box3: fix gamebook audio, accents, and crashes

The BOX-3 touch gamebook had several blocking bugs that made the
voiced+foley story packs unusable. This fixes them end to end.

- Stack overflow on story start: gb_play_task kept a 2 KB buffer on
  its 4 KB stack and rebooted the board. Buffers are now static (the
  task is a singleton) and the stack is 6 KB.
- No sound at all: the ES8311 codec was never initialised — main.c
  wrote raw I2S to a muted DAC. Output (and the mic) now go through
  the BSP esp_codec_dev (bsp_audio_codec_speaker/microphone_init,
  esp_codec_dev_write/read). The power amplifier (GPIO46) is also
  forced on and volume set to 100, as the es8311 pa_pin handling did
  not drive it in practice.
- First passage silent: gb_play re-mounted the SD that the gamebook
  had already mounted; the failed second bsp_sdcard_mount tore down
  the VFS so the first fopen failed. Now we open the file first and
  only mount if that fails. A short silence lead-in also primes the
  codec/PA ramp.
- No accents: the gamebook used ASCII-only lv_font_montserrat_14/24.
  Added custom Montserrat fonts (font_fr_14/24, Latin-1 + œŒŸ « » ’ …,
  uncompressed so they render without LV_USE_FONT_COMPRESSED).
- Mic/voice streaming gated behind CONFIG_BOX3_VOICE_STREAMING
  (default off): as a gamebook the BOX-3 needs no mic, and the
  constant codec reads + bridge reconnects interfered with playback
  and flooded the log.

plip_virtual + cmd_exec handle types switched from i2s_chan_handle_t
to esp_codec_dev_handle_t accordingly.

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" "font_fr_14.c" "font_fr_24.c"
     INCLUDE_DIRS "."
     PRIV_REQUIRES
         driver
@@ -13,5 +13,6 @@ idf_component_register(
         spiffs
         scenario_mesh
         espressif__esp-box-3
+        espressif__esp_codec_dev
         lvgl__lvgl
 )

box3_voice/main/Kconfig.projbuild

@@ -1,5 +1,15 @@
 menu "Zacus BOX-3 Voice Configuration"
 
+    config BOX3_VOICE_STREAMING
+        bool "Enable microphone + voice-bridge streaming"
+        default n
+        help
+            Start the mic capture task and the WebSocket voice-bridge client.
+            Off by default: as a touch gamebook the BOX-3 needs no microphone,
+            and the continuous codec reads + bridge reconnect attempts interfere
+            with speaker playback (the first narration clip came out silent) and
+            flood the log. Enable only when using the BOX-3 as a voice device.
+
     config ZACUS_WIFI_SSID
         string "WiFi SSID"
         default "zacus-net"

box3_voice/main/cmd_exec.c

@@ -10,7 +10,7 @@
 //
 // Objective 2 — play: real WAV streaming to I2S
 //   1. Mount SD card (best-effort, /sdcard). Parse WAV header (PCM-16).
-//   2. Stream PCM samples via s_spk_handle (extern, same I2S channel as TTS).
+//   2. Stream PCM samples via the ES8311 codec (esp_codec_dev), same path as TTS.
 //   3. If SD absent / file not found → fallback bip + log.
 //   4. Embedded test cue always available at virtual path "embedded://" or
 //      when the requested path starts with "embedded:" — proves WAV decode + I2S.
@@ -33,16 +33,16 @@
 
 #include "embedded_wav.h"   // s_embedded_wav[], EMBEDDED_WAV_SIZE
 
-// audio_play_tone and s_spk_handle are defined in main.c
+// audio_play_tone is defined in main.c (writes through the ES8311 codec).
 extern void audio_play_tone(float frequency, int duration_ms);
 
-// Speaker I2S handle — obtained from main.c via a weak accessor.
-// We declare a weak symbol here; main.c must define cmd_exec_set_spk_handle()
-// and call it after speaker_init().  Alternatively we use the driver directly.
-#include "driver/i2s_std.h"
-static i2s_chan_handle_t s_spk_handle_local = NULL;
+// Speaker codec (ES8311) handle — obtained from main.c after speaker_init().
+// Audio is streamed through esp_codec_dev so the ES8311 DAC is actually driven
+// (a raw I2S write leaves the codec muted).
+#include "esp_codec_dev.h"
+static esp_codec_dev_handle_t s_spk_handle_local = NULL;
 
