ peaking EQ magnitude (FilterGraph) - more accurate than 1/sqrt(1+(Q*A)^2)
    double w0 = 2.0 * M_PI * fc / 48000.0;
    double alpha = std::sin(w0) / (2.0 * q);
    double A = std::pow(10.0, sens_db / 40.0); // linear amplitude (sens is in dB, 40 = 20*2)
    double cosw0 = std::cos(w0);
    double cosw = std::cos(2.0 * M_PI * f / 48000.0);
    // Peaking EQ magnitude squared from RBJ: |H|^2 = (1 + ...)/...
    // Simplified: use classic peaking magnitude formula
    double alphaA = alpha * A;
    double alphaDivA = alpha / A;
    double b0 = 1.0 + alphaA, b1 = -2.0*cosw0, b2 = 1.0 - alphaA;
    double a0 = 1.0 + alphaDivA, a1 = -2.0*cosw0, a2 = 1.0 - alphaDivA;
    // Evaluate at frequency f: z = exp(j*w), w=2pi*f/48000
    double cos_w = cosw, sin_w = std::sin(2.0 * M_PI * f / 48000.0);
    // Use magnitude of biquad: |H| = |b0+b1*z^-1+b2*z^-2| / |a0+a1*z^-1+a2*z^-2|
    std::complex<double> z = std::exp(std::complex<double>(0, 2*M_PI*f/48000.0));
    std::complex<double> z1 = 1.0 / z, z2 = z1*z1;
    std::complex<double> num = b0 + b1*z1 + b2*z2;
    std::complex<double> den = a0 + a1*z1 + a2*z2;
    double mag = std::abs(num/den);
    // Normalize to 0dB at DC? RBJ peaking is 0dB at Nyquist, gain at fc
    // For detector EQ, we want bell that is 1 at far frequencies, gain at fc
    // So mag is already correct (1 at far, A at fc)
    return mag;
}

void FramedDetector::processFrame(const std::complex<double>* spectrum, float* mask) {
    size_t half = nfft_ / 2;
    if (wsum_ == 0.0) {
        double s = 0.0;
        for (size_t i = 0; i < nfft_; i++) {
            s += std::sqrt(0.5 * (1.0 - std::cos(2.0 * M_PI * i / (nfft_ - 1))));
        }
        wsum_ = s;
    }

    double tatt = 0.011, trel = 0.08;
    double att = std::exp(-1.0 * (nfft_ / 4) / (tatt * sample_rate_));
    double rel = std::exp(-1.0 * (nfft_ / 4) / (trel * sample_rate_));

    // RT_ENV=live (NOTES 22n): detector envelope from live tables kRTAtt/kRTRel —
    // attack as feed-forward, release as retention (~tau 2s at hop rate). This is
    // the slow adaptation the real plugin exhibits on sustained content.
    static const int env_live = getenv("RT_ENV") ? atoi(getenv("RT_ENV")) : 0;

    for (size_t k = 0; k <= half; k++) {
        double a_cur = 2.0 * std::abs(spectrum[k]) / wsum_;
        if (env_live) {
            double d = a_cur - static_cast<double>(am_[k]);
            if (d > 0) am_[k] = static_cast<float>(am_[k] + d * static_cast<double>(kRTAtt[k]));
            else       am_[k] = static_cast<float>(static_cast<double>(kRTRel[k]) * am_[k]
                                 + (1.0 - static_cast<double>(kRTRel[k])) * a_cur);
        } else {
            double am = am_[k];
            if (a_cur > am) am = att * am + (1.0 - att) * a_cur;
            else am = rel * am + (1.0 - rel) * a_cur;
            am_[k] = static_cast<float>(am);
        }
    }
    // EQ before detector moved to lvl calc in process_band_structural (not am_ state)

