feat(box3): audio narration for the touch gamebook

Context:
The BOX-3 touch gamebook showed text + tappable choices but was silent.
We want narration through its ES8311 speaker while the text is on
screen — and tapping a choice must cut it instantly.

Approach:
Add a dedicated async, interruptible WAV player to cmd_exec (which owns
the speaker handle), streaming straight from the SD in small chunks so
any length plays without a big malloc. The gamebook plays each passage
WAV on render; a new tap (or returning to the menu) stops it.

Changes:
- cmd_exec.c/.h: cmd_exec_play_file_async(path) + cmd_exec_stop_play().
  One persistent task streams a 16 kHz mono 16-bit WAV from the SD in
  2 KB chunks (RAM-safe), and a new request / stop flag breaks the
  stream mid-clip so playback is interruptible.
- gamebook.c: on entering a passage, play /sdcard/gamebook/<wav> if the
  pack has audio (non-fatal when absent → stays text-only); stop the
  narration when returning to the library.

Impact:
The BOX-3 reads the story aloud (16 kHz, matches AUDIO_SAMPLE_RATE)
while showing the text, and a tap cuts to the next passage. Audio is
optional — text-only packs still work.

box3_voice/main/CMakeLists.txt

@@ -1,5 +1,5 @@
 idf_component_register(
-    SRCS "main.c" "voice_ws_client.c" "scenario_server.c" "plip_virtual.c" "plip_ui.c" "stimulus.c" "cmd_exec.c"
+    SRCS "main.c" "voice_ws_client.c" "scenario_server.c" "plip_virtual.c" "plip_ui.c" "stimulus.c" "cmd_exec.c" "gamebook.c"
     INCLUDE_DIRS "."
     PRIV_REQUIRES
         driver

box3_voice/main/cmd_exec.c

@@ -488,6 +488,74 @@ static void play_task(void *arg)
     vTaskDelete(NULL);
 }
 
+// ─── Gamebook narration player (async, streamed, interruptible) ──────────────
+// One persistent task plays one WAV at a time, streaming straight from the SD
+// in small chunks (no full-file malloc → any length, RAM-safe). A new request
+// or a stop flag breaks the stream so tapping a choice cuts the narration.
+
+static TaskHandle_t   s_gb_task = NULL;
+static char           s_gb_path[160];
+static volatile bool  s_gb_new  = false;
+static volatile bool  s_gb_stop = false;
+
+static void gb_play_task(void *arg)
+{
+    (void) arg;
+    uint8_t hdr[64];
+    uint8_t buf[2048];
+    for (;;) {
+        if (!s_gb_new) { vTaskDelay(pdMS_TO_TICKS(20)); continue; }
+        char path[160];
+        strncpy(path, s_gb_path, sizeof(path) - 1);
+        path[sizeof(path) - 1] = '\0';
+        s_gb_new  = false;
+        s_gb_stop = false;
+
+        ensure_sd_mounted();
+        FILE *f = fopen(path, "rb");
+        if (!f) { ESP_LOGW(TAG, "gb: open %s failed", path); continue; }
+        size_t hn = fread(hdr, 1, sizeof(hdr), f);
+        wav_info_t wi;
+        if (parse_wav_header(hdr, hn, &wi) != ESP_OK || wi.bits_per_sample != 16) {
+            ESP_LOGW(TAG, "gb: bad/usupported WAV %s", path);
+            fclose(f);
+            continue;
+        }
+        fseek(f, wi.data_offset, SEEK_SET);
+        size_t remaining = wi.data_size;
+        while (remaining > 0 && !s_gb_stop && !s_gb_new) {
+            size_t want = remaining < sizeof(buf) ? remaining : sizeof(buf);
+            size_t got = fread(buf, 1, want, f);
+            if (got == 0) break;
+            size_t written = 0;
+            if (s_spk_handle_local) {
+                i2s_channel_write(s_spk_handle_local, buf, got, &written,
+                                  pdMS_TO_TICKS(500));
+            }
+            remaining -= got;
+        }
+        fclose(f);
+    }
+}
+
+void cmd_exec_play_file_async(const char *path)
+{
+    if (!path || !path[0]) return;
+    strncpy(s_gb_path, path, sizeof(s_gb_path) - 1);
+    s_gb_path[sizeof(s_gb_path) - 1] = '\0';
+    s_gb_stop = true;    // interrupt whatever is playing
+    s_gb_new  = true;    // … and pick up the new clip
+    if (!s_gb_task) {
+        xTaskCreate(gb_play_task, "gb_play", 4096, NULL, 4, &s_gb_task);
+    }
+}
+
+void cmd_exec_stop_play(void)
+{
+    s_gb_stop = true;
+    s_gb_new  = false;
+}
+
 // ─── op=play ──────────────────────────────────────────────────────────────────
 
 static esp_err_t exec_play(const cJSON *a)

box3_voice/main/cmd_exec.h

@@ -34,6 +34,14 @@ esp_err_t cmd_exec_handle(const char *data, size_t len);
 // Must be called from main.c after speaker_init(), before any CMD arrives.
 void cmd_exec_set_spk_handle(i2s_chan_handle_t h);
 
+// Play a 16 kHz mono 16-bit WAV from the SD asynchronously, streamed in chunks
+// (RAM-safe for long narration) and INTERRUPTIBLE: a new call stops the current
+// clip and starts the new one. Used by the touch gamebook for passage audio.
+void cmd_exec_play_file_async(const char *path);
+
+// Stop the async narration (no-op if nothing is playing).
+void cmd_exec_stop_play(void);
+
 #ifdef __cplusplus
 }
 #endif

box3_voice/main/gamebook.c

@@ -0,0 +1,234 @@
+// gamebook.c — see gamebook.h. Touch CYOA for the ESP32-S3-BOX-3.
+#include "gamebook.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "cJSON.h"
+#include "esp_log.h"
+#include "esp_heap_caps.h"
+
+#include "bsp/esp-bsp.h"
+#include "lvgl.h"
+#include "cmd_exec.h"   // async WAV narration player
+
+static const char *TAG = "gamebook";
+
+#define GB_DIR      "/sdcard/gamebook"
+#define GB_LIBRARY  GB_DIR "/library.json"
+#define GB_JSON_MAX (256 * 1024)        // expanded books are ~80 KB of JSON
+
+// Colours (Workbench-ish palette).
+#define COL_BG     0x101820
+#define COL_TITLE  0xFFCC55
+#define COL_TEXT   0xF0F0F0
+#define COL_BTN    0x224466
+
+static cJSON   *s_lib_root  = NULL;     // owns library.json
+static cJSON   *s_lib       = NULL;     // borrowed: root->"library"
+static int      s_lib_n     = 0;
+
+static cJSON   *s_book      = NULL;     // owns the current <id>.json
+static cJSON   *s_passages  = NULL;     // borrowed: book->"passages"
+
+static lv_obj_t *s_root     = NULL;     // scrollable full-screen column
+
+// ── PSRAM allocators for cJSON (book trees are large) ───────────────────────
+static void *gb_malloc(size_t sz) { return heap_caps_malloc(sz, MALLOC_CAP_SPIRAM); }
+static void  gb_free(void *p)      { heap_caps_free(p); }
+
+static cJSON *load_json(const char *path)
+{
+    FILE *f = fopen(path, "rb");
+    if (!f) { ESP_LOGW(TAG, "open %s failed", path); return NULL; }
+    fseek(f, 0, SEEK_END);
+    long sz = ftell(f);
+    rewind(f);
+    if (sz <= 0 || sz > GB_JSON_MAX) { fclose(f); ESP_LOGW(TAG, "%s size %ld", path, sz); return NULL; }
+    char *buf = heap_caps_malloc((size_t)sz + 1, MALLOC_CAP_SPIRAM);
+    if (!buf) { fclose(f); return NULL; }
+    size_t rd = fread(buf, 1, (size_t)sz, f);
+    fclose(f);
+    buf[rd] = '\0';
+    cJSON *root = cJSON_Parse(buf);
+    heap_caps_free(buf);
+    if (!root) ESP_LOGW(TAG, "malformed JSON: %s", path);
+    return root;
+}
+
+// ── Root container (created once, cleared per view) ─────────────────────────
+// All UI functions below assume the LVGL lock is held (gamebook_init takes it;
+// button callbacks already run inside the LVGL task / lock).
+static void ensure_root(void)
+{
+    if (s_root) { lv_obj_clean(s_root); return; }
+    lv_obj_t *scr = lv_screen_active();
+    s_root = lv_obj_create(scr);
+    lv_obj_set_size(s_root, LV_PCT(100), LV_PCT(100));
+    lv_obj_set_pos(s_root, 0, 0);
+    lv_obj_set_style_bg_color(s_root, lv_color_hex(COL_BG), 0);
+    lv_obj_set_style_bg_opa(s_root, LV_OPA_COVER, 0);
+    lv_obj_set_style_border_width(s_root, 0, 0);
+    lv_obj_set_style_radius(s_root, 0, 0);
+    lv_obj_set_style_pad_all(s_root, 8, 0);
+    lv_obj_set_style_pad_row(s_root, 8, 0);
+    lv_obj_set_flex_flow(s_root, LV_FLEX_FLOW_COLUMN);
+    lv_obj_set_scroll_dir(s_root, LV_DIR_VER);
+}
+
+static lv_obj_t *add_title(const char *txt)
+{
+    lv_obj_t *l = lv_label_create(s_root);
+    lv_obj_set_style_text_font(l, &lv_font_montserrat_24, 0);
+    lv_obj_set_style_text_color(l, lv_color_hex(COL_TITLE), 0);
+    lv_label_set_long_mode(l, LV_LABEL_LONG_WRAP);
+    lv_obj_set_width(l, LV_PCT(100));
+    lv_label_set_text(l, txt);
+    return l;
+}
+
+static lv_obj_t *add_text(const char *txt)
+{
+    lv_obj_t *l = lv_label_create(s_root);
+    lv_obj_set_style_text_font(l, &lv_font_montserrat_14, 0);
+    lv_obj_set_style_text_color(l, lv_color_hex(COL_TEXT), 0);
+    lv_label_set_long_mode(l, LV_LABEL_LONG_WRAP);
+    lv_obj_set_width(l, LV_PCT(100));
+    lv_label_set_text(l, txt);
+    return l;
+}
+
+static lv_obj_t *add_button(const char *txt, lv_event_cb_t cb, void *user)
+{
+    lv_obj_t *b = lv_button_create(s_root);
+    lv_obj_set_width(b, LV_PCT(100));
+    lv_obj_set_style_bg_color(b, lv_color_hex(COL_BTN), 0);
+    lv_obj_set_style_pad_all(b, 10, 0);
+    lv_obj_add_event_cb(b, cb, LV_EVENT_CLICKED, user);
+    lv_obj_t *l = lv_label_create(b);
+    lv_obj_set_style_text_font(l, &lv_font_montserrat_14, 0);
+    lv_label_set_long_mode(l, LV_LABEL_LONG_WRAP);
+    lv_obj_set_width(l, LV_PCT(100));
+    lv_label_set_text(l, txt);
+    return b;
+}
+
+// Forward decls
+static void show_library(void);
+static void enter_passage(const char *pid);
+
+static void tile_cb(lv_event_t *e);
+static void choice_cb(lv_event_t *e);
+static void menu_cb(lv_event_t *e);
+
+// ── Library menu ────────────────────────────────────────────────────────────
+static void show_library(void)
+{
+    cmd_exec_stop_play();   // silence any passage narration
+    if (s_book) { cJSON_Delete(s_book); s_book = NULL; s_passages = NULL; }
+    ensure_root();
+    add_title("Choisis ton histoire");
+    for (int i = 0; i < s_lib_n; i++) {
+        const cJSON *t = cJSON_GetObjectItem(cJSON_GetArrayItem(s_lib, i), "title");
+        add_button(cJSON_IsString(t) ? t->valuestring : "?",
+                   tile_cb, (void *)(intptr_t)i);
+    }
+    lv_obj_scroll_to_y(s_root, 0, LV_ANIM_OFF);
+}
+
+// ── Passage ─────────────────────────────────────────────────────────────────
+static void enter_passage(const char *pid)
+{
+    const cJSON *p = cJSON_GetObjectItem(s_passages, pid);
+    if (!cJSON_IsObject(p)) { ESP_LOGW(TAG, "passage '%s' missing", pid); return; }
+    const cJSON *screen  = cJSON_GetObjectItem(p, "screen");
+    const cJSON *text    = cJSON_GetObjectItem(p, "text");
+    const cJSON *choices = cJSON_GetObjectItem(p, "choices");
+
+    ensure_root();
+    add_title(cJSON_IsString(screen) ? screen->valuestring : "");
+    add_text(cJSON_IsString(text) ? text->valuestring : "");
+
+    // Narration: play this passage's WAV from the SD if the pack has audio.
+    // Interruptible — tapping a choice cuts it and plays the next passage.
+    const cJSON *wav = cJSON_GetObjectItem(p, "wav");
+    if (cJSON_IsString(wav) && wav->valuestring[0]) {
+        char path[160];
+        snprintf(path, sizeof(path), "%s/%s", GB_DIR, wav->valuestring);
+        cmd_exec_play_file_async(path);
+    }
+    int n = cJSON_IsArray(choices) ? cJSON_GetArraySize(choices) : 0;
+    if (n == 0) {
+        add_button("Revenir au menu", menu_cb, NULL);
+    } else {
+        for (int i = 0; i < n; i++) {
+            const cJSON *c = cJSON_GetArrayItem(choices, i);
+            const cJSON *lbl = cJSON_GetObjectItem(c, "label");
+            const cJSON *g   = cJSON_GetObjectItem(c, "goto");
+            // goto valuestring stays valid while s_book is alive → use as user_data
+            add_button(cJSON_IsString(lbl) ? lbl->valuestring : "?",
+                       choice_cb, cJSON_IsString(g) ? g->valuestring : NULL);
+        }
+    }
+    lv_obj_scroll_to_y(s_root, 0, LV_ANIM_OFF);
+    ESP_LOGI(TAG, "passage '%s' (%d choices)", pid, n);
+}
+
+static void load_book(int idx)
+{
+    const cJSON *e = cJSON_GetArrayItem(s_lib, idx);
+    const cJSON *file = cJSON_GetObjectItem(e, "book");
+    if (!cJSON_IsString(file)) return;
+    char path[128];
+    snprintf(path, sizeof(path), "%s/%s", GB_DIR, file->valuestring);
+    cJSON *book = load_json(path);
+    if (!book) return;
+    const cJSON *passages = cJSON_GetObjectItem(book, "passages");
+    const cJSON *start = cJSON_GetObjectItem(book, "start");
+    if (!cJSON_IsObject(passages) || !cJSON_IsString(start)) {
+        cJSON_Delete(book); return;
+    }
+    if (s_book) cJSON_Delete(s_book);
+    s_book = book;
+    s_passages = (cJSON *)passages;
+    ESP_LOGI(TAG, "load book #%d @ '%s'", idx, start->valuestring);
+    enter_passage(start->valuestring);
+}
+
+// ── Touch callbacks (run inside the LVGL task → lock already held) ───────────
+static void tile_cb(lv_event_t *e)   { load_book((int)(intptr_t)lv_event_get_user_data(e)); }
+static void menu_cb(lv_event_t *e)   { (void)e; show_library(); }
+static void choice_cb(lv_event_t *e)
+{
+    const char *goto_id = (const char *)lv_event_get_user_data(e);
+    if (goto_id) enter_passage(goto_id);
+}
+
+// ── Public init ─────────────────────────────────────────────────────────────
+void gamebook_init(void)
+{
+    cJSON_Hooks hooks = { .malloc_fn = gb_malloc, .free_fn = gb_free };
+    cJSON_InitHooks(&hooks);   // keep the big book trees in PSRAM
+
+    if (bsp_sdcard_mount() != ESP_OK) {
+        ESP_LOGW(TAG, "no SD card — gamebook disabled");
+        return;
+    }
+    s_lib_root = load_json(GB_LIBRARY);
+    if (!s_lib_root) { ESP_LOGW(TAG, "no library.json — gamebook disabled"); return; }
+    s_lib = cJSON_GetObjectItem(s_lib_root, "library");
+    if (!cJSON_IsArray(s_lib) || cJSON_GetArraySize(s_lib) == 0) {
+        cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL;
+        ESP_LOGW(TAG, "empty library");
+        return;
+    }
+    s_lib_n = cJSON_GetArraySize(s_lib);
+
+    bsp_display_brightness_set(80);
+    if (bsp_display_lock(1000)) {
+        show_library();
+        bsp_display_unlock();
+    }
+    ESP_LOGI(TAG, "touch gamebook up (%d stories)", s_lib_n);
+}