-void cmd_exec_set_spk_handle(i2s_chan_handle_t h)
+void cmd_exec_set_spk_handle(esp_codec_dev_handle_t h)
 {
     s_spk_handle_local = h;
 }
@@ -344,13 +344,11 @@ static esp_err_t stream_pcm16(const uint8_t *pcm, size_t byte_len,
         size_t offset = 0;
         while (offset < byte_len) {
             size_t chunk = (byte_len - offset < 1024) ? (byte_len - offset) : 1024;
-            size_t written = 0;
-            esp_err_t ret = i2s_channel_write(s_spk_handle_local,
-                                               pcm + offset, chunk,
-                                               &written, pdMS_TO_TICKS(500));
-            if (ret != ESP_OK) {
-                ESP_LOGW(TAG, "I2S write error: %s", esp_err_to_name(ret));
-                return ret;
+            int ret = esp_codec_dev_write(s_spk_handle_local,
+                                          (void *)(pcm + offset), (int)chunk);
+            if (ret != ESP_CODEC_DEV_OK) {
+                ESP_LOGW(TAG, "codec write error: %d", ret);
+                return ESP_FAIL;
             }
             offset += chunk;
         }
@@ -488,6 +486,93 @@ 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;
+    /* The task is a singleton (guarded by s_gb_task), so these large buffers
+     * are kept static — on the stack they ate ~2.1 KB and, together with the
+     * FATFS fread internals + i2s + logging, overflowed the 4 KB stack and
+     * rebooted the board when a story started. */
+    static uint8_t hdr[64];
+    static 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;
+
+        /* The SD is normally already mounted (the gamebook mounts it at init).
+         * Try the file directly first — re-mounting an already-mounted card via
+         * bsp_sdcard_mount() FAILS and tears down the existing VFS, which made
+         * the very first narration clip silent. Only mount if the open fails. */
+        FILE *f = fopen(path, "rb");
+        if (!f) {
+            ensure_sd_mounted();
+            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);
+        /* Wake the codec/PA before the real audio: after idle the ES8311 mutes
+         * and ramps up on the first write, swallowing the start of the clip.
+         * A short silent lead-in lets the unmute settle so the narration plays
+         * from the very first word. */
+        memset(buf, 0, sizeof(buf));
+        for (int i = 0; i < 4 && s_spk_handle_local && !s_gb_stop && !s_gb_new; i++) {
+            esp_codec_dev_write(s_spk_handle_local, buf, (int)sizeof(buf));
+        }
+        size_t remaining = wi.data_size;
+        ESP_LOGI(TAG, "gb: play %s (%u B, %u Hz)", path,
+                 (unsigned)wi.data_size, (unsigned)wi.sample_rate);
+        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;
+            if (s_spk_handle_local) {
+                esp_codec_dev_write(s_spk_handle_local, buf, (int)got);
+            }
+            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", 6144, 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

@@ -18,7 +18,7 @@
 
 #include <stddef.h>
 #include "esp_err.h"
-#include "driver/i2s_std.h"
+#include "esp_codec_dev.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -30,9 +30,17 @@ extern "C" {
 // ESP_ERR_NOT_SUPPORTED for unknown ops (all logged, no crash).
 esp_err_t cmd_exec_handle(const char *data, size_t len);
 
-// Provide the speaker I2S channel handle so exec_play() can stream audio.
+// Provide the speaker codec (ES8311) handle so exec_play() can stream audio.
 // Must be called from main.c after speaker_init(), before any CMD arrives.
-void cmd_exec_set_spk_handle(i2s_chan_handle_t h);
+void cmd_exec_set_spk_handle(esp_codec_dev_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
 }

box3_voice/main/gamebook.c

@@ -0,0 +1,238 @@
+// 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";
+
+// Custom fonts with French glyphs (ASCII + Latin-1 + ’ + …). See main/font_fr_*.c.
+LV_FONT_DECLARE(font_fr_14);
+LV_FONT_DECLARE(font_fr_24);
+
+#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, &font_fr_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, &font_fr_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, &font_fr_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"
@@ -39,8 +40,12 @@ static const char *TAG = "zacus-voice";
 
 /* --------------- Shared state --------------- */
 
-static i2s_chan_handle_t s_spk_handle = NULL;   /* Speaker TX channel */
-static i2s_chan_handle_t s_mic_handle = NULL;   /* Microphone RX channel */
+/* Audio now goes through the ES8311 (speaker) / ES7210 (mic) codecs via the
+ * BSP + esp_codec_dev. bsp_audio_init() owns the single shared I2S port; we
+ * must NOT create our own raw I2S channels (that left the ES8311 DAC unpowered
+ * and the speaker silent). */
+static esp_codec_dev_handle_t s_spk_codec = NULL;   /* ES8311 output (DAC) */
+static esp_codec_dev_handle_t s_mic_codec = NULL;   /* ES7210 input (ADC)  */
 static volatile bool     s_wifi_connected = false;
 static volatile bool     s_voice_streaming = false;
 