    // Detector cascade 529c60: per-band pre-processor on complex twin-filtered
    // spectrum. Computes magnitudes, Haar-smooths, applies sin-peak floor.
    static const int casc_on = getenv("RT_CASC") ? atoi(getenv("RT_CASC")) : 0;

    for (size_t k = 0; k <= half; k++) mask[k] = 1.0f;

    if (is_internal_grid(nfft_, sample_rate_)) {
        // RT_FAITHFUL=1 (NOTES 22w): BLOCKMAP_529fe0 transcription path
        static const int faithful = getenv("RT_FAITHFUL") ? atoi(getenv("RT_FAITHFUL")) : 0;
        static const fnfaith::Params fparams = faithful ? fnfaith::params_from_env()
                                                        : fnfaith::Params{};
        // same scale_factor as process_band_structural (line ~79)
        constexpr float sf = 15.0f * 440.95f / 2048.0f;
        for (size_t b = 0; b < bands_.size(); b++) {
            std::vector<float> band_mask(nfft_, 1.0f);
            if (faithful) {
                fnfaith::band_mask_faithful(am_.data(), res_[b].data(), half + 1,
                                            sample_rate_, sf, fparams,
                                            band_mask.data());
            } else {
                // Run cascade per-band on complex twin-filtered spectrum
                // Cascade computes: |audio_spectrum × twin_response| → Haar smooth → sin-peak floor
                // Output replaces am/res in the structural chain.
                static thread_local std::vector<float> casc_curve;
                fprintf(stderr, "DBG_CASC casc_on=%d nfft=%zu twin=%zu b=%zu bands=%zu\n", casc_on, nfft_, twin_resp_complex_.size(), b, bands_.size());
                if (casc_on && nfft_ == 4096 && twin_resp_complex_.size() > b) {
                    size_t nbin = half + 1;
                    std::vector<float> complex_input(2 * nbin);
                    casc_curve.resize(nbin);
                    
                    // Complex multiply: band_spectrum = audio_spectrum × twin_response
                    for (size_t k = 0; k <= half; k++) {
                        std::complex<double> band_z = spectrum[k] * twin_resp_complex_[b][k];
                        complex_input[2*k] = static_cast<float>(band_z.real());
                        complex_input[2*k+1] = static_cast<float>(band_z.imag());
                    }

                    fn529fe0::cascade_detect(
                        complex_input.data(),
                        casc_curve.data(),
                        cascade_states_[b],
                        nbin,
                        2,     // Haar iterations
                        0.0f,  // sin_peak_param (0 = no floor; set >0 for Step 9 floor)
                        48000.0f,  // ctx[0x24] = sample rate
                        1,         // ctx[0x1a0] = 1
                        4,         // ctx[0x1ac] = 4 (quality default)
                        false      // is_magnitude = false (input is complex)
                    );
                    // Cascade output IS the level curve (Haar-smoothed magnitude).
                    // Use it directly as am_ replacement — pass res=1.0 so level = am*1
                    // (twin response already baked into cascade output).
                    process_band_structural_am(am_.data(), res_[b].data(), bands_[b],
                                               band_mask.data(), nfft_, sample_rate_,
                                               bands_.size(), casc_curve.data(), true);
                } else {
                    process_band_structural(am_.data(), res_[b].data(), bands_[b],
                                            band_mask.data(), nfft_, sample_rate_, bands_.size());
                }
            }
            for (size_t k = 0; k <= half; k++) {
                mask[k] = std::min(band_mask[k], mask[k]);
            }
        }
    } else {
        for (size_t b = 0; b < bands_.size(); b++) {
            double rp = RP0 * std::pow(static_cast<double>(bands_[b].q), DRP);
            double fk = 0.0;
            double fstep = (sample_rate_ * 0.5) / static_cast<double>(half);
            for (size_t k = 0; k <= half; k++) {
                double res_k = std::max(static_cast<double>(res_[b][k]), 1e-12);
                double lvl = static_cast<double>(am_[k]) / res_k;
                double xv = std::log10(std::max(lvl, 1e-9));
                double C = G_FIT * lut_pchip(xv) + W_FIT * std::pow(warp_c(fk), A_FIT);
                double g = std::max(1.0 - C, 1e-9) * std::pow(res_k, rp);
                mask[k] = std::min(static_cast<float>(g), mask[k]);
                fk += fstep;
            }
        }
    }

    for (size_t k = half + 1; k < nfft_; k++) {
        mask[k] = mask[nfft_ - k];
    }
}