box3_voice/main/gamebook.h

@@ -0,0 +1,27 @@
+#pragma once
+// gamebook — "livre dont tu es le héros" TACTILE pour l'ESP32-S3-BOX-3.
+//
+// Écran 320x240 + tactile : on lit le passage à l'écran et on TOUCHE un bouton
+// pour choisir. Réutilise le pack gamebook du master sur la SD :
+//   /sdcard/gamebook/{library.json, <id>.json}
+// où chaque passage porte {screen, text, choices:[{label, goto}]} (le champ
+// "wav" éventuel est ignoré — version texte/tactile, pas d'audio requis).
+//
+// S'appuie sur la stack LVGL déjà démarrée par bsp_display_start() : on
+// construit une UUI (menu d'histoires en liste, puis pages tactiles) et la
+// navigation se fait dans les callbacks de boutons LVGL. 100% local, hors-ligne.
+
+#include <stdbool.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Monte la SD, charge la bibliothèque et affiche le menu tactile des histoires.
+// À appeler une fois depuis app_main, après bsp_display_start() et le serveur
+// de fichiers (pour que la SD soit accessible). No-op si aucun pack présent.
+void gamebook_init(void);
+
+#ifdef __cplusplus
+}
+#endif

box3_voice/main/main.c

@@ -28,6 +28,7 @@
 #include "scenario_server.h"
 #include "plip_virtual.h"
 #include "plip_ui.h"
+#include "gamebook.h"
 #include "stimulus.h"
 #include "scenario_mesh.h"
 #include "cmd_exec.h"
@@ -463,6 +464,11 @@ void app_main(void)
         } else {
             ESP_LOGW(TAG, "stimulus_init failed — QR/melody generator off");
         }
+
+        /* Touch gamebook: if a story pack is on the SD (/sdcard/gamebook/),
+         * take over the screen with a tap-to-choose "livre dont tu es le
+         * héros". No-op (and the phone UI stays) when no pack is present. */
+        gamebook_init();
     }
 
     /* Start the ESP-NOW receiver so the master can relay scenarios to us even

idf_zacus/components/display_ui/display_ui.cpp

@@ -248,6 +248,30 @@ static bool     s_browser_isdir[BROWSER_MAX_ENTRIES];
 static uint8_t  s_browser_count = 0;
 static volatile uint8_t s_browser_sel = 0;
 static volatile bool s_browser_open = false;
+
+// ── Gamebook view ("livre dont vous êtes le héros") ─────────────────────────
+static lv_obj_t *s_scr_gamebook;
+static lv_obj_t *s_gb_title;          // page title, top
+static lv_obj_t *s_gb_body;           // full passage text, wrapped
+static lv_obj_t *s_gb_menu;           // choice lines, bottom
+static volatile bool s_gamebook_open = false;
+static lv_obj_t *s_gb_body_cont;      // scrollable container holding s_gb_body
+static char s_gb_title_buf[64];
+static char s_gb_body_buf[1100];
+static char s_gb_menu_buf[160];
+static volatile int  s_gb_scroll_req = 0;   // pending scroll: +pages down / -pages up
+static volatile bool s_gb_scroll_home = false;  // reset scroll to top on new passage
+
+// ── Gamebook library (tile grid) ────────────────────────────────────────────
+#define LIB_TILES 6
+static lv_obj_t *s_scr_library;
+static lv_obj_t *s_lib_tile[LIB_TILES];
+static lv_obj_t *s_lib_label[LIB_TILES];
+static volatile bool s_library_open = false;
+static char s_lib_title_buf[LIB_TILES][48];
+static int  s_lib_count_buf = 0;
+static int  s_lib_sel_buf   = 0;
+
 static volatile bool s_browser_reload = false;
 
 // ─── Intro — faithful port of the original cracktro ──────────────────────────
@@ -761,6 +785,93 @@ scroller:
     lv_obj_set_pos(s_intro_scroll, sx, sy);
 }
 
+static void build_library_screen(void) {
+    s_scr_library = lv_obj_create(NULL);
+    lv_obj_set_style_bg_color(s_scr_library, lv_color_hex(0x000000), 0);
+    lv_obj_set_style_bg_opa(s_scr_library, LV_OPA_COVER, 0);
+    lv_obj_clear_flag(s_scr_library, LV_OBJ_FLAG_SCROLLABLE);
+
+    lv_obj_t *hdr = lv_label_create(s_scr_library);
+    lv_obj_set_style_text_color(hdr, lv_color_hex(COL_CODE), 0);
+    lv_obj_set_style_text_font(hdr, &lv_font_montserrat_14, 0);
+    lv_label_set_text(hdr, "Choisis ton histoire");
+    lv_obj_align(hdr, LV_ALIGN_TOP_MID, 0, 4);
+
+    for (int i = 0; i < LIB_TILES; i++) {
+        int col = i % 2, row = i / 2;
+        lv_obj_t *t = lv_obj_create(s_scr_library);
+        lv_obj_set_size(t, 228, 84);
+        lv_obj_set_pos(t, 8 + col * 236, 26 + row * 94);
+        lv_obj_set_style_radius(t, 8, 0);
+        lv_obj_set_style_bg_color(t, lv_color_hex(0x101820), 0);
+        lv_obj_set_style_bg_opa(t, LV_OPA_COVER, 0);
+        lv_obj_set_style_border_width(t, 2, 0);
+        lv_obj_set_style_border_color(t, lv_color_hex(0x445566), 0);
+        lv_obj_set_style_pad_all(t, 6, 0);
+        lv_obj_clear_flag(t, LV_OBJ_FLAG_SCROLLABLE);
+        lv_obj_t *l = lv_label_create(t);
+        lv_obj_set_style_text_color(l, lv_color_hex(COL_VALUE), 0);
+        lv_obj_set_style_text_font(l, &lv_font_montserrat_14, 0);
+        lv_label_set_long_mode(l, LV_LABEL_LONG_WRAP);
+        lv_obj_set_width(l, 208);
+        lv_obj_set_style_text_align(l, LV_TEXT_ALIGN_CENTER, 0);
+        lv_label_set_text(l, "");
+        lv_obj_center(l);
+        s_lib_tile[i] = t;
+        s_lib_label[i] = l;
+    }
+}
+
+static void build_gamebook_screen(void) {
+    s_scr_gamebook = lv_obj_create(NULL);
+    lv_obj_set_style_bg_color(s_scr_gamebook, lv_color_hex(0x000000), 0);
+    lv_obj_set_style_bg_opa(s_scr_gamebook, LV_OPA_COVER, 0);
+
+    // Compact title line at the very top.
+    s_gb_title = lv_label_create(s_scr_gamebook);
+    lv_obj_set_style_text_color(s_gb_title, lv_color_hex(COL_CODE), 0);
+    lv_obj_set_style_text_font(s_gb_title, &lv_font_montserrat_14, 0);
+    lv_label_set_long_mode(s_gb_title, LV_LABEL_LONG_DOT);
+    lv_obj_set_width(s_gb_title, DUI_HOR_RES - 12);
+    lv_obj_set_style_text_align(s_gb_title, LV_TEXT_ALIGN_CENTER, 0);
+    lv_label_set_text(s_gb_title, "");
+    lv_obj_align(s_gb_title, LV_ALIGN_TOP_MID, 0, 4);
+
+    // Choice selector: a single line pinned at the very bottom. Left/Right cycle
+    // the selection, click confirms. Drawn last (foreground).
+    s_gb_menu = lv_label_create(s_scr_gamebook);
+    lv_obj_set_style_text_color(s_gb_menu, lv_color_hex(COL_CODE), 0);
+    lv_obj_set_style_bg_color(s_gb_menu, lv_color_hex(0x182230), 0);
+    lv_obj_set_style_bg_opa(s_gb_menu, LV_OPA_COVER, 0);
+    lv_obj_set_style_pad_all(s_gb_menu, 5, 0);
+    lv_obj_set_style_text_font(s_gb_menu, &lv_font_montserrat_14, 0);
+    lv_label_set_long_mode(s_gb_menu, LV_LABEL_LONG_DOT);
+    lv_obj_set_width(s_gb_menu, DUI_HOR_RES);
+    lv_obj_set_style_text_align(s_gb_menu, LV_TEXT_ALIGN_CENTER, 0);
+    lv_label_set_text(s_gb_menu, "");
+    lv_obj_align(s_gb_menu, LV_ALIGN_BOTTOM_MID, 0, 0);
+
+    // Reading area: a vertically-scrollable container filling the whole height
+    // between the title and the choice line. Up/Down scroll it by half-pages.
+    s_gb_body_cont = lv_obj_create(s_scr_gamebook);
+    lv_obj_set_style_bg_color(s_gb_body_cont, lv_color_hex(0x000000), 0);
+    lv_obj_set_style_bg_opa(s_gb_body_cont, LV_OPA_COVER, 0);
+    lv_obj_set_style_border_width(s_gb_body_cont, 0, 0);
+    lv_obj_set_style_pad_all(s_gb_body_cont, 6, 0);
+    lv_obj_set_size(s_gb_body_cont, DUI_HOR_RES, 264);   // 24 (title) .. 288 (menu)
+    lv_obj_set_pos(s_gb_body_cont, 0, 24);
+    lv_obj_set_scroll_dir(s_gb_body_cont, LV_DIR_VER);
+    lv_obj_set_scrollbar_mode(s_gb_body_cont, LV_SCROLLBAR_MODE_AUTO);
+
+    s_gb_body = lv_label_create(s_gb_body_cont);
+    lv_obj_set_style_text_color(s_gb_body, lv_color_hex(COL_VALUE), 0);
+    lv_obj_set_style_text_font(s_gb_body, &lv_font_montserrat_14, 0);
+    lv_label_set_long_mode(s_gb_body, LV_LABEL_LONG_WRAP);
+    lv_obj_set_width(s_gb_body, DUI_HOR_RES - 24);   // container width minus padding
+    lv_label_set_text(s_gb_body, "");
+    lv_obj_align(s_gb_body, LV_ALIGN_TOP_LEFT, 0, 0);
+}
+
 static void build_status_screen(void) {
     lv_obj_t *scr = s_scr_status;
     lv_obj_set_style_bg_color(scr, lv_color_hex(COL_BG), 0);
@@ -853,10 +964,49 @@ static void apply_status(const display_status_t *s) {
     // Active step → scene view; idle → status view. The 5-way buttons can
     // override (1=force status, 2=force scene; 0=auto). Fade 240 ms, the
     // original ui_manager default transition (SceneTransition::kFade, 240).
+    // Gamebook page: refresh its labels from the latest show() buffers, then
+    // apply any pending manual scroll (Up/Down) on the reading container.
+    if (s_gamebook_open) {
+        lv_label_set_text(s_gb_title, s_gb_title_buf);
+        lv_label_set_text(s_gb_body,  s_gb_body_buf);
+        lv_label_set_text(s_gb_menu,  s_gb_menu_buf);
+        if (s_gb_scroll_home) {
+            lv_obj_scroll_to_y(s_gb_body_cont, 0, LV_ANIM_OFF);
+            s_gb_scroll_home = false;
+        }
+        int req = s_gb_scroll_req;
+        if (req != 0) {
+            s_gb_scroll_req = 0;
+            // +1 page = read further down (content moves up): negative dy.
+            lv_obj_scroll_by(s_gb_body_cont, 0, -req * 130, LV_ANIM_ON);
+        }
+    }
+    // Library grid: refresh tile titles + highlight the selected one.
+    if (s_library_open) {
+        for (int i = 0; i < LIB_TILES; i++) {
+            if (i < s_lib_count_buf) {
+                lv_label_set_text(s_lib_label[i], s_lib_title_buf[i]);
+                lv_obj_clear_flag(s_lib_tile[i], LV_OBJ_FLAG_HIDDEN);
+            } else {
+                lv_obj_add_flag(s_lib_tile[i], LV_OBJ_FLAG_HIDDEN);
+            }
+            bool on = (i == s_lib_sel_buf);
+            lv_obj_set_style_border_color(s_lib_tile[i],
+                lv_color_hex(on ? COL_CODE : 0x445566), 0);
+            lv_obj_set_style_border_width(s_lib_tile[i], on ? 4 : 2, 0);
+            lv_obj_set_style_bg_color(s_lib_tile[i],
+                lv_color_hex(on ? 0x223040 : 0x101820), 0);
+        }
+    }
+
     extern volatile uint8_t g_dui_view_override;
     lv_obj_t *want;
     if (!s_intro_done) {
         want = s_scr_intro;            // boot intro plays out first
+    } else if (s_library_open) {
+        want = s_scr_library;          // story picker owns the screen
+    } else if (s_gamebook_open) {
+        want = s_scr_gamebook;         // gamebook owns the screen while running
     } else if (s_browser_open) {
         want = s_scr_browser;          // file browser on top of everything
     } else if (s_shell_open) {
@@ -964,6 +1114,8 @@ static void display_task(void *arg) {
     build_status_screen();
     build_scene_screen();
     build_shell_screen();
+    build_gamebook_screen();
+    build_library_screen();
     build_browser_screen();
     build_intro_screen();
 