@@ -113,11 +118,10 @@ static esp_err_t wifi_init_sta(void)
 // generator can sequence a melody for the master's microphone.
 void audio_play_tone(float frequency, int duration_ms)
 {
-    if (!s_spk_handle || duration_ms <= 0) return;
+    if (!s_spk_codec || duration_ms <= 0) return;
     const int total_samples = AUDIO_SAMPLE_RATE * duration_ms / 1000;
     const float amplitude = 16000.0f;
     int16_t buffer[256];
-    size_t bytes_written = 0;
     int sample_idx = 0;
     while (sample_idx < total_samples) {
         int chunk = (total_samples - sample_idx < 256)
@@ -126,29 +130,27 @@ void audio_play_tone(float frequency, int duration_ms)
             float t = (float)(sample_idx + i) / (float)AUDIO_SAMPLE_RATE;
             buffer[i] = (int16_t)(amplitude * sinf(2.0f * M_PI * frequency * t));
         }
-        i2s_channel_write(s_spk_handle, buffer, chunk * sizeof(int16_t),
-                          &bytes_written, portMAX_DELAY);
+        esp_codec_dev_write(s_spk_codec, buffer, chunk * sizeof(int16_t));
         sample_idx += chunk;
     }
 }
 
 static void audio_test_tone(void)
 {
-    if (!s_spk_handle) {
+    if (!s_spk_codec) {
         ESP_LOGW(TAG, "Speaker not initialized, skipping test tone");
         return;
     }
 
     ESP_LOGI(TAG, "Playing 440 Hz test tone (1 second)...");
 
-    /* Generate 1 second of 440 Hz sine wave using persistent speaker handle */
+    /* Generate 1 second of 440 Hz sine wave through the ES8311 codec */
     const int duration_ms = 1000;
     const int total_samples = AUDIO_SAMPLE_RATE * duration_ms / 1000;
     const float frequency = 440.0f;
     const float amplitude = 16000.0f;
 
     int16_t buffer[256];
-    size_t bytes_written = 0;
     int sample_idx = 0;
 
     while (sample_idx < total_samples) {
@@ -157,7 +159,7 @@ static void audio_test_tone(void)
             float t = (float)(sample_idx + i) / (float)AUDIO_SAMPLE_RATE;
             buffer[i] = (int16_t)(amplitude * sinf(2.0f * M_PI * frequency * t));
         }
-        i2s_channel_write(s_spk_handle, buffer, chunk * sizeof(int16_t), &bytes_written, portMAX_DELAY);
+        esp_codec_dev_write(s_spk_codec, buffer, chunk * sizeof(int16_t));
         sample_idx += chunk;
     }
 
@@ -173,15 +175,13 @@ static void tts_audio_callback(const uint8_t *data, size_t len)
      * Called from WebSocket event context with incoming TTS PCM16 data.
      * Write directly to the speaker I2S channel.
      */
-    if (!s_spk_handle || len == 0) {
+    if (!s_spk_codec || len == 0) {
         return;
     }
 
-    size_t bytes_written = 0;
-    esp_err_t ret = i2s_channel_write(s_spk_handle, data, len,
-                                       &bytes_written, pdMS_TO_TICKS(200));
-    if (ret != ESP_OK) {
-        ESP_LOGW(TAG, "TTS write to speaker failed: %s", esp_err_to_name(ret));
+    int ret = esp_codec_dev_write(s_spk_codec, (void *)data, (int)len);
+    if (ret != ESP_CODEC_DEV_OK) {
+        ESP_LOGW(TAG, "TTS write to speaker failed: %d", ret);
     }
 }
 
@@ -189,34 +189,42 @@ static void tts_audio_callback(const uint8_t *data, size_t len)
 
 static esp_err_t speaker_init(void)
 {
-    /* Enable power amplifier */
+    /* Init the ES8311 output codec via the BSP. This also runs bsp_audio_init()
+     * (shared duplex I2S port) + bsp_i2c_init() and configures the power
+     * amplifier (BSP_POWER_AMP_IO = GPIO46). The ES8311 DAC is configured over
+     * I2C here — without this the speaker stays muted no matter what we push on
+     * I2S. */
+    s_spk_codec = bsp_audio_codec_speaker_init();
+    if (!s_spk_codec) {
+        ESP_LOGE(TAG, "bsp_audio_codec_speaker_init failed — speaker unavailable");
+        return ESP_FAIL;
+    }
+
+    /* All speaker audio in this app is 16 kHz / 16-bit / mono PCM. */
+    esp_codec_dev_sample_info_t fs = {
+        .bits_per_sample = AUDIO_BITS,
+        .channel         = AUDIO_CHANNELS,
+        .channel_mask    = 0,
+        .sample_rate     = AUDIO_SAMPLE_RATE,
+        .mclk_multiple   = 0,
+    };
+    int ret = esp_codec_dev_open(s_spk_codec, &fs);
+    if (ret != ESP_CODEC_DEV_OK) {
+        ESP_LOGE(TAG, "esp_codec_dev_open (speaker) failed: %d", ret);
+        return ESP_FAIL;
+    }
+
+    /* Output volume 0..100. */
+    esp_codec_dev_set_out_vol(s_spk_codec, 100);
+
+    /* Belt-and-suspenders: the es8311 driver is configured with pa_pin=GPIO46
+     * (pa_reverted=false) and should raise it on open, but the speaker stays
+     * silent in practice — force the power amplifier on explicitly. */
     gpio_set_direction(BOX3_PA_ENABLE, GPIO_MODE_OUTPUT);
     gpio_set_level(BOX3_PA_ENABLE, 1);
 
-    i2s_chan_config_t chan_cfg = I2S_CHANNEL_DEFAULT_CONFIG(I2S_NUM_0, I2S_ROLE_MASTER);
-    ESP_ERROR_CHECK(i2s_new_channel(&chan_cfg, &s_spk_handle, NULL));
-
-    i2s_std_config_t std_cfg = {
-        .clk_cfg  = I2S_STD_CLK_DEFAULT_CONFIG(AUDIO_SAMPLE_RATE),
-        .slot_cfg = I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG(I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO),
-        .gpio_cfg = {
-            .mclk = BOX3_I2S_MCLK,
-            .bclk = BOX3_I2S_BCLK,
-            .ws   = BOX3_I2S_WS,
-            .dout = BOX3_I2S_DOUT,
-            .din  = I2S_GPIO_UNUSED,
-            .invert_flags = {
-                .mclk_inv = false,
-                .bclk_inv = false,
-                .ws_inv   = false,
-            },
-        },
-    };
-
-    ESP_ERROR_CHECK(i2s_channel_init_std_mode(s_spk_handle, &std_cfg));
-    ESP_ERROR_CHECK(i2s_channel_enable(s_spk_handle));
-
-    ESP_LOGI(TAG, "Speaker I2S initialized (I2S_NUM_0)");
+    ESP_LOGI(TAG, "Speaker codec (ES8311) initialized @ %d Hz, vol=100, PA on",
+             AUDIO_SAMPLE_RATE);
     return ESP_OK;
 }
 
@@ -226,43 +234,45 @@ static void mic_monitor_task(void *arg)
 {
     ESP_LOGI(TAG, "Starting mic capture task...");
 
-    i2s_chan_config_t chan_cfg = I2S_CHANNEL_DEFAULT_CONFIG(I2S_NUM_1, I2S_ROLE_MASTER);
-    ESP_ERROR_CHECK(i2s_new_channel(&chan_cfg, NULL, &s_mic_handle));
+    /* Init the ES7210 input codec via the BSP. It shares the same I2S port
+     * (already created by speaker_init -> bsp_audio_init); creating a separate
+     * raw I2S channel here would conflict with that shared port. */
+    s_mic_codec = bsp_audio_codec_microphone_init();
+    if (!s_mic_codec) {
+        ESP_LOGE(TAG, "bsp_audio_codec_microphone_init failed — mic unavailable");
+        vTaskDelete(NULL);
+        return;
+    }
 