@@ -1145,9 +1297,67 @@ extern "C" void display_ui_set_status(const display_status_t *s) {
     }
 }
 
+static display_ui_key_hook_t s_key_hook = nullptr;
+
+extern "C" void display_ui_set_key_hook(display_ui_key_hook_t hook) {
+    s_key_hook = hook;
+}
+
+extern "C" void display_ui_gamebook_show(const char *title, const char *body,
+                                         const char *menu, bool reset_scroll) {
+    if (s_mutex) xSemaphoreTake(s_mutex, portMAX_DELAY);
+    snprintf(s_gb_title_buf, sizeof(s_gb_title_buf), "%s", title ? title : "");
+    snprintf(s_gb_body_buf,  sizeof(s_gb_body_buf),  "%s", body  ? body  : "");
+    snprintf(s_gb_menu_buf,  sizeof(s_gb_menu_buf),  "%s", menu  ? menu  : "");
+    if (reset_scroll) s_gb_scroll_home = true;   // new passage → back to top
+    s_gamebook_open = true;
+    s_dirty = true;
+    if (s_mutex) xSemaphoreGive(s_mutex);
+}
+
+extern "C" void display_ui_gamebook_scroll(int dir) {
+    // dir > 0 : read further down ; dir < 0 : back up. Accumulated and applied
+    // on the display task (this runs on the buttons task).
+    s_gb_scroll_req += (dir > 0) ? 1 : -1;
+    s_dirty = true;
+}
+
+extern "C" void display_ui_gamebook_hide(void) {
+    s_gamebook_open = false;
+    s_gb_scroll_req = 0;
+    s_dirty = true;
+}
+
+extern "C" void display_ui_library_show(const char *const *titles, int count,
+                                        int sel) {
+    if (count > LIB_TILES) count = LIB_TILES;
+    if (count < 0) count = 0;
+    if (s_mutex) xSemaphoreTake(s_mutex, portMAX_DELAY);
+    s_lib_count_buf = count;
+    s_lib_sel_buf   = (count > 0) ? ((sel % count + count) % count) : 0;
+    for (int i = 0; i < LIB_TILES; i++) {
+        snprintf(s_lib_title_buf[i], sizeof(s_lib_title_buf[i]), "%s",
+                 (i < count && titles && titles[i]) ? titles[i] : "");
+    }
+    s_library_open  = true;
+    s_gamebook_open = false;   // library and a running story are exclusive
+    s_dirty = true;
+    if (s_mutex) xSemaphoreGive(s_mutex);
+}
+
+extern "C" void display_ui_library_hide(void) {
+    s_library_open = false;
+    s_dirty = true;
+}
+
 extern "C" void display_ui_handle_key(uint8_t key) {
     static uint8_t s_brightness = 100;
 
+    // Takeover mode (e.g. gamebook): if the hook consumes the key, stop here.
+    if (s_key_hook && s_key_hook(key)) {
+        return;
+    }
+
     // Any key skips the boot intro.
     if (!s_intro_done) {
         s_intro_done = true;

idf_zacus/components/display_ui/include/display_ui.h

@@ -78,6 +78,59 @@ void display_ui_camera_frame(const uint8_t *gray, int width, int height);
  */
 void display_ui_handle_key(uint8_t key);
 
+/**
+ * @brief Install an optional key interceptor.
+ *
+ * If a hook is set and returns true for a given key, the key is consumed and
+ * the normal shell/scene handling is skipped. Lets a takeover mode (e.g. the
+ * gamebook) own the 5-way pad without display_ui having to depend on it.
+ * Pass NULL to remove the hook. Thread-safe (single pointer write).
+ */
+typedef bool (*display_ui_key_hook_t)(uint8_t key);
+void display_ui_set_key_hook(display_ui_key_hook_t hook);
+
+/**
+ * @brief Show a full-screen gamebook page (takes over the display).
+ *
+ * Dedicated "livre dont vous êtes le héros" view: a centred title, the full
+ * passage text wrapped over multiple lines, and a choice menu at the bottom.
+ * Forces this view until display_ui_gamebook_hide(). Thread-safe (buffers
+ * copied under the mutex; the actual LVGL render runs on the display task).
+ * Strings should be ASCII (the embedded fonts have no accents).
+ *
+ * @param title  short page title (top)
+ * @param body   full passage text (wrapped); may be long
+ * @param menu   choice lines (bottom), e.g. "[OK] ...\n[<>] ..."
+ */
+void display_ui_gamebook_show(const char *title, const char *body,
+                              const char *menu, bool reset_scroll);
+
+/**
+ * @brief Scroll the gamebook reading area (Up/Down buttons).
+ * @param dir  >0 = read further down, <0 = back up. Half-page per call.
+ */
+void display_ui_gamebook_scroll(int dir);
+
+/** @brief Leave the gamebook view and return to the normal scene/status flow. */
+void display_ui_gamebook_hide(void);
+
+/**
+ * @brief Show the gamebook library as a grid of up to 6 tiles.
+ *
+ * Each tile shows a story title; the tile at index `sel` is highlighted. The
+ * gamebook owns the pad and calls this on every cursor move. Forces the
+ * library view until a story is loaded or display_ui_library_hide(). Titles
+ * should be ASCII.
+ *
+ * @param titles  array of `count` story titles
+ * @param count   number of stories (clamped to 6)
+ * @param sel     highlighted tile index
+ */
+void display_ui_library_show(const char *const *titles, int count, int sel);
+
+/** @brief Leave the library view. */
+void display_ui_library_hide(void);
+
 /**
  * @brief Pop the pending shell-app launch request, if any.
  *

idf_zacus/components/game_endpoint/CMakeLists.txt

@@ -17,9 +17,11 @@ idf_component_register(
         puzzle_state
         media_manager
         sd_storage
+        gamebook
     PRIV_REQUIRES
         local_puzzles
         p7_coffre
         p5_morse
         p6_nfc
+        npc_engine
 )

idf_zacus/components/game_endpoint/game_endpoint.c

@@ -27,6 +27,8 @@
 #include "nvs_flash.h"
 
 #include "hints_client.h"
+#include "gamebook.h"
+#include "npc_engine.h"
 #include "scenario_mesh.h"
 #include "puzzle_binding.h"
 #include "local_puzzles.h"
@@ -807,6 +809,16 @@ esp_err_t game_endpoint_apply_step(const char *step_id,
     strncpy(s_current_step_id, step_id, sizeof(s_current_step_id) - 1);
     s_current_step_id[sizeof(s_current_step_id) - 1] = '\0';
 
+    // Sync the NPC/hints engine (and, through it, the voice gateway) with the
+    // scene this step enters. This is what makes the phone NPCs hint on — and
+    // scripted calls auto-ring for — the current scene, regardless of whether
+    // the step was driven by the local game loop or an external POST /game/step.
+    // Best-effort: notify is async + server-side idempotent and never blocks
+    // the step change. Steps with no scene_id (or an unknown one) are skipped.
+    if (s_current_scene.scene_id[0] != '\0') {
+        (void) npc_engine_set_scene_by_id(s_current_scene.scene_id, 0);
+    }
+
     // P7 coffre: fire the actuator when the final win step is reached.
     // p7_coffre_unlock() is a no-op stub when CONFIG_ZACUS_P7_COFFRE_ENABLE=n.
     if (strcmp(step_id, "STEP_FINAL_WIN") == 0) {
@@ -983,6 +995,29 @@ static esp_err_t handle_media_play_post(httpd_req_t *req) {
     return send_error(req, "500 Internal Server Error", esp_err_to_name(err));
 }
 
+// ─── POST /game/gamebook[?action=stop] ───────────────────────────────────────
+//
+// Start the "livre dont vous êtes le héros" mode (loads /sdcard/gamebook/
+// gamebook.json, plays the start passage, takes over the 5-way pad). With
+// ?action=stop it leaves gamebook mode. No body required.
+static esp_err_t handle_gamebook_post(httpd_req_t *req) {
+    char query[32] = {0}, action[16] = {0};
+    if (httpd_req_get_url_query_str(req, query, sizeof(query)) == ESP_OK) {
+        httpd_query_key_value(query, "action", action, sizeof(action));
+    }
+    if (strcmp(action, "stop") == 0) {
+        gamebook_stop();
+        return send_json(req, "200 OK", "{\"gamebook\":\"stopped\"}");
+    }
+    esp_err_t err = gamebook_start();
+    if (err != ESP_OK) {
+        return send_error(req, "503 Service Unavailable",
+                          err == ESP_ERR_NOT_FOUND ? "gamebook_not_on_sd"
+                                                   : "gamebook_load_failed");
+    }
+    return send_json(req, "200 OK", "{\"gamebook\":\"started\"}");
+}
+
 // ─── GET /game/puzzle_state ──────────────────────────────────────────────────
 //
 // Returns {"step_id":"STEP_X"|null, "solved":[1,3], "code":"125"}.
@@ -1138,6 +1173,43 @@ static esp_err_t handle_file_post(httpd_req_t *req) {
     return send_json(req, "200 OK", resp);
 }
 
+// ─── DELETE /game/file?path=sd/<…>|apps/<…> — remove a staged file ───────────
+// Same path routing/whitelist as the POST handler. Lets the host clean stale
+// assets (e.g. orphan gamebook WAVs) off the SD without reflashing.
+static esp_err_t handle_file_delete(httpd_req_t *req) {
+    char query[160], path_param[96];
+    if (httpd_req_get_url_query_str(req, query, sizeof(query)) != ESP_OK ||
+        httpd_query_key_value(query, "path", path_param,
+                              sizeof(path_param)) != ESP_OK) {
+        return send_error(req, "400 Bad Request", "missing path param");
+    }
+    const bool to_sd   = (strncmp(path_param, "sd/", 3) == 0);
+    const bool to_apps = (strncmp(path_param, "apps/", 5) == 0);
+    if ((!to_sd && !to_apps) || strstr(path_param, "..") ||
+        path_param[strlen(path_param) - 1] == '/') {
+        return send_error(req, "403 Forbidden", "path must be under apps/ or sd/");
+    }
+    char full[192];
+    if (to_sd) {
+        if (!sd_storage_ready()) {
+            return send_error(req, "503 Service Unavailable", "sd_not_mounted");
+        }
+        snprintf(full, sizeof(full), "/sdcard/%s", path_param + 3);
+    } else {
+        if (mount_storage_lazy() != ESP_OK) {
+            return send_error(req, "503 Service Unavailable", "storage_unavailable");
+        }
+        snprintf(full, sizeof(full), "/littlefs/%s", path_param);
+    }
+    if (unlink(full) != 0) {
+        return send_error(req, "404 Not Found", "remove failed");
+    }
+    ESP_LOGI(TAG, "file removed: %s", full);
+    char resp[192];
+    snprintf(resp, sizeof(resp), "{\"status\":\"ok\",\"removed\":\"%s\"}", path_param);
+    return send_json(req, "200 OK", resp);
+}
+
 void game_endpoint_get_scene(scene_binding_t *out) {
     if (!out) return;
     // Lock-free snapshot, same class as get_puzzle_status: written on the
@@ -1209,6 +1281,12 @@ esp_err_t game_endpoint_init(httpd_handle_t server) {
         .handler  = handle_file_post,
         .user_ctx = NULL,
     };
+    static const httpd_uri_t uri_file_delete = {
+        .uri      = "/game/file",
+        .method   = HTTP_DELETE,
+        .handler  = handle_file_delete,
+        .user_ctx = NULL,
+    };
     static const httpd_uri_t uri_espnow_cmd_post = {
         .uri      = "/game/espnow/cmd",
         .method   = HTTP_POST,
@@ -1221,6 +1299,12 @@ esp_err_t game_endpoint_init(httpd_handle_t server) {
         .handler  = handle_media_play_post,
         .user_ctx = NULL,
     };
+    static const httpd_uri_t uri_gamebook_post = {
+        .uri      = "/game/gamebook",
+        .method   = HTTP_POST,
+        .handler  = handle_gamebook_post,
+        .user_ctx = NULL,
+    };
 
     // Bring up the ESP-NOW mesh transport. The master is primarily a sender
     // (the relay handler) but we also register the apply adapter so a peer
@@ -1285,6 +1369,10 @@ esp_err_t game_endpoint_init(httpd_handle_t server) {
     if (err != ESP_OK) {
         ESP_LOGW(TAG, "register POST /game/step: %s", esp_err_to_name(err));
     }
+    err = httpd_register_uri_handler(server, &uri_file_delete);
+    if (err != ESP_OK) {
+        ESP_LOGW(TAG, "register DELETE /game/file: %s", esp_err_to_name(err));
+    }
     err = httpd_register_uri_handler(server, &uri_file_post);
     if (err != ESP_OK) {
         ESP_LOGW(TAG, "register /game/file: %s", esp_err_to_name(err));
@@ -1297,6 +1385,10 @@ esp_err_t game_endpoint_init(httpd_handle_t server) {
     if (err != ESP_OK) {
         ESP_LOGW(TAG, "register POST /game/media/play: %s", esp_err_to_name(err));
     }
+    err = httpd_register_uri_handler(server, &uri_gamebook_post);
+    if (err != ESP_OK) {
+        ESP_LOGW(TAG, "register POST /game/gamebook: %s", esp_err_to_name(err));
+    }
 
     ESP_LOGI(TAG, "game endpoint registered "
                   "(GET+POST /game/group_profile, POST /game/scenario%s, "

idf_zacus/components/game_endpoint/include/puzzle_binding.h

@@ -55,6 +55,7 @@ esp_err_t puzzle_binding_from_ir(const char *ir_json, const char *step_id,
 #define SB_MAX_TITLE    48
 #define SB_MAX_SUBTITLE 64
 #define SB_MAX_SYMBOL   16
+#define SB_MAX_SCENE_ID 40
 
 typedef enum {
     SB_FX_PULSE = 0,   // default (original SceneEffect::kPulse)
@@ -65,6 +66,8 @@ typedef enum {
 
 typedef struct {
     bool present;                     // step has a scene object
+    char scene_id[SB_MAX_SCENE_ID];   // canonical SCENE_* id (sibling of step id;
+                                      // set even when no display scene object)
     char title[SB_MAX_TITLE];
     char subtitle[SB_MAX_SUBTITLE];
     char symbol[SB_MAX_SYMBOL];

idf_zacus/components/game_endpoint/puzzle_binding.c

@@ -239,9 +239,15 @@ esp_err_t scene_binding_from_ir(const char *ir_json, const char *step_id,
     }
     if (!step) { cJSON_Delete(root); return ESP_ERR_NOT_FOUND; }
 
+    // Canonical scene id (SCENE_*), sibling of the step "id". Captured
+    // unconditionally so npc_engine / the voice gateway can be synced even
+    // on steps that carry no display "scene" object.
+    copy_scene_str(out->scene_id, sizeof(out->scene_id),
+                   cJSON_GetObjectItemCaseSensitive(step, "scene_id"));
+
     const cJSON *scene = cJSON_GetObjectItemCaseSensitive(step, "scene");
     if (!scene || !cJSON_IsObject(scene)) {
-        // No scene — present stays false, ESP_OK.
+        // No scene object — present stays false, ESP_OK (scene_id may be set).
         cJSON_Delete(root);
         return ESP_OK;
     }

idf_zacus/components/gamebook/CMakeLists.txt

@@ -0,0 +1,11 @@
+idf_component_register(
+    SRCS
+        "gamebook.c"
+    INCLUDE_DIRS
+        "include"
+    REQUIRES
+        json
+        display_ui
+        media_manager
+        log
+)

idf_zacus/components/gamebook/gamebook.c

@@ -0,0 +1,293 @@
+// gamebook.c — see gamebook.h. Standalone gamebook player + library for the
+// Freenove master. On boot it shows the library (a grid of story tiles);
+// picking one loads and plays that book; finishing returns to the library.
+#include "gamebook.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "cJSON.h"
+#include "esp_log.h"
+
+#include "display_ui.h"
+#include "media_manager.h"
+
+static const char *TAG = "gamebook";
+
+#define GAMEBOOK_DIR   "/sdcard/gamebook"
+#define LIBRARY_JSON   GAMEBOOK_DIR "/library.json"
+#define JSON_MAX       (64 * 1024)   /* per-file JSON sanity cap */
+#define LIB_MAX        6             /* matches the display tile grid */
+
+typedef enum { GB_OFF, GB_LIBRARY, GB_STORY } gb_mode_t;
+
+// Library state
+static cJSON   *s_lib_root  = NULL;  /* owns library.json while loaded   */
+static cJSON   *s_lib       = NULL;  /* borrowed: root->"library" array  */
+static int      s_lib_n     = 0;
+static int      s_lib_sel   = 0;
+
+// Story state
+static cJSON   *s_book      = NULL;  /* owns the current <book>.json      */
+static cJSON   *s_passages  = NULL;  /* borrowed: book->"passages"        */
+static char     s_title[48] = {0};
+static char     s_current[48] = {0};
+static int      s_sel       = 0;     /* highlighted choice index          */
+
+static volatile gb_mode_t s_mode = GB_OFF;
+
+bool gamebook_active(void) { return s_mode != GB_OFF; }
+
+/* Read a whole JSON file from the SD into a cJSON tree (caller frees). */
+static cJSON *load_json(const char *path)
+{
+    FILE *f = fopen(path, "rb");
+    if (!f) { ESP_LOGW(TAG, "open %s failed", path); return NULL; }
+    fseek(f, 0, SEEK_END);
+    long sz = ftell(f);
+    rewind(f);
+    if (sz <= 0 || sz > JSON_MAX) { fclose(f); ESP_LOGW(TAG, "%s size %ld", path, sz); return NULL; }
+    char *buf = malloc((size_t)sz + 1);
+    if (!buf) { fclose(f); return NULL; }
+    size_t rd = fread(buf, 1, (size_t)sz, f);
+    fclose(f);
+    buf[rd] = '\0';
+    cJSON *root = cJSON_Parse(buf);
+    free(buf);
+    if (!root) ESP_LOGW(TAG, "malformed JSON: %s", path);
+    return root;
+}
+
+// ── Library ──────────────────────────────────────────────────────────────────
+
+static void render_library(void)
+{
+    const char *titles[LIB_MAX] = {0};
+    int n = (s_lib_n < LIB_MAX) ? s_lib_n : LIB_MAX;
+    for (int i = 0; i < n; i++) {
+        const cJSON *t = cJSON_GetObjectItem(cJSON_GetArrayItem(s_lib, i), "title");
+        titles[i] = cJSON_IsString(t) ? t->valuestring : "?";
+    }
+    display_ui_library_show(titles, n, s_lib_sel);
+}
+
+static void free_story(void)
+{
+    media_manager_stop();
+    if (s_book) { cJSON_Delete(s_book); s_book = NULL; }
+    s_passages = NULL;
+    s_current[0] = '\0';
+}
+
+static esp_err_t open_library(void)
+{
+    free_story();
+    if (s_lib_root) { cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL; }
+
+    s_lib_root = load_json(LIBRARY_JSON);
+    if (!s_lib_root) return ESP_ERR_NOT_FOUND;
+    s_lib = cJSON_GetObjectItem(s_lib_root, "library");
+    if (!cJSON_IsArray(s_lib) || cJSON_GetArraySize(s_lib) == 0) {
+        cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL;
+        ESP_LOGW(TAG, "library.json has no stories");
+        return ESP_ERR_INVALID_ARG;
+    }
+    s_lib_n = cJSON_GetArraySize(s_lib);
+    s_lib_sel = 0;
+    s_mode = GB_LIBRARY;
+    display_ui_gamebook_hide();
+    render_library();
+    ESP_LOGI(TAG, "library open (%d stories)", s_lib_n);
+    return ESP_OK;
+}
+
+// ── Story ─────────────────────────────────────────────────────────────────────
+
+static const cJSON *cur_passage(void) { return cJSON_GetObjectItem(s_passages, s_current); }
+
+static int cur_choice_count(void)
+{
+    const cJSON *ch = cJSON_GetObjectItem(cur_passage(), "choices");
+    return cJSON_IsArray(ch) ? cJSON_GetArraySize(ch) : 0;
+}
+
+/* (Re)draw the current page. `home` true = new passage (reset scroll to top);
+ * false = same passage, only the choice cursor moved (keep scroll position).
+ * The bottom line shows ONE choice with < > arrows (Left/Right cycles). */
+static void render_page(bool home)
+{
+    const cJSON *p = cur_passage();
+    if (!cJSON_IsObject(p)) return;
+    const cJSON *screen  = cJSON_GetObjectItem(p, "screen");
+    const cJSON *text    = cJSON_GetObjectItem(p, "text");
+    const cJSON *choices = cJSON_GetObjectItem(p, "choices");
+    int n = cJSON_IsArray(choices) ? cJSON_GetArraySize(choices) : 0;
+
+    char menu[160];
+    if (n == 0) {
+        snprintf(menu, sizeof(menu), "~ Fin ~   (clic = bibliotheque)");
+    } else {
+        const cJSON *lbl = cJSON_GetObjectItem(cJSON_GetArrayItem(choices, s_sel),
+                                               "label");
+        const char *txt = cJSON_IsString(lbl) ? lbl->valuestring : "?";
+        if (n > 1) {
+            /* Down cycles the answer, Click validates. */
+            snprintf(menu, sizeof(menu), "%d/%d  %s   (bas / clic)",
+                     s_sel + 1, n, txt);
+        } else {
+            snprintf(menu, sizeof(menu), "%s   (clic)", txt);
+        }
+    }
+    display_ui_gamebook_show(
+        cJSON_IsString(screen) ? screen->valuestring : s_title,
+        cJSON_IsString(text)   ? text->valuestring   : "",
+        menu, home);
+}
+
+static void enter_passage(const char *pid)
+{
+    const cJSON *p = cJSON_GetObjectItem(s_passages, pid);
+    if (!cJSON_IsObject(p)) { ESP_LOGW(TAG, "passage '%s' not found", pid); return; }
+    snprintf(s_current, sizeof(s_current), "%s", pid);
+    s_sel = 0;
+    render_page(true);
+
+    const cJSON *wav = cJSON_GetObjectItem(p, "wav");
+    if (cJSON_IsString(wav) && wav->valuestring[0]) {
+        char path[96];
+        snprintf(path, sizeof(path), "%s/%s", GAMEBOOK_DIR, wav->valuestring);
+        media_manager_stop();   /* a choice skips the narration */
+        media_manager_play(path);
+    }
+    ESP_LOGI(TAG, "passage '%s' (%d choices)", pid, cur_choice_count());
+}
+
+/* Load story #idx from the library and play it. */
+static void load_book(int idx)
+{
+    const cJSON *e = cJSON_GetArrayItem(s_lib, idx);
+    const cJSON *file = cJSON_GetObjectItem(e, "book");
+    if (!cJSON_IsString(file)) { ESP_LOGW(TAG, "library[%d] has no book", idx); return; }
+
+    char path[128];
+    snprintf(path, sizeof(path), "%s/%s", GAMEBOOK_DIR, file->valuestring);
+    cJSON *book = load_json(path);
+    if (!book) return;
+    const cJSON *passages = cJSON_GetObjectItem(book, "passages");
+    const cJSON *start = cJSON_GetObjectItem(book, "start");
+    const cJSON *title = cJSON_GetObjectItem(book, "title");
+    if (!cJSON_IsObject(passages) || !cJSON_IsString(start)) {
+        ESP_LOGW(TAG, "%s missing passages/start", path);
+        cJSON_Delete(book);
+        return;
+    }
+    free_story();
+    s_book = book;
+    s_passages = (cJSON *)passages;
+    snprintf(s_title, sizeof(s_title), "%s",
+             cJSON_IsString(title) ? title->valuestring : "");
+    s_mode = GB_STORY;
+    display_ui_library_hide();
+    ESP_LOGI(TAG, "load book \"%s\" @ '%s'", s_title, start->valuestring);
+    enter_passage(start->valuestring);
+}
+
+// ── Input ─────────────────────────────────────────────────────────────────────
+
+/* MEASURED pad mapping on this hardware (ADC ladder, see calibration):
+ *   key1 = CLICK (centre, ~0 mV)
+ *   key2 = LEFT  (gauche, ~700 mV)
+ *   key4 = DOWN  (bas,    ~1350 mV)
+ *   key5 = RIGHT (droite, ~1994 mV)
+ *   key3 is never produced; the physical UP button is electrically dead
+ *   (held = no voltage change), so nothing can depend on it. */
+#define PAD_CLICK 1
+#define PAD_LEFT  2
+#define PAD_DOWN  4
+#define PAD_RIGHT 5
+
+/* Library nav: Down/Right → next tile, Left → previous, Click → open.
+ * Selection wraps, so the working buttons reach every story without Up. */
+static bool library_key(uint8_t key)
+{
+    if (s_lib_n <= 0) return true;
+    switch (key) {
+    case PAD_DOWN: case PAD_RIGHT:
+        s_lib_sel = (s_lib_sel + 1) % s_lib_n; render_library(); break;
+    case PAD_LEFT:
+        s_lib_sel = (s_lib_sel - 1 + s_lib_n) % s_lib_n; render_library(); break;
+    case PAD_CLICK:
+        load_book(s_lib_sel); break;
+    default: break;
+    }
+    return true;
+}
+
+/* Story nav (Up is dead, so it's never used):
+ *   Left  → scroll the reading text UP
+ *   Right → scroll the reading text DOWN
+ *   Down  → next answer (cycles through all choices, wraps around)
+ *   Click → validate the highlighted answer
+ * On an ending page (no choices): Left/Right still scroll, Click → library. */
+static bool story_key(uint8_t key)
+{
+    if (key == PAD_LEFT)  { display_ui_gamebook_scroll(-1); return true; }  /* scroll up   */
+    if (key == PAD_RIGHT) { display_ui_gamebook_scroll(+1); return true; }  /* scroll down */
+
+    int n = cur_choice_count();
+    if (n == 0) {                       /* ending → click returns to library */
+        if (key == PAD_CLICK) open_library();
+        return true;
+    }
+    switch (key) {
+    case PAD_DOWN: s_sel = (s_sel + 1) % n; render_page(false); break;  /* next answer */
+    case PAD_CLICK: {                                                   /* validate */
+        const cJSON *choices = cJSON_GetObjectItem(cur_passage(), "choices");
+        const cJSON *g = cJSON_GetObjectItem(cJSON_GetArrayItem(choices, s_sel), "goto");
+        if (cJSON_IsString(g)) {
+            ESP_LOGI(TAG, "validate #%d -> '%s'", s_sel, g->valuestring);
+            enter_passage(g->valuestring);
+        }
+        break;
+    }
+    default: break;
+    }
+    return true;
+}
+
+static bool gamebook_key_hook(uint8_t key)
+{
+    switch (s_mode) {
+    case GB_LIBRARY: return library_key(key);
+    case GB_STORY:   return story_key(key);
+    default:         return false;   /* not active → let the shell have it */
+    }
+}
+
+// ── Public API ────────────────────────────────────────────────────────────────
+
+esp_err_t gamebook_start(void)
+{
+    return open_library();
+}
+
+void gamebook_stop(void)
+{
+    if (s_mode == GB_OFF) return;
+    s_mode = GB_OFF;
+    free_story();
+    if (s_lib_root) { cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL; }
+    display_ui_gamebook_hide();
+    display_ui_library_hide();
+    ESP_LOGI(TAG, "stopped");
+}
+
+void gamebook_init(void)
+{
+    /* Only register the pad hook here — the SD and media_manager aren't up yet
+     * at this point in boot. main.c calls gamebook_start() once they are, to
+     * boot into the library. */
+    display_ui_set_key_hook(gamebook_key_hook);
+    ESP_LOGI(TAG, "key hook installed");
+}

idf_zacus/components/gamebook/include/gamebook.h

@@ -0,0 +1,42 @@
+#pragma once
+// gamebook — "livre dont vous êtes le héros" mode for the Freenove master.
+//
+// Reads /sdcard/gamebook/gamebook.json (built by tools/gamebook/build_gamebook.py),
+// plays each passage's WAV narration from the SD via media_manager, shows the
+// passage title + choice menu on the display, and navigates with the 5-way pad
+// (it installs a display_ui key hook so it owns the buttons while active).
+// Fully local: no model, no gateway.
+//
+// JSON shape:
+//   { "title": "...", "start": "intro",
+//     "passages": { "intro": { "wav": "intro.wav", "screen": "...",
+//                              "menu": "[OK] ... | [<>] ...",
+//                              "choices": [ {"key": 1, "goto": "machine"}, ... ] },
+//                   ... } }
+// key codes match the 5-way pad: 1=SELECT 2=DOWN 3=MENU 4=LEFT/RIGHT 5=UP.
+// MENU (3) always quits the gamebook.
+
+#include "esp_err.h"
+#include <stdbool.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Install the display_ui key hook. Call once at boot, after display_ui_init().
+void gamebook_init(void);
+
+// Load /sdcard/gamebook/gamebook.json and enter the start passage (show text +
+// play its WAV). Returns ESP_ERR_* if the SD/JSON is missing or malformed.
+esp_err_t gamebook_start(void);
+
+// Leave gamebook mode: stop playback, release the JSON, hand the pad back to
+// the shell. Idempotent.
+void gamebook_stop(void);
+
+// True while a gamebook is running (used by the key hook to claim the pad).
+bool gamebook_active(void);
+
+#ifdef __cplusplus
+}
+#endif

idf_zacus/components/npc_engine/CMakeLists.txt

@@ -22,6 +22,7 @@ idf_component_register(
     REQUIRES
         media_manager
         hints_client
+        esp_http_client
         esp_timer
         esp_system
         nvs_flash

idf_zacus/components/npc_engine/include/npc_engine.h

@@ -169,6 +169,16 @@ esp_err_t npc_engine_trigger_cue(const char *cue_id);
 // and primes the stuck timer.
 esp_err_t npc_engine_set_step(uint8_t step_id, uint32_t expected_duration_ms);
 