-    i2s_std_config_t std_cfg = {
-        .clk_cfg  = I2S_STD_CLK_DEFAULT_CONFIG(AUDIO_SAMPLE_RATE),
-        .slot_cfg = I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG(I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO),
-        .gpio_cfg = {
-            .mclk = I2S_GPIO_UNUSED,
-            .bclk = BOX3_I2S_BCLK,
-            .ws   = BOX3_I2S_WS,
-            .dout = I2S_GPIO_UNUSED,
-            .din  = BOX3_I2S_DIN,
-            .invert_flags = {
-                .mclk_inv = false,
-                .bclk_inv = false,
-                .ws_inv   = false,
-            },
-        },
+    esp_codec_dev_sample_info_t fs = {
+        .bits_per_sample = AUDIO_BITS,
+        .channel         = AUDIO_CHANNELS,
+        .channel_mask    = 0,
+        .sample_rate     = AUDIO_SAMPLE_RATE,
+        .mclk_multiple   = 0,
     };
-
-    ESP_ERROR_CHECK(i2s_channel_init_std_mode(s_mic_handle, &std_cfg));
-    ESP_ERROR_CHECK(i2s_channel_enable(s_mic_handle));
+    int oret = esp_codec_dev_open(s_mic_codec, &fs);
+    if (oret != ESP_CODEC_DEV_OK) {
+        ESP_LOGE(TAG, "esp_codec_dev_open (mic) failed: %d", oret);
+        vTaskDelete(NULL);
+        return;
+    }
+    /* ES7210 input gain (dB). 30 dB is a sane default for the BOX-3 mic array. */
+    esp_codec_dev_set_in_gain(s_mic_codec, 30.0f);
 
     int16_t buffer[AUDIO_FRAME_SAMPLES];
-    size_t bytes_read = 0;
     int rms_log_counter = 0;
 
     while (1) {
-        esp_err_t ret = i2s_channel_read(s_mic_handle, buffer,
-                                          AUDIO_FRAME_SAMPLES * sizeof(int16_t),
-                                          &bytes_read, pdMS_TO_TICKS(1000));
-        if (ret != ESP_OK) {
-            ESP_LOGW(TAG, "Mic read error: %s", esp_err_to_name(ret));
+        int ret = esp_codec_dev_read(s_mic_codec, buffer,
+                                     AUDIO_FRAME_SAMPLES * sizeof(int16_t));
+        if (ret != ESP_CODEC_DEV_OK) {
+            ESP_LOGW(TAG, "Mic read error: %d", ret);
+            vTaskDelay(pdMS_TO_TICKS(20));
             continue;
         }
 
-        int samples = bytes_read / sizeof(int16_t);
+        int samples = AUDIO_FRAME_SAMPLES;
 
         /* If voice streaming is active, send PCM to bridge */
         if (s_voice_streaming) {
@@ -422,14 +432,19 @@ void app_main(void)
     /* Initialize persistent speaker output (used for test tone + TTS playback) */
     speaker_init();
 
-    /* Pass speaker handle to CMD executor for real WAV playback */
-    cmd_exec_set_spk_handle(s_spk_handle);
+    /* Pass speaker codec handle to CMD executor for real WAV playback */
+    cmd_exec_set_spk_handle(s_spk_codec);
 
     /* Play test tone to verify audio output */
     audio_test_tone();
 
-    /* Start mic capture task (reads I2S mic, streams to bridge when active) */
+#if CONFIG_BOX3_VOICE_STREAMING
+    /* Start mic capture task (reads I2S mic, streams to bridge when active).
+     * Off by default: as a touch gamebook the BOX-3 needs no mic, and the
+     * continuous codec reads + bridge reconnects interfered with the speaker
+     * (the first narration clip came out silent) and spammed the log. */
     xTaskCreatePinnedToCore(mic_monitor_task, "mic_capture", 4096, NULL, 5, NULL, 1);
+#endif
 
     /* Start button handler (BOOT button toggles voice streaming) */
     xTaskCreate(button_task, "button", 2048, NULL, 4, NULL);
@@ -438,8 +453,10 @@ void app_main(void)
     ESP_LOGI(TAG, "Connecting to WiFi...");
     wifi_init_sta();
 
+#if CONFIG_BOX3_VOICE_STREAMING
     /* Start voice bridge connection task (waits for WiFi, then connects WS) */
     xTaskCreate(voice_bridge_task, "voice_bridge", 6144, NULL, 5, NULL);
+#endif
 
     /* Start the scenario hot-load HTTP server (POST /game/scenario).
      * httpd_start binds to all netifs — works as soon as the WiFi STA has an IP. */
@@ -449,7 +466,7 @@ void app_main(void)
         /* Phone-less PLIP annex: same REST contract as PLIP_FIRMWARE
          * (POST /ring /stop /play, GET /status), ring on the speaker,
          * BOOT button as the virtual hook switch. */
-        if (plip_virtual_init(scenario_server_handle(), s_spk_handle) != ESP_OK) {
+        if (plip_virtual_init(scenario_server_handle(), s_spk_codec) != ESP_OK) {
             ESP_LOGW(TAG, "plip_virtual_init failed — virtual phone unavailable");
         } else if (plip_ui_init() != ESP_OK) {
             /* REST/ESP-NOW phone still works headless if the UI fails. */
@@ -463,6 +480,11 @@ 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

box3_voice/main/plip_virtual.c

@@ -38,7 +38,7 @@ typedef struct {
     char reason[32];  /* "pickup" | "hangup" | "ring_timeout" | "dial:<n>" */
 } hook_event_t;
 