+// Same bridge as npc_engine_set_step(), but keyed by the canonical scene
+// string id (e.g. "SCENE_WARNING") rather than its numeric index. Resolves
+// the id against the engine's internal scene table and forwards to
+// npc_engine_set_step(). Lets the game endpoint sync the NPC/hints engine
+// (and the voice gateway) straight from the scenario IR's `scene_id`.
+// Returns ESP_ERR_NOT_FOUND when the id matches no known scene (no notify
+// is sent in that case), ESP_ERR_INVALID_ARG on a NULL/empty id.
+esp_err_t npc_engine_set_scene_by_id(const char *scene_id,
+                                     uint32_t expected_duration_ms);
+
 // Request a hint for `puzzle_id` at escalation `level` (0..3). The engine
 // invokes `cb` with the resulting text. The current implementation is a
 // LOCAL STUB: it returns a hardcoded French placeholder synchronously.
@@ -184,6 +194,12 @@ const npc_state_t *npc_engine_state(void);
 // Thin wrapper kept here to give callers a single npc_engine_* surface.
 esp_err_t npc_engine_set_group_profile(const char *profile);
 
+// Point the engine at the voice gateway (tools/zacus-gateway). On each scene
+// change the engine POSTs the active SCENE_* to {base_url}/game/step?scene=… so
+// the phone NPCs disguise that scene's hint. Best-effort; pass NULL/"" to
+// disable. `token` is sent as a Bearer header (gateway require_token).
+void npc_engine_set_gateway(const char *base_url, const char *token);
+
 // Write the active puzzle id (e.g. "SCENE_LA_DETECTOR") into `out`,
 // truncated to `cap` bytes (NUL-terminated). Falls back to "SCENE_NPC"
 // when no scene is active or the engine is not initialised so callers

idf_zacus/components/npc_engine/npc_engine.c

@@ -17,16 +17,78 @@
 #include "npc_engine.h"
 
 #include <stdio.h>
+#include <stdlib.h>
 #include <string.h>
 
 #include "esp_err.h"
+#include "esp_http_client.h"
 #include "esp_log.h"
+#include "freertos/FreeRTOS.h"
+#include "freertos/task.h"
 
 #include "media_manager.h"
 #include "hints_client.h"
 
 static const char *TAG = "npc_engine";
 
+// Voice gateway (tools/zacus-gateway) — we push the active SCENE_* to it at each
+// scene change so the phone NPCs disguise THIS scene's hint. Best-effort, set by
+// npc_engine_set_gateway(); empty = disabled (no-op, e.g. CI / dry runs).
+static char s_gw_url[128]   = {0};
+static char s_gw_token[80]  = {0};
+
+void npc_engine_set_gateway(const char *base_url, const char *token) {
+    if (base_url) snprintf(s_gw_url, sizeof(s_gw_url), "%s", base_url);
+    if (token)    snprintf(s_gw_token, sizeof(s_gw_token), "%s", token);
+}
+
+// One-shot worker: POST {gateway}/game/step?scene=<scene> then self-delete.
+// Runs on its own task so a slow/unreachable gateway NEVER stalls the scene
+// transition (the game must not lag 2 s on every scene change if the gateway is
+// offline). `arg` is a heap copy of the scene, freed here. Best-effort.
+static void gw_notify_task(void *arg) {
+    char *scene = (char *) arg;
+    char url[256];
+    snprintf(url, sizeof(url), "%s/game/step?scene=%s", s_gw_url, scene);
+    esp_http_client_config_t cfg = {
+        .url = url,
+        .method = HTTP_METHOD_POST,
+        .timeout_ms = 2000,
+    };
+    esp_http_client_handle_t cli = esp_http_client_init(&cfg);
+    if (cli) {
+        if (s_gw_token[0]) {
+            char auth[112];
+            snprintf(auth, sizeof(auth), "Bearer %s", s_gw_token);
+            esp_http_client_set_header(cli, "Authorization", auth);
+        }
+        esp_err_t err = esp_http_client_perform(cli);
+        if (err != ESP_OK) {
+            ESP_LOGW(TAG, "gateway /game/step notify failed: %s", esp_err_to_name(err));
+        } else {
+            ESP_LOGI(TAG, "gateway scene -> %s (HTTP %d)", scene,
+                     esp_http_client_get_status_code(cli));
+        }
+        esp_http_client_cleanup(cli);
+    }
+    free(scene);
+    vTaskDelete(NULL);
+}
+
+// Fire-and-forget the gateway scene notification on a detached task so a scene
+// change is never blocked by the network. No-op if no gateway configured.
+static void notify_gateway_scene(const char *scene) {
+    if (!s_gw_url[0] || !scene || !scene[0]) return;
+    size_t n = strlen(scene) + 1;
+    char *copy = malloc(n);
+    if (!copy) return;
+    memcpy(copy, scene, n);
+    if (xTaskCreate(gw_notify_task, "gw_notify", 4096, copy, 3, NULL) != pdPASS) {
+        ESP_LOGW(TAG, "gw_notify task spawn failed");
+        free(copy);
+    }
+}
+
 // ─── Core: scene/trigger/mood lookup tables (verbatim Arduino) ──────────────
 
 static const char *const kSceneIds[] = {
@@ -372,10 +434,25 @@ esp_err_t npc_engine_set_step(uint8_t step_id, uint32_t expected_duration_ms) {
                      (unsigned) prev_scene, (unsigned) step_id);
         }
         (void) hints_client_puzzle_start(puzzle_id);
+        // Keep the voice gateway in sync so the phone NPCs hint on THIS scene.
+        notify_gateway_scene(puzzle_id);
     }
     return ESP_OK;
 }
 
+esp_err_t npc_engine_set_scene_by_id(const char *scene_id,
+                                     uint32_t expected_duration_ms) {
+    if (scene_id == NULL || scene_id[0] == '\0') return ESP_ERR_INVALID_ARG;
+    for (uint8_t i = 0; i < kSceneCount; i++) {
+        if (strcmp(scene_id, kSceneIds[i]) == 0) {
+            return npc_engine_set_step(i, expected_duration_ms);
+        }
+    }
+    ESP_LOGW(TAG, "set_scene_by_id: unknown scene \"%s\" (no kSceneIds match)",
+             scene_id);
+    return ESP_ERR_NOT_FOUND;
+}
+
 esp_err_t npc_engine_set_group_profile(const char *profile) {
     // Thin pass-through. hints_client validates the value and logs the
     // outcome. Kept on npc_engine so the rest of the firmware doesn't

idf_zacus/components/ota_server/ota_server.c

@@ -310,8 +310,9 @@ esp_err_t ota_server_init(void) {
     // HTTP server config
     httpd_config_t config = HTTPD_DEFAULT_CONFIG();
     config.server_port     = OTA_SERVER_PORT;
-    config.max_uri_handlers = 20;   // ota + voice_hook + game (incl. /game/step,
-                                    // /game/puzzle_state, /game/file, relay)
+    config.max_uri_handlers = 24;   // ota + voice_hook + game (incl. /game/step,
+                                    // /game/puzzle_state, /game/file POST+DELETE,
+                                    // relay, /game/gamebook)
                                     // + headroom
     config.uri_match_fn    = httpd_uri_match_wildcard;
     config.stack_size      = 8192;

idf_zacus/main/CMakeLists.txt

@@ -23,6 +23,7 @@ idf_component_register(
         esp_event
         espressif__mdns
         display_ui
+        gamebook
         puzzle_state
         p7_coffre
         p5_morse

idf_zacus/main/main.c

@@ -55,6 +55,7 @@
 #include "mic_broker.h"
 #include "board_pins_mediakit.h"
 #include "display_ui.h"
+#include "gamebook.h"
 #include "p7_coffre.h"
 #include "p5_morse.h"
 #include "p6_nfc.h"
@@ -63,6 +64,12 @@
 // this to NVS so the field operator can repoint the firmware without a flash.
 #define ZACUS_HINTS_BASE_URL  "http://192.168.0.150:8302"
 
+// Voice gateway (tools/zacus-gateway) — the engine POSTs each scene change to
+// {URL}/game/step?scene=SCENE_* so the phone NPCs disguise that scene's hint.
+// Hardcoded for now like the URLs above — moves to NVS in the same follow-up.
+#define ZACUS_GATEWAY_BASE_URL  "http://192.168.0.175:8401"
+#define ZACUS_GATEWAY_TOKEN     "testtoken"
+
 // Slice 7: voice-bridge WebSocket on the MacStudio (Tailscale address).
 // Hardcoded here for the same reason as ZACUS_HINTS_BASE_URL — moves to
 // NVS in a follow-up slice. The bridge endpoint is documented in
@@ -430,6 +437,8 @@ void app_main(void) {
         if (btn_err != ESP_OK) {
             ESP_LOGW(TAG, "buttons_input_init: %s", esp_err_to_name(btn_err));
         }
+        // Gamebook mode owns the 5-way pad when started (POST /game/gamebook).
+        gamebook_init();
     }
 
     bool sta_ok = wifi_bring_up();
@@ -514,6 +523,13 @@ void app_main(void) {
             ESP_LOGI(TAG, "media smoke play -> %s",
                      esp_err_to_name(play_err));
 
+            // Boot into the gamebook library (the 6-story tile picker) when the
+            // SD pack is present. Best-effort: no /sdcard/gamebook/library.json
+            // -> the normal shell stays in charge.
+            if (gamebook_start() == ESP_OK) {
+                ESP_LOGI(TAG, "gamebook library up (boot picker)");
+            }
+
             // Slice 4: bring up the ported npc_engine. Cue table is empty
             // at this stage — wiring the scenario IR-driven cue catalog is
             // a follow-up slice. The engine still boots, accepts ticks
@@ -530,6 +546,9 @@ void app_main(void) {
                 ESP_LOGE(TAG, "npc_engine_init failed: %s",
                          esp_err_to_name(npc_err));
             }
+            // Push each scene change to the voice gateway so the phone NPCs
+            // disguise the CURRENT scene's hint (best-effort, see main.c URL).
+            npc_engine_set_gateway(ZACUS_GATEWAY_BASE_URL, ZACUS_GATEWAY_TOKEN);
 
             // Slice 5: bring up the hints HTTP client (so npc_engine can
             // route hint requests through the real backend) and the voice

plip_voice/main/CMakeLists.txt

@@ -13,6 +13,7 @@ idf_component_register(
         "dtmf.c"
         "turn_client.c"
         "slic.c"
+        "plip_gamebook.c"
     INCLUDE_DIRS "."
     PRIV_REQUIRES
         driver

plip_voice/main/audio.c

@@ -137,6 +137,12 @@ static void ensure_sd_mounted(void)
     }
 }
 
+bool audio_ensure_sd(void)
+{
+    ensure_sd_mounted();
+    return s_sd_mounted;
+}
+
 /* ── WAV parser ──────────────────────────────────────────────────────────── */
 
 static esp_err_t parse_wav_header(const uint8_t *buf, size_t len, wav_info_t *out)
@@ -599,66 +605,113 @@ int audio_capture_wav(uint8_t *out, size_t out_max, int max_ms, int silence_ms)
     const int max_frames     = (max_ms     / 20);
     const int silence_frames = (silence_ms / 20);
 
-    /* VAD thresholds (raw S16). Tuned for the SLIC handset mic at +24 dB PGA:
-     * speech peaks ~5 % FS / RMS ~1-2 %, noise floor ~0.6 %. Onset must sit
-     * between (catch quiet speech), silence ABOVE the noise floor (so the turn
-     * ends when the caller stops instead of running to max_ms). */
-    const int32_t  vad_onset_rms_sq  = 550 * 550;    /* ~1.7% FS² — close-talk speech above room ambient */
-    const int32_t  vad_silence_rms_sq = 400 * 400;   /* ~1.2% FS² — above room ambient so the turn ends */
+    /* Calibrated to the QUIET handset voice (DC-blocked: body 0.16-0.8 % FS RMS,
+     * loud syllables ~2-4 %, noise floor ~0.3 %). The old onset (1.4 %) NEVER
+     * triggered → "no sustained voice"; the old silence (0.6 %) sat ABOVE the
+     * body → cut mid-sentence. Now: onset catches a real syllable just above the
+     * floor (3-frame confirm guards clicks); silence sits just above the floor so
+     * the turn ends only on a true pause, holding the whole sentence. */
+    const int32_t  vad_onset_rms_sq   = 230 * 230;   /* ~0.70% FS² — quiet-voice onset */
+    const int32_t  vad_silence_rms_sq = 110 * 110;   /* ~0.34% FS² — just above noise floor */
 
-    bool voice_started = false;
     int  silent_frames = 0;
     int  total_frames  = 0;
 
+    /* One-pole DC blocker (high-pass ~80 Hz) applied to the captured mono.
+     * The SLIC/handset path carries a huge DC/sub-audio offset (~80 % FS,
+     * measured). Left in, it (a) swamps the faint voice and (b) keeps the VAD
+     * RMS permanently above the silence threshold, so the capture never ends
+     * and always runs the full max_ms — and the gateway then sees only a DC
+     * transient, transcribing empty. Removing it lets the VAD track the real
+     * speech envelope (onset/silence work) and hands the gateway a clean,
+     * voice-dominated signal. Telephone band (300-3400 Hz) is untouched.
+     * y[n] = x[n] - x[n-1] + R*y[n-1]; R=0.97 → cutoff ~80 Hz at 16 kHz. */
+    float       dc_x1 = 0.0f, dc_y1 = 0.0f;
+    const float dc_R  = 0.97f;
+
     ESP_LOGI(TAG, "capture: RX enabled, max_ms=%d silence_ms=%d", max_ms, silence_ms);
 