-static i2s_chan_handle_t        s_spk;
+static esp_codec_dev_handle_t   s_spk;
 static volatile plip_hook_state_t s_state = PLIP_HOOK_ON;
 static volatile bool            s_ring_stop = false;
 static TaskHandle_t             s_ring_task = NULL;
@@ -107,7 +107,6 @@ static void ring_burst(int duration_ms)
 {
     const int total = AUDIO_SAMPLE_RATE * duration_ms / 1000;
     int16_t buffer[256];
-    size_t written = 0;
     int idx = 0;
     while (idx < total && !s_ring_stop) {
         int chunk = (total - idx < 256) ? (total - idx) : 256;
@@ -116,8 +115,9 @@ static void ring_burst(int duration_ms)
             buffer[i] = (int16_t) (RING_AMPLITUDE *
                                    sinf(2.0f * (float) M_PI * RING_FREQ_HZ * t));
         }
-        i2s_channel_write(s_spk, buffer, chunk * sizeof(int16_t), &written,
-                          pdMS_TO_TICKS(500));
+        if (s_spk) {
+            esp_codec_dev_write(s_spk, buffer, chunk * sizeof(int16_t));
+        }
         idx += chunk;
     }
 }
@@ -279,7 +279,7 @@ plip_hook_state_t plip_virtual_state(void)
     return s_state;
 }
 
-esp_err_t plip_virtual_init(httpd_handle_t server, i2s_chan_handle_t spk)
+esp_err_t plip_virtual_init(httpd_handle_t server, esp_codec_dev_handle_t spk)
 {
     if (!server || !spk) return ESP_ERR_INVALID_ARG;
     s_spk = spk;

box3_voice/main/plip_virtual.h

@@ -16,7 +16,7 @@
 #pragma once
 
 #include <stdbool.h>
-#include "driver/i2s_std.h"
+#include "esp_codec_dev.h"
 #include "esp_err.h"
 #include "esp_http_server.h"
 
@@ -34,7 +34,7 @@ typedef enum {
 // Register the REST handlers on `server` and keep `spk` for the ring tone.
 // Call once after scenario_server_start(); the speaker channel must be
 // enabled (speaker_init() in main.c).
-esp_err_t plip_virtual_init(httpd_handle_t server, i2s_chan_handle_t spk);
+esp_err_t plip_virtual_init(httpd_handle_t server, esp_codec_dev_handle_t spk);
 
 // Forward a BOOT-button press. Returns true when the press was consumed as
 // a hook transition (pickup/hangup) — the caller should then skip its own

idf_zacus/components/display_ui/display_ui.cpp

@@ -255,9 +255,12 @@ 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[512];
+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
@@ -824,37 +827,49 @@ static void build_gamebook_screen(void) {
     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_VALUE), 0);
-    lv_obj_set_style_text_font(s_gb_title, &lv_font_montserrat_24, 0);
+    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 - 16);
+    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, 6);
+    lv_obj_align(s_gb_title, LV_ALIGN_TOP_MID, 0, 4);
 
-    // Body band: top area, height-bounded so it can never run into the menu.
-    s_gb_body = lv_label_create(s_scr_gamebook);
+    // 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_size(s_gb_body, DUI_HOR_RES - 24, 196);
+    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, 12, 44);
-
-    // Menu band: reserved strip at the bottom, drawn last (foreground), with a
-    // faint background so the choices stand out from the body text.
-    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(0x101820), 0);
-    lv_obj_set_style_bg_opa(s_gb_menu, LV_OPA_COVER, 0);
-    lv_obj_set_style_pad_all(s_gb_menu, 6, 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_WRAP);
-    lv_obj_set_width(s_gb_menu, DUI_HOR_RES);
-    lv_label_set_text(s_gb_menu, "");
-    lv_obj_align(s_gb_menu, LV_ALIGN_BOTTOM_LEFT, 0, 0);
-    lv_obj_move_foreground(s_gb_menu);
+    lv_obj_align(s_gb_body, LV_ALIGN_TOP_LEFT, 0, 0);
 }
 