-    for (int f = 0; f < max_frames && pcm_written + frame_out_bytes <= pcm_max; f++) {
+    /* ── Phase A: wait for SUSTAINED voice before committing to a capture ─────
+     * Only start recording when the caller actually speaks. We require the
+     * DC-blocked RMS to stay above the onset threshold for ONSET_CONFIRM
+     * consecutive frames (~60 ms) — a single loud frame is rejected as a
+     * transient (e.g. the click when the PA mutes just before capture). Audio
+     * is discarded during this phase. If no sustained voice arrives within
+     * max_ms we return an empty WAV so the caller posts nothing (the NPC must
+     * not "reply" to silence). The listen loop simply calls us again. */
+    const int ONSET_CONFIRM = 3;
+    int  onset_run  = 0;
+    bool got_onset  = false;
+    for (int f = 0; f < max_frames; f++) {
         size_t bytes_read = 0;
         ret = i2s_channel_read(s_mic_handle, rx_buf, frame_in_bytes,
                                &bytes_read, pdMS_TO_TICKS(100));
-        if (ret != ESP_OK || bytes_read == 0) {
-            ESP_LOGW(TAG, "capture: read error f=%d: %s", f, esp_err_to_name(ret));
-            continue;
-        }
-
-        /* Downmix stereo → mono, compute RMS². */
-        int16_t *out_frame = (int16_t *)(pcm_out + pcm_written);
-        int64_t  rms_sq    = 0;
-        int      n_stereo  = (int)(bytes_read / (2 * sizeof(int16_t)));
+        if (ret != ESP_OK || bytes_read == 0) continue;
+        int64_t rms_sq = 0;
+        int     n_stereo = (int)(bytes_read / (2 * sizeof(int16_t)));
         for (int i = 0; i < n_stereo; i++) {
-            int32_t l = rx_buf[i * 2];
-            int32_t r = rx_buf[i * 2 + 1];
-            int16_t mono = (int16_t)((l + r) / 2);
-            out_frame[i] = mono;
-            rms_sq += (int64_t)mono * mono;
+            float xf = (float)((rx_buf[i * 2] + rx_buf[i * 2 + 1]) / 2);
+            float yf = xf - dc_x1 + dc_R * dc_y1;     /* DC-blocking high-pass */
+            dc_x1 = xf; dc_y1 = yf;
+            int32_t s = (int32_t)yf;
+            rms_sq += (int64_t)s * s;
         }
         rms_sq /= (n_stereo > 0 ? n_stereo : 1);
-
-        /* VAD logic. */
-        if (!voice_started) {
-            if (rms_sq >= vad_onset_rms_sq) {
-                voice_started = true;
-                silent_frames = 0;
-                ESP_LOGI(TAG, "capture: voice onset at frame %d (rms²=%"PRId64")", f, rms_sq);
+        if (rms_sq >= vad_onset_rms_sq) {
+            if (++onset_run >= ONSET_CONFIRM) {
+                got_onset = true;
+                ESP_LOGI(TAG, "capture: sustained voice onset at frame %d (rms²=%"PRId64")",
+                         f, rms_sq);
+                break;
             }
-            /* Before voice onset: still accumulate (to avoid clipping onset).
-             * But don't count towards silence timeout yet. */
         } else {
+            onset_run = 0;
+        }
+    }
+
+    if (got_onset) {
+        /* ── Phase B: record the utterance until silence_ms of silence ──────── */
+        for (int f = 0; f < max_frames && pcm_written + frame_out_bytes <= pcm_max; f++) {
+            size_t bytes_read = 0;
+            ret = i2s_channel_read(s_mic_handle, rx_buf, frame_in_bytes,
+                                   &bytes_read, pdMS_TO_TICKS(100));
+            if (ret != ESP_OK || bytes_read == 0) {
+                ESP_LOGW(TAG, "capture: read error f=%d: %s", f, esp_err_to_name(ret));
+                continue;
+            }
+
+            /* Downmix stereo → mono (DC-blocked), compute RMS². */
+            int16_t *out_frame = (int16_t *)(pcm_out + pcm_written);
+            int64_t  rms_sq    = 0;
+            int      n_stereo  = (int)(bytes_read / (2 * sizeof(int16_t)));
+            for (int i = 0; i < n_stereo; i++) {
+                float   xf = (float)((rx_buf[i * 2] + rx_buf[i * 2 + 1]) / 2);
+                float   yf = xf - dc_x1 + dc_R * dc_y1;
+                dc_x1 = xf; dc_y1 = yf;
+                if (yf > 32767.0f)       yf = 32767.0f;
+                else if (yf < -32768.0f) yf = -32768.0f;
+                int16_t mono = (int16_t)yf;
+                out_frame[i] = mono;
+                rms_sq += (int64_t)mono * mono;
+            }
+            rms_sq /= (n_stereo > 0 ? n_stereo : 1);
+
+            pcm_written += (size_t)n_stereo * sizeof(int16_t);
+            total_frames = f + 1;
+
             if (rms_sq < vad_silence_rms_sq) {
-                silent_frames++;
-                if (silent_frames >= silence_frames) {
-                    total_frames = f + 1;
-                    ESP_LOGI(TAG, "capture: VAD end (silence %d frames at f=%d)", silent_frames, f);
-                    pcm_written += (size_t)n_stereo * sizeof(int16_t);
+                if (++silent_frames >= silence_frames) {
+                    ESP_LOGI(TAG, "capture: VAD end (silence %d frames at f=%d)",
+                             silent_frames, f);
                     break;
                 }
             } else {
                 silent_frames = 0;
             }
         }
-
-        pcm_written += (size_t)n_stereo * sizeof(int16_t);
-        total_frames = f + 1;
+    } else {
+        ESP_LOGI(TAG, "capture: no sustained voice in %d ms — nothing to send", max_ms);
+        /* pcm_written stays 0 → empty WAV → caller skips posting. */
     }
 
     /* Leave RX enabled (full-duplex, as at boot). Disabling it here would make

plip_voice/main/audio.h

@@ -39,6 +39,12 @@ esp_err_t audio_play_async(const char *path);
 /* Stop current playback immediately. */
 esp_err_t audio_stop(void);
 
+/* Mount the microSD card at /sdcard on demand (best-effort, idempotent) and
+ * report whether it is mounted. Unlike playback, this does NOT depend on the
+ * hook state — it lets the REST layer stage a voice pack onto the SD while the
+ * handset is down. Returns true if /sdcard is usable. */
+bool audio_ensure_sd(void);
+
 /* True while the audio worker is playing a clip (tone/WAV). The conversation
  * LISTEN loop polls this to stay half-duplex: never capture while playing
  * (avoids the earpiece→mic feedback that saturated the line). */

plip_voice/main/conversation.c

@@ -23,6 +23,7 @@
 #include "phone.h"
 #include "slic.h"
 #include "turn_client.h"
+#include "plip_gamebook.h"
 #if CONFIG_PLIP_DIAL_DTMF
 #include "dtmf.h"
 #endif
@@ -47,7 +48,9 @@
 #define CAPTURE_MAX_IRAM    (128 * 1024)
 #define CAPTURE_MAX_MS_PSRAM  8000
 #define CAPTURE_MAX_MS_IRAM   4000
-#define CAPTURE_SILENCE_MS     600   /* end-of-speech VAD: shorter = snappier reply (was 800) */
+#define CAPTURE_SILENCE_MS     900   /* end-of-speech: needs to ride through brief
+                                      * mid-sentence pauses so a whole phrase is
+                                      * captured (quiet handset voice). 900 ms. */
 #define REPLY_POLL_MS          200   /* interval for checking hook during playback */
 #define REPLY_PLAYBACK_EXTRA_MS 500  /* safety margin added to computed WAV duration */
 #define BETWEEN_TURNS_MS       300   /* short pause between capture rounds */
@@ -66,6 +69,7 @@ typedef enum {
     STATE_BUSY,
     STATE_GREET,       /* fetching + playing NPC greeting (Stage 2) */
     STATE_CONNECTED,   /* in-call — Stage 3 will add listen loop     */
+    STATE_GAMEBOOK,    /* audio "livre dont vous êtes le héros"      */
 } conv_state_t;
 
 static volatile conv_state_t s_state    = STATE_IDLE;
@@ -82,12 +86,16 @@ static char s_sid[32] = {0};
  * NOT re-read dialer_current() for the greeting/reply — that polluted number
  * would 404 at the gateway. Capture the clean routed number here once. */
 static char s_number[16] = {0};
+/* Scene this call hints on (e.g. "SCENE_WARNING"), locked at pickup like the
+ * number. Empty for player-dialled calls. Drives the local SD hint clip. */
+static char s_scene[40] = {0};
 
 /* Incoming-call mode: an NPC "calls" the phone. conversation_arm_incoming()
- * stores the caller's persona number and rings; when the handset is then picked
- * up from IDLE, we skip the outgoing dialtone/dial/ringback and go straight to
- * GREET with this number (the NPC speaks first). */
+ * stores the caller's persona number (and optional scene) and rings; when the
+ * handset is then picked up from IDLE, we skip the outgoing dialtone/dial/
+ * ringback and go straight to GREET with this number (the NPC speaks first). */
 static char s_incoming_number[16] = {0};
+static char s_incoming_scene[40] = {0};
 static volatile bool s_incoming_armed = false;
 
 /* Known numbers: ringback when dialed */
@@ -103,15 +111,32 @@ static bool is_known(const char *num)
     return false;
 }
 
+/* "SCENE_WARNING" → "warning": strip the SCENE_ prefix and lowercase, matching
+ * the SD pack filenames from tools/tts/generate_plip_sd_pack.py. */
+static void scene_slug(const char *scene, char *out, size_t cap)
+{
+    const char *p = scene;
+    if (strncmp(p, "SCENE_", 6) == 0) p += 6;
+    size_t i = 0;
+    for (; p[i] && i + 1 < cap; i++) {
+        char c = p[i];
+        out[i] = (c >= 'A' && c <= 'Z') ? (char)(c - 'A' + 'a') : c;
+    }
+    out[i] = '\0';
+}
+
 static void go_idle(void)
 {
     tones_stop();
     audio_stop();
     audio_pa_set(false);
     dialer_reset();
+    plip_gamebook_end();   /* no-op if the gamebook wasn't running */
 #if CONFIG_PLIP_DIAL_DTMF
     dtmf_stop();
 #endif
+    s_scene[0] = '\0';   /* clear scripted-call scene so a later dial-out call
+                          * doesn't inherit it (CONNECTED would skip listening). */
     s_state = STATE_IDLE;
     ESP_LOGI(TAG, "-> IDLE");
 }
@@ -124,10 +149,15 @@ static void enter_incoming_greet(void)
 {
     s_incoming_armed = false;
     phone_ring_stop();
+    slic_ring_stop();   /* belt-and-suspenders: kill the physical bell directly in
+                         * case phone.c's s_ringing flag desynced from slic.c's
+                         * (otherwise phone_ring_stop early-returns and the bell
+                         * keeps ringing through the answered call). */
     tones_stop();
     dialer_reset();
     s_offhook = true;
     snprintf(s_number, sizeof(s_number), "%s", s_incoming_number);
+    snprintf(s_scene, sizeof(s_scene), "%s", s_incoming_scene);
     snprintf(s_sid, sizeof(s_sid), "%lld", (long long)esp_timer_get_time());
     audio_pa_set(true);
     s_state = STATE_GREET;
@@ -158,6 +188,17 @@ static void conv_task(void *arg)
                     if (s_incoming_armed) {
                         /* INCOMING call answered (phone.c detected the edge). */
                         enter_incoming_greet();
+                    } else if (plip_gamebook_available()) {
+                        /* Gamebook phone: a story pack is on the SD → pick up
+                         * launches the audio "livre dont vous êtes le héros".
+                         * Digits dialed choose the story then the answers. */
+                        dialer_reset();
+#if CONFIG_PLIP_DIAL_DTMF
+                        dtmf_start();
+#endif
+                        plip_gamebook_begin();
+                        s_state = STATE_GAMEBOOK;
+                        ESP_LOGI(TAG, "off-hook -> GAMEBOOK");
                     } else {
                         /* Outgoing call: dial tone, wait for digits. */
                         dialer_reset();
@@ -197,6 +238,21 @@ static void conv_task(void *arg)
             }
             break;
 
+        case STATE_GAMEBOOK:
+            if (!s_offhook) {
+                go_idle();
+                break;
+            }
+            /* Each dialed digit drives one choice. Consume a single digit at a
+             * time (take the first, reset) so the player can dial again right
+             * away — even over the narration, which gets interrupted. */
+            if (!dialer_idle()) {
+                int d = dialer_current()[0] - '0';
+                dialer_reset();
+                plip_gamebook_feed_digit(d);
+            }
+            break;
+
         case STATE_DIALING:
             if (!s_offhook) {
                 go_idle();
@@ -251,18 +307,57 @@ static void conv_task(void *arg)
                 go_idle();
                 break;
             }
-            /* Fetch greeting WAV from gateway and enqueue playback */
-            if (turn_client_greeting(s_sid, s_number,
-                                     "/spiffs/turn.wav")) {
-                audio_play_async("/spiffs/turn.wav");
-            } else {
-                ESP_LOGW(TAG, "turn_client_greeting failed — proceeding silent");
+            /* Prefer the local SD voice pack (no gateway, no model in RAM):
+             * play /sdcard/voice/greet_<number>.wav when present. Fall back to
+             * the live gateway greeting only when the pack has no clip for this
+             * NPC number (see tools/tts/generate_plip_sd_pack.py). */
+            {
+                char sd_greet[64];
+                snprintf(sd_greet, sizeof(sd_greet),
+                         "/sdcard/voice/greet_%s.wav", s_number);
+                FILE *tf = NULL;
+                if (audio_ensure_sd() && (tf = fopen(sd_greet, "rb")) != NULL) {
+                    fclose(tf);
+                    audio_play_async(sd_greet);
+                    ESP_LOGI(TAG, "GREET: local SD pack %s", sd_greet);
+                } else if (turn_client_greeting(s_sid, s_number,
+                                                "/spiffs/turn.wav")) {
+                    audio_play_async("/spiffs/turn.wav");
+                } else {
+                    ESP_LOGW(TAG, "no SD clip + gateway greeting failed — silent");
+                }
+            }
+            /* Scripted offline call: chain this scene's hint from the SD pack
+             * right after the greeting (the audio queue plays them back-to-back).
+             * Fully local — no gateway, no model. */
+            if (s_scene[0] != '\0' && audio_ensure_sd()) {
+                char slug[40], hint[80];
+                scene_slug(s_scene, slug, sizeof(slug));
+                snprintf(hint, sizeof(hint),
+                         "/sdcard/voice/hint_%s_l1_0.wav", slug);
+                FILE *hf = fopen(hint, "rb");
+                if (hf != NULL) {
+                    fclose(hf);
+                    audio_play_async(hint);
+                    ESP_LOGI(TAG, "GREET: chained SD hint %s", hint);
+                } else {
+                    ESP_LOGI(TAG, "GREET: no SD hint for scene %s (%s)",
+                             s_scene, hint);
+                }
             }
             s_state = STATE_CONNECTED;
             ESP_LOGI(TAG, "-> CONNECTED");
             break;
 
         case STATE_CONNECTED:
+            /* Scripted offline call (greeting + SD hint already queued): there's
+             * no live two-way conversation to run — just let the clips play out
+             * and wait for the player to hang up. Skips the gateway/model listen
+             * loop entirely. The global on-hook check returns us to IDLE. */
+            if (s_scene[0] != '\0') {
+                vTaskDelay(pdMS_TO_TICKS(REPLY_POLL_MS));
+                break;
+            }
 #if CONFIG_PLIP_VOICE_REPLY
             /*
              * Stage 3 — LISTEN loop.
@@ -313,17 +408,13 @@ static void conv_task(void *arg)
                     if (!s_offhook) break;
                     vTaskDelay(pdMS_TO_TICKS(200));   /* let the I2S DMA tail drain */
 
-                    /* Kill the earpiece amp during capture: at full volume + 24 dB
-                     * mic gain the playback couples into the handset mic at ~50 %
-                     * FS and swamps the caller's voice (capture transcribed empty).
-                     * PA off = no echo path; restored before the reply plays. */
-                    audio_pa_set(false);
-                    vTaskDelay(pdMS_TO_TICKS(60));   /* PA mute settle */
-
-                    /* Capture caller utterance — earpiece muted, nothing plays. */
+                    /* Capture the caller utterance with the PA LEFT ON. The loop
+                     * already waits for any playback to finish above, so nothing
+                     * is driving the earpiece during capture — there is no echo to
+                     * mute. Cutting the PA here was found to collapse the mic level
+                     * (captures came back near-silent / DC only), so keep it on. */
                     int n = audio_capture_wav(cap_buf, cap_max,
                                               cap_ms, CAPTURE_SILENCE_MS);
-                    audio_pa_set(true);              /* restore for filler/reply */
                     if (!s_offhook) break;   /* hung up during capture */
 
                     if (n <= 44) {
@@ -402,12 +493,14 @@ void conversation_on_hook_change(bool offhook)
     s_hook_changed = true;
 }
 
-void conversation_arm_incoming(const char *number)
+void conversation_arm_incoming(const char *number, const char *scene)
 {
     snprintf(s_incoming_number, sizeof(s_incoming_number), "%s",
              (number && number[0]) ? number : "17");
+    snprintf(s_incoming_scene, sizeof(s_incoming_scene), "%s",
+             (scene && scene[0]) ? scene : "");
     s_incoming_armed = true;
     phone_ring_start();
-    ESP_LOGI(TAG, "incoming call armed (num=%s) — ringing until pickup",
-             s_incoming_number);
+    ESP_LOGI(TAG, "incoming call armed (num=%s scene=%s) — ringing until pickup",
+             s_incoming_number, s_incoming_scene[0] ? s_incoming_scene : "-");
 }

plip_voice/main/conversation.h

@@ -6,5 +6,8 @@ void conversation_on_hook_change(bool offhook);
 
 /* Arm an INCOMING call from NPC persona `number` and start ringing. When the
  * handset is picked up from idle, the conversation jumps straight to the
- * greeting (persona speaks first) instead of the outgoing dial-tone path. */
-void conversation_arm_incoming(const char *number);
+ * greeting (persona speaks first) instead of the outgoing dial-tone path.
+ * `scene` (optional, e.g. "SCENE_WARNING") makes the phone play that scene's
+ * hint from its local SD pack right after the greeting; pass NULL/"" for a
+ * greeting-only call. */
+void conversation_arm_incoming(const char *number, const char *scene);

plip_voice/main/net.c

@@ -22,8 +22,10 @@
 #include <string.h>
 #include <stdio.h>
 #include <errno.h>
+#include <sys/stat.h>
 
 #include "audio.h"
+#include "tones.h"
 #include "cmd_exec.h"
 #include "phone.h"
 #include "conversation.h"
@@ -52,7 +54,9 @@
 
 #define SPIFFS_LABEL  "storage"
 #define SPIFFS_BASE   "/spiffs"
-#define MAX_BODY      (256 * 1024)
+#define MAX_BODY      (3 * 1024 * 1024) /* streamed in 2 KB chunks → RAM-safe;
+                                          raised to fit full-length neural-TTS
+                                          (Kyutai) gamebook narration WAVs */
 
 static volatile bool    s_connected  = false;
 static httpd_handle_t   s_httpd      = NULL;
@@ -208,14 +212,39 @@ static esp_err_t handle_file_post(httpd_req_t *req)
     if (req->content_len <= 0 || req->content_len > MAX_BODY)
         return send_json(req, "413 Payload Too Large", "{\"error\":\"too large\"}");
 
-    ensure_spiffs();
+    /* Route to the microSD when the path is sd-prefixed ("sd/…", "sdcard/…"
+     * or "/sdcard/…"); otherwise stage in SPIFFS (default). The SD is mounted
+     * on demand here so a voice pack can be staged without an off-hook play. */
+    const char *base;
+    const char *rel = fpath;
+    bool to_sd = false;
+    if      (strncmp(fpath, "sd/", 3) == 0)       { rel = fpath + 3; to_sd = true; }
+    else if (strncmp(fpath, "sdcard/", 7) == 0)   { rel = fpath + 7; to_sd = true; }
+    else if (strncmp(fpath, "/sdcard/", 8) == 0)  { rel = fpath + 8; to_sd = true; }
 
-    /* Build full SPIFFS path. */
-    char full[160];
-    if (fpath[0] == '/') {
-        snprintf(full, sizeof(full), "%s%s", SPIFFS_BASE, fpath);
+    if (to_sd) {
+        if (!audio_ensure_sd())
+            return send_json(req, "503 Service Unavailable", "{\"error\":\"no sd card\"}");
+        base = PLIP_SD_MOUNT;   /* "/sdcard" */
     } else {
-        snprintf(full, sizeof(full), "%s/%s", SPIFFS_BASE, fpath);
+        ensure_spiffs();
+        base = SPIFFS_BASE;
+    }
+    while (*rel == '/') rel++;   /* avoid a double slash when joining */
+
+    char full[192];
+    snprintf(full, sizeof(full), "%s/%s", base, rel);
+
+    /* Create the immediate parent dir (one level) so a pack can live in a
+     * subfolder like /sdcard/voice/<file>.wav — fopen won't create it. FAT
+     * supports mkdir; SPIFFS is flat so we only do this for the SD path. */
+    if (to_sd) {
+        char *slash = strrchr(full, '/');
+        if (slash && slash != full) {
+            *slash = '\0';
+            mkdir(full, 0775);   /* best-effort; ignore EEXIST */
+            *slash = '/';
+        }
     }
 
     FILE *f = fopen(full, "wb");
@@ -253,6 +282,44 @@ static esp_err_t handle_file_post(httpd_req_t *req)
     return send_json(req, "200 OK", resp);
 }
 
+/* ── DELETE /game/file?path=...  (remove a staged file; SD or SPIFFS) ─────── */
+/* Same path routing as the POST handler (sd/ prefix → /sdcard, else /spiffs).
+ * Lets the host clean stale clips when regenerating the SD voice pack. */
+static esp_err_t handle_file_delete(httpd_req_t *req)
+{
+    char query[128] = {0}, fpath[128] = {0};
+    httpd_req_get_url_query_str(req, query, sizeof(query));
+    if (httpd_query_key_value(query, "path", fpath, sizeof(fpath)) != ESP_OK)
+        return send_json(req, "400 Bad Request", "{\"error\":\"missing path param\"}");
+
+    const char *base, *rel = fpath;
+    bool to_sd = false;
+    if      (strncmp(fpath, "sd/", 3) == 0)       { rel = fpath + 3; to_sd = true; }
+    else if (strncmp(fpath, "sdcard/", 7) == 0)   { rel = fpath + 7; to_sd = true; }
+    else if (strncmp(fpath, "/sdcard/", 8) == 0)  { rel = fpath + 8; to_sd = true; }
+
+    if (to_sd) {
+        if (!audio_ensure_sd())
+            return send_json(req, "503 Service Unavailable", "{\"error\":\"no sd card\"}");
+        base = PLIP_SD_MOUNT;
+    } else {
+        ensure_spiffs();
+        base = SPIFFS_BASE;
+    }
+    while (*rel == '/') rel++;
+
+    char full[192];
+    snprintf(full, sizeof(full), "%s/%s", base, rel);
+    if (remove(full) != 0) {
+        ESP_LOGW(TAG, "remove %s failed: errno=%d", full, errno);
+        return send_json(req, "404 Not Found", "{\"error\":\"remove failed\"}");
+    }
+    ESP_LOGI(TAG, "file removed: %s", full);
+    char resp[256];
+    snprintf(resp, sizeof(resp), "{\"status\":\"ok\",\"removed\":\"%s\"}", full);
+    return send_json(req, "200 OK", resp);
+}
+
 /* ── POST /voice/capture ──────────────────────────────────────────────────── */
 
 #define CAP_MAX_MS_DEFAULT      5000
@@ -482,10 +549,13 @@ static esp_err_t handle_debug_ring(httpd_req_t *req)
 {
     /* /debug/ring[?number=NN] — arm an INCOMING call from persona NN (default
      * 17) and ring. On pickup the conversation jumps straight to the greeting. */
-    char query[48] = {0}, number[16] = {0};
+    char query[96] = {0}, number[16] = {0}, scene[40] = {0};
     httpd_req_get_url_query_str(req, query, sizeof(query));
     httpd_query_key_value(query, "number", number, sizeof(number));
-    conversation_arm_incoming(number);  /* defaults to "17" if empty; starts ringing */
+    httpd_query_key_value(query, "scene", scene, sizeof(scene));
+    /* scene (optional) lets the phone play that scene's hint from its SD pack
+     * after the greeting — fully offline, no model. */
+    conversation_arm_incoming(number, scene[0] ? scene : NULL);
     char resp[64];
     snprintf(resp, sizeof(resp), "{\"ok\":true,\"ringing\":true,\"incoming\":\"%s\"}",
              number[0] ? number : "17");
@@ -600,6 +670,61 @@ static esp_err_t handle_debug_dacvol(httpd_req_t *req)
     return send_json(req, "200 OK", resp);
 }
 
+/* ── GET /debug/miccap?ms=4000  (TEMP DIAG: raw fixed-duration mic capture) ───
+ * No VAD, DC-blocked, PA on, tones stopped. Lets us measure the true handset
+ * mic level while the caller speaks continuously, independent of VAD timing. */
+static esp_err_t handle_debug_miccap(httpd_req_t *req)
+{
+    char query[64] = {0};
+    httpd_req_get_url_query_str(req, query, sizeof(query));
+    char val[16] = {0};
+    int ms = 4000;
+    if (httpd_query_key_value(query, "ms", val, sizeof(val)) == ESP_OK) {
+        int v = atoi(val);
+        if (v > 0 && v <= 8000) ms = v;
+    }
+    tones_stop();
+    audio_pa_set(true);
+    if (audio_capture_begin(ms, ms) != 0) {
+        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "capture init failed");
+        return ESP_FAIL;
+    }
+    httpd_resp_set_type(req, "audio/wav");
+    const uint32_t cap_rate = 16000; const uint16_t cap_ch = 1, cap_bits = 16;
+    const uint32_t byte_rate = cap_rate * cap_ch * (cap_bits / 8);
+    const uint16_t block_align = (uint16_t)(cap_ch * (cap_bits / 8));
+    const uint32_t big = 0xFFFFFFFFu; uint16_t fmt_pcm = 1; uint32_t fmt_size = 16;
+    uint8_t hdr[44];
+    memcpy(hdr, "RIFF", 4);  memcpy(hdr + 4, &big, 4);
+    memcpy(hdr + 8, "WAVE", 4);  memcpy(hdr + 12, "fmt ", 4);
+    memcpy(hdr + 16, &fmt_size, 4);  memcpy(hdr + 20, &fmt_pcm, 2);
+    memcpy(hdr + 22, &cap_ch, 2);  memcpy(hdr + 24, &cap_rate, 4);
+    memcpy(hdr + 28, &byte_rate, 4);  memcpy(hdr + 32, &block_align, 2);
+    memcpy(hdr + 34, &cap_bits, 2);  memcpy(hdr + 36, "data", 4);  memcpy(hdr + 40, &big, 4);
+    httpd_resp_send_chunk(req, (const char *)hdr, sizeof(hdr));
+
+    const int max_frames = ms / 20;
+    int16_t mono[320];
+    float dc_x1 = 0.0f, dc_y1 = 0.0f; const float dc_R = 0.97f;
+    for (int f = 0; f < max_frames; f++) {
+        int64_t rms_sq = 0;
+        int n = audio_capture_read_frame(mono, 320, &rms_sq);
+        if (n < 0) break;
+        if (n == 0) continue;
+        for (int i = 0; i < n; i++) {
+            float xf = (float)mono[i];
+            float yf = xf - dc_x1 + dc_R * dc_y1;
+            dc_x1 = xf; dc_y1 = yf;
+            if (yf > 32767.0f) yf = 32767.0f; else if (yf < -32768.0f) yf = -32768.0f;
+            mono[i] = (int16_t)yf;
+        }
+        if (httpd_resp_send_chunk(req, (const char *)mono, (ssize_t)(n * sizeof(int16_t))) != ESP_OK) break;
+    }
+    audio_capture_end();
+    httpd_resp_send_chunk(req, NULL, 0);
+    return ESP_OK;
+}
+
 /* ── GET /debug/getfile?path=/x.wav  (read a SPIFFS file back for diagnosis) ── */
 
 static esp_err_t handle_debug_getfile(httpd_req_t *req)
@@ -756,7 +881,7 @@ esp_err_t net_init(void)
     /* Start HTTP server. */
     httpd_config_t hcfg = HTTPD_DEFAULT_CONFIG();
     hcfg.server_port      = 80;
-    hcfg.max_uri_handlers = 16;
+    hcfg.max_uri_handlers = 18;
     hcfg.stack_size       = 8192;
 