 static void build_status_screen(void) {
@@ -949,11 +964,22 @@ 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.
+    // 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) {
@@ -1278,18 +1304,27 @@ extern "C" void display_ui_set_key_hook(display_ui_key_hook_t hook) {
 }
 
 extern "C" void display_ui_gamebook_show(const char *title, const char *body,
-                                         const char *menu) {
+                                         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;
 }
 

idf_zacus/components/display_ui/include/display_ui.h

@@ -103,7 +103,13 @@ void display_ui_set_key_hook(display_ui_key_hook_t hook);
  * @param menu   choice lines (bottom), e.g. "[OK] ...\n[<>] ..."
  */
 void display_ui_gamebook_show(const char *title, const char *body,
-                              const char *menu);
+                              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);

idf_zacus/components/game_endpoint/game_endpoint.c

@@ -1173,6 +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
@@ -1244,6 +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,
@@ -1326,6 +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));

idf_zacus/components/gamebook/gamebook.c

@@ -9,6 +9,7 @@
 
 #include "cJSON.h"
 #include "esp_log.h"
+#include "esp_heap_caps.h"
 
 #include "display_ui.h"
 #include "media_manager.h"
@@ -17,7 +18,7 @@ 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 JSON_MAX       (256 * 1024)  /* expanded books are ~80 KB of JSON */
 #define LIB_MAX        6             /* matches the display tile grid */
 
 typedef enum { GB_OFF, GB_LIBRARY, GB_STORY } gb_mode_t;
@@ -39,6 +40,10 @@ static volatile gb_mode_t s_mode = GB_OFF;
 
 bool gamebook_active(void) { return s_mode != GB_OFF; }
 
+/* PSRAM allocators for cJSON — book trees are large (see gamebook_init). */
+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); }
+
 /* Read a whole JSON file from the SD into a cJSON tree (caller frees). */
 static cJSON *load_json(const char *path)
 {
@@ -48,13 +53,13 @@ static cJSON *load_json(const char *path)
     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);
+    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);
-    free(buf);
+    heap_caps_free(buf);
     if (!root) ESP_LOGW(TAG, "malformed JSON: %s", path);
     return root;
 }
@@ -112,8 +117,10 @@ static int cur_choice_count(void)
     return cJSON_IsArray(ch) ? cJSON_GetArraySize(ch) : 0;
 }
 
-/* (Re)draw the current page: title + wrapped text + choice list with a cursor. */
-static void render_page(void)
+/* (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;
@@ -122,23 +129,25 @@ static void render_page(void)
     const cJSON *choices = cJSON_GetObjectItem(p, "choices");
     int n = cJSON_IsArray(choices) ? cJSON_GetArraySize(choices) : 0;
 
-    char menu[256];
+    char menu[160];
     if (n == 0) {
         snprintf(menu, sizeof(menu), "~ Fin ~   (clic = bibliotheque)");
     } else {
-        size_t off = 0;
-        for (int i = 0; i < n && off < sizeof(menu); i++) {
-            const cJSON *lbl = cJSON_GetObjectItem(cJSON_GetArrayItem(choices, i),
-                                                   "label");
-            off += snprintf(menu + off, sizeof(menu) - off, "%s%s\n",
-                            i == s_sel ? "> " : "  ",
-                            cJSON_IsString(lbl) ? lbl->valuestring : "?");
+        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);
+        menu, home);
 }
 
 static void enter_passage(const char *pid)
@@ -147,7 +156,7 @@ static void enter_passage(const char *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();
+    render_page(true);
 
     const cJSON *wav = cJSON_GetObjectItem(p, "wav");
     if (cJSON_IsString(wav) && wav->valuestring[0]) {
@@ -191,43 +200,58 @@ static void load_book(int idx)
 
 // ── Input ─────────────────────────────────────────────────────────────────────
 
-/* Library grid nav: 2 columns. up/down = ±2 (row), left/right = ±1 (col),
- * click = open the highlighted story. Codes: 1=click 2=down 3=left 4=right
- * 5=up (5-way ladder). */
+/* 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)
 {
-    int s = s_lib_sel;
+    if (s_lib_n <= 0) return true;
     switch (key) {
-    case 5: s -= 2; break;   /* up    */
-    case 2: s += 2; break;   /* down  */
-    case 3: s -= 1; break;   /* left  */
-    case 4: s += 1; break;   /* right */
-    case 1:                  /* click → open */
-        load_book(s_lib_sel);
-        return true;
-    default: return true;
+    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;
     }
-    if (s >= 0 && s < s_lib_n) { s_lib_sel = s; render_library(); }
     return true;
 }
 