     esp_err_t ret = httpd_start(&s_httpd, &hcfg);
@@ -777,6 +902,10 @@ esp_err_t net_init(void)
         .uri = "/game/file", .method = HTTP_POST,
         .handler = handle_file_post, .user_ctx = NULL,
     };
+    static const httpd_uri_t uri_file_delete = {
+        .uri = "/game/file", .method = HTTP_DELETE,
+        .handler = handle_file_delete, .user_ctx = NULL,
+    };
     static const httpd_uri_t uri_cmd = {
         .uri = "/game/cmd", .method = HTTP_POST,
         .handler = handle_cmd_post, .user_ctx = NULL,
@@ -801,6 +930,10 @@ esp_err_t net_init(void)
         .uri = "/debug/getfile", .method = HTTP_GET,
         .handler = handle_debug_getfile, .user_ctx = NULL,
     };
+    static const httpd_uri_t uri_debug_miccap = {
+        .uri = "/debug/miccap", .method = HTTP_GET,
+        .handler = handle_debug_miccap, .user_ctx = NULL,
+    };
     static const httpd_uri_t uri_debug_dacvol = {
         .uri = "/debug/dacvol", .method = HTTP_GET,
         .handler = handle_debug_dacvol, .user_ctx = NULL,
@@ -828,12 +961,14 @@ esp_err_t net_init(void)
     httpd_register_uri_handler(s_httpd, &uri_status);
     httpd_register_uri_handler(s_httpd, &uri_scenario);
     httpd_register_uri_handler(s_httpd, &uri_file);
+    httpd_register_uri_handler(s_httpd, &uri_file_delete);
     httpd_register_uri_handler(s_httpd, &uri_cmd);
     httpd_register_uri_handler(s_httpd, &uri_capture);
     httpd_register_uri_handler(s_httpd, &uri_debug_regs);
     httpd_register_uri_handler(s_httpd, &uri_debug_dial);
     httpd_register_uri_handler(s_httpd, &uri_debug_vol);
     httpd_register_uri_handler(s_httpd, &uri_debug_getfile);
+    httpd_register_uri_handler(s_httpd, &uri_debug_miccap);
     httpd_register_uri_handler(s_httpd, &uri_debug_dacvol);
     httpd_register_uri_handler(s_httpd, &uri_debug_ring);
     httpd_register_uri_handler(s_httpd, &uri_debug_ringstop);

plip_voice/main/phone.c

@@ -34,20 +34,23 @@
 #define TAG "phone"
 
 #define DEBOUNCE_MS          30
-#define HANGUP_THRESHOLD_MS  1500  /* open > this → real hangup, not a pulse.
-                                    * Raised from 500 ms: the A1S cradle contact
-                                    * is marginal and produces ~500 ms spurious
-                                    * opens that were resetting the call before
-                                    * the NPC greeting. A real hangup holds the
-                                    * line open indefinitely, so it still fires
-                                    * (~1.5 s later). Rotary pulses (~60 ms) and
-                                    * closed inter-digit gaps are unaffected. */
+#define HANGUP_THRESHOLD_MS  2500  /* open > this → real hangup, not a pulse/flicker.
+                                    * HOOK DEBOUNCE: the A1S cradle contact is
+                                    * marginal and flickers open for up to ~2 s
+                                    * mid-call; treating those as a hangup dropped
+                                    * the live conversation. Requiring the line to
+                                    * stay open ≥ 2.5 s before hanging up rides
+                                    * through the flickers. A real hangup holds the
+                                    * line open indefinitely so it still fires
+                                    * (~2.5 s later). Rotary pulses (~60 ms) and a
+                                    * closed line are unaffected. */
 #define TASK_STACK           4096
 #define TASK_PRIO            5
-#define RESYNC_STABLE_MS     600   /* if the raw hook level stably disagrees with
-                                    * s_offhook for this long (no pulse), the edge
-                                    * detection missed/flapped a transition →
-                                    * self-correct s_offhook to the physical level */
+#define RESYNC_PICKUP_MS     600   /* off-hook (pickup) self-correct: fast, so an
+                                    * incoming call answers promptly. */
+#define RESYNC_HANGUP_MS     HANGUP_THRESHOLD_MS  /* on-hook (hangup) self-correct:
+                                    * same long debounce as the prolonged-open path
+                                    * so a flickering contact never drops the call. */
 
 /* Rotary pulse hardening (CONFIG_PLIP_DIAL_PULSE) */
 #define PULSE_MIN_WIDTH_MS   20    /* ignore opens shorter than this (glitch filter) */
@@ -316,7 +319,11 @@ poll_sleep:
             bool phys_off = (gpio_get_level(hook_gpio) == HOOK_OFFHOOK_LEVEL);
             if (!pulse_active && phys_off != s_offhook) {
                 resync_ms += 10;
-                if (resync_ms >= RESYNC_STABLE_MS) {
+                /* Hook debounce: a PICKUP (→off-hook) is confirmed fast so calls
+                 * answer promptly; a HANGUP (→on-hook) needs the long debounce so
+                 * a flickering cradle contact can't drop a live call. */
+                int need = phys_off ? RESYNC_PICKUP_MS : RESYNC_HANGUP_MS;
+                if (resync_ms >= need) {
                     ESP_LOGW(TAG, "hook resync: physical=%s but s_offhook=%d — correcting",
                              phys_off ? "off-hook" : "on-hook", (int)s_offhook);
                     s_offhook  = phys_off;

plip_voice/main/plip_gamebook.c

@@ -0,0 +1,171 @@
+// plip_gamebook.c — see plip_gamebook.h. Audio CYOA for the PLIP phone.
+#include "plip_gamebook.h"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "cJSON.h"
+#include "esp_log.h"
+
+#include "audio.h"
+
+static const char *TAG = "plip_gb";
+
+#define GB_DIR      "/sdcard/gamebook"
+#define GB_LIBRARY  GB_DIR "/library.json"
+#define GB_MENU_WAV GB_DIR "/menu.wav"
+#define GB_JSON_MAX (64 * 1024)
+
+typedef enum { GB_OFF, GB_MENU, GB_STORY } gb_mode_t;
+
+static cJSON   *s_lib_root  = NULL;   // owns library.json
+static cJSON   *s_lib       = NULL;   // borrowed: root->"library"
+static int      s_lib_n     = 0;
+
+static cJSON   *s_book      = NULL;   // owns the current <id>.json
+static cJSON   *s_passages  = NULL;   // borrowed: book->"passages"
+static char     s_current[48] = {0};
+
+static gb_mode_t s_mode = GB_OFF;
+
+bool plip_gamebook_active(void) { return s_mode != GB_OFF; }
+
+static cJSON *load_json(const char *path)
+{
+    FILE *f = fopen(path, "rb");
+    if (!f) { ESP_LOGW(TAG, "open %s failed", path); return NULL; }
+    fseek(f, 0, SEEK_END);
+    long sz = ftell(f);
+    rewind(f);
+    if (sz <= 0 || sz > GB_JSON_MAX) { fclose(f); return NULL; }
+    char *buf = malloc((size_t)sz + 1);
+    if (!buf) { fclose(f); return NULL; }
+    size_t rd = fread(buf, 1, (size_t)sz, f);
+    fclose(f);
+    buf[rd] = '\0';
+    cJSON *root = cJSON_Parse(buf);
+    free(buf);
+    if (!root) ESP_LOGW(TAG, "malformed JSON: %s", path);
+    return root;
+}
+
+static void play_wav(const char *file)   // file = bare name under GB_DIR
+{
+    char path[96];
+    snprintf(path, sizeof(path), "%s/%s", GB_DIR, file);
+    audio_stop();                 // interrupt any current narration
+    audio_play_async(path);
+}
+
+bool plip_gamebook_available(void)
+{
+    if (!audio_ensure_sd()) return false;
+    FILE *f = fopen(GB_LIBRARY, "rb");
+    if (!f) return false;
+    fclose(f);
+    return true;
+}
+
+static void free_story(void)
+{
+    if (s_book) { cJSON_Delete(s_book); s_book = NULL; }
+    s_passages = NULL;
+    s_current[0] = '\0';
+}
+
+static void enter_menu(void)
+{
+    free_story();
+    s_mode = GB_MENU;
+    audio_stop();
+    audio_play_async(GB_MENU_WAV);   // absolute path
+    ESP_LOGI(TAG, "menu (%d stories)", s_lib_n);
+}
+
+// Play a passage by id (its WAV narrates the text + numbered choices).
+static void enter_passage(const char *pid)
+{
+    const cJSON *p = cJSON_GetObjectItem(s_passages, pid);
+    if (!cJSON_IsObject(p)) { ESP_LOGW(TAG, "passage '%s' missing", pid); return; }
+    snprintf(s_current, sizeof(s_current), "%s", pid);
+    const cJSON *wav = cJSON_GetObjectItem(p, "wav");
+    if (cJSON_IsString(wav) && wav->valuestring[0]) play_wav(wav->valuestring);
+    const cJSON *ch = cJSON_GetObjectItem(p, "choices");
+    ESP_LOGI(TAG, "passage '%s' (%d choices)", pid,
+             cJSON_IsArray(ch) ? cJSON_GetArraySize(ch) : 0);
+}
+
+static void load_book(int idx)
+{
+    const cJSON *e = cJSON_GetArrayItem(s_lib, idx);
+    const cJSON *file = cJSON_GetObjectItem(e, "book");
+    if (!cJSON_IsString(file)) return;
+    char path[128];
+    snprintf(path, sizeof(path), "%s/%s", GB_DIR, file->valuestring);
+    cJSON *book = load_json(path);
+    if (!book) return;
+    const cJSON *passages = cJSON_GetObjectItem(book, "passages");
+    const cJSON *start = cJSON_GetObjectItem(book, "start");
+    if (!cJSON_IsObject(passages) || !cJSON_IsString(start)) {
+        cJSON_Delete(book); return;
+    }
+    free_story();
+    s_book = book;
+    s_passages = (cJSON *)passages;
+    s_mode = GB_STORY;
+    ESP_LOGI(TAG, "load book #%d @ '%s'", idx, start->valuestring);
+    enter_passage(start->valuestring);
+}
+
+void plip_gamebook_begin(void)
+{
+    plip_gamebook_end();   // clean any previous run
+
+    s_lib_root = load_json(GB_LIBRARY);
+    if (!s_lib_root) return;
+    s_lib = cJSON_GetObjectItem(s_lib_root, "library");
+    if (!cJSON_IsArray(s_lib) || cJSON_GetArraySize(s_lib) == 0) {
+        cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL;
+        return;
+    }
+    s_lib_n = cJSON_GetArraySize(s_lib);
+    audio_pa_set(true);    // earpiece amplifier on
+    enter_menu();
+}
+
+void plip_gamebook_feed_digit(int d)
+{
+    if (s_mode == GB_MENU) {
+        if (d >= 1 && d <= s_lib_n) load_book(d - 1);
+        return;
+    }
+    if (s_mode == GB_STORY) {
+        const cJSON *p = cJSON_GetObjectItem(s_passages, s_current);
+        const cJSON *ch = cJSON_GetObjectItem(p, "choices");
+        int n = cJSON_IsArray(ch) ? cJSON_GetArraySize(ch) : 0;
+        if (n == 0) {                 // ending: 0 (or any) → back to menu
+            enter_menu();
+            return;
+        }
+        if (d >= 1 && d <= n) {
+            const cJSON *g = cJSON_GetObjectItem(cJSON_GetArrayItem(ch, d - 1),
+                                                 "goto");
+            if (cJSON_IsString(g)) {
+                ESP_LOGI(TAG, "choice %d -> '%s'", d, g->valuestring);
+                enter_passage(g->valuestring);
+            }
+        }
+        // out-of-range digit: ignore (player can dial again)
+    }
+}
+
+void plip_gamebook_end(void)
+{
+    if (s_mode == GB_OFF && !s_lib_root && !s_book) return;
+    s_mode = GB_OFF;
+    audio_stop();
+    free_story();
+    if (s_lib_root) { cJSON_Delete(s_lib_root); s_lib_root = NULL; s_lib = NULL; }
+    ESP_LOGI(TAG, "ended");
+}

plip_voice/main/plip_gamebook.h

@@ -0,0 +1,44 @@
+#pragma once
+// plip_gamebook — "livre dont vous êtes le héros" AUDIO pour le téléphone PLIP.
+//
+// Pas d'écran : tout passe par l'écouteur. On décroche, le narrateur lit le
+// menu des histoires ("pour le dragon, faites le 1..."), on compose un chiffre
+// au cadran pour choisir une histoire, puis à chaque passage le narrateur lit
+// le texte ET les choix numérotés ("pour examiner la machine, faites le 1..."),
+// et on compose le chiffre du choix. Raccrocher arrête tout.
+//
+// Lit /sdcard/gamebook/{library.json, menu.wav, <id>.json, <id>_<passage>.wav}
+// (pack audio "téléphone" produit par tools/gamebook/build_phone_gamebook.py,
+// où la narration des WAV inclut déjà les invites "faites le N"). 100% local,
+// aucun modèle, aucun réseau.
+//
+// Intégration : conversation.c lance plip_gamebook_begin() au décroché (si un
+// pack est présent sur la SD), pousse chaque chiffre composé via
+// plip_gamebook_feed_digit(), et appelle plip_gamebook_end() au raccroché.
+
+#include <stdbool.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// La SD est-elle montée et un pack gamebook présent ? (à tester au décroché)
+bool plip_gamebook_available(void);
+
+// Démarre le mode livre-jeu : charge la bibliothèque et joue le menu audio.
+void plip_gamebook_begin(void);
+
+// Traite un chiffre composé (0..9). Menu → choisit l'histoire ; en histoire →
+// choisit la réponse ; sur une fin → 0 revient au menu. Interrompt la
+// narration en cours pour enchaîner.
+void plip_gamebook_feed_digit(int digit);
+
+// Quitte le mode livre-jeu (raccroché) : stoppe l'audio, libère la mémoire.
+void plip_gamebook_end(void);
+
+// Vrai tant qu'une partie est en cours.
+bool plip_gamebook_active(void);
+
+#ifdef __cplusplus
+}
+#endif

plip_voice/sdkconfig.defaults

@@ -52,3 +52,7 @@ CONFIG_PLIP_HOOK_ACTIVE_HIGH=y
 # Example: CONFIG_PLIP_GATEWAY_URL="http://192.168.0.175:8401"
 #          CONFIG_PLIP_GATEWAY_TOKEN="testtoken"
 # Default (Kconfig): http://192.168.0.50:8401 with empty token.
+
+# Long file names on the SD (voice sample pack uses >8.3 names)
+CONFIG_FATFS_LFN_HEAP=y
+CONFIG_FATFS_MAX_LFN=255