-/* Story nav: D-pad moves the choice cursor, click confirms. Ending + click
- * returns to the library. */
+/* 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 → back to library on click */
-        if (key == 1) open_library();
+    if (n == 0) {                       /* ending → click returns to library */
+        if (key == PAD_CLICK) open_library();
         return true;
     }
     switch (key) {
-    case 5: case 3: s_sel = (s_sel - 1 + n) % n; render_page(); break;  /* up/left  */
-    case 2: case 4: s_sel = (s_sel + 1) % n;     render_page(); break;  /* down/right */
-    case 1: {                           /* click → confirm */
+    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, "select #%d -> '%s'", s_sel, g->valuestring);
+            ESP_LOGI(TAG, "validate #%d -> '%s'", s_sel, g->valuestring);
             enter_passage(g->valuestring);
         }
         break;
@@ -266,6 +290,14 @@ void gamebook_stop(void)
 
 void gamebook_init(void)
 {
+    /* Keep the large book trees in PSRAM — an expanded book is ~80 KB of JSON,
+     * whose parsed node tree would exhaust the internal heap on this firmware. */
+    static cJSON_Hooks hooks = {
+        .malloc_fn = gb_malloc,
+        .free_fn   = gb_free,
+    };
+    cJSON_InitHooks(&hooks);
+
     /* 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. */

idf_zacus/components/ota_server/ota_server.c

@@ -310,9 +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 = 22;   // ota + voice_hook + game (incl. /game/step,
-                                    // /game/puzzle_state, /game/file, relay,
-                                    // /game/gamebook)
+    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;

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/Kconfig.projbuild

@@ -84,6 +84,16 @@ menu "PLIP Voice Configuration"
             Can be combined with PLIP_DIAL_PULSE: whichever source detects a
             digit first wins.  Default off — enable for touch-tone handsets.
 
+    config PLIP_AUTO_RING
+        bool "Auto story-ring (rings by itself every 15-30 min)"
+        default n
+        help
+            When enabled, the phone rings on its own at a random 15-30 minute
+            interval; picking up launches a random audio story straight away
+            (no menu). If nobody answers within 1 minute the bell stops and the
+            next ring is rescheduled. Default off — enable for the escape-room
+            ambience where the phone calls the players.
+
     config PLIP_GATEWAY_URL
         string "NPC Gateway Base URL"
         default "http://192.168.0.50:8401"

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
@@ -32,6 +33,7 @@
 #include "esp_log.h"
 #include "esp_timer.h"
 #include "esp_heap_caps.h"
+#include "esp_random.h"
 
 #include <stdio.h>
 #include <string.h>
@@ -68,6 +70,7 @@ 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;
@@ -96,6 +99,15 @@ static char s_incoming_number[16] = {0};
 static char s_incoming_scene[40] = {0};
 static volatile bool s_incoming_armed = false;
 
+/* Auto story-ring: the phone rings on its own at a random interval; picking up
+ * launches a random audio story straight away. If nobody answers, the bell
+ * stops after STORY_RING_TIMEOUT_MS and the next ring is rescheduled. */
+#define STORY_RING_MIN_MS      (15 * 60 * 1000)   /* 15 min */
+#define STORY_RING_SPAN_MS     (15 * 60 * 1000)   /* + up to 15 min → 30 max */
+#define STORY_RING_TIMEOUT_MS  (60 * 1000)        /* give up after 1 min */
+static volatile bool s_story_ring_armed = false;
+static int64_t       s_story_ring_start_us = 0;
+
 /* Known numbers: ringback when dialed */
 static const char *KNOWN[] = {
     "12", "3615", "15", "17", "18", "0142738200", NULL
@@ -125,10 +137,16 @@ static void scene_slug(const char *scene, char *out, size_t cap)
 
 static void go_idle(void)
 {
+    if (s_story_ring_armed) {   /* hung up mid auto-ring → silence the bell */
+        phone_ring_stop();
+        slic_ring_stop();
+        s_story_ring_armed = false;
+    }
     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
@@ -161,6 +179,53 @@ static void enter_incoming_greet(void)
     ESP_LOGI(TAG, "INCOMING pickup -> GREET num=%s sid=%s", s_number, s_sid);
 }
 
+/* Answered an auto story-ring: stop the bell and launch a random story right
+ * into STATE_GAMEBOOK (no menu). Drives s_offhook directly because the
+ * debounced GPIO hook detection is unreliable while the bell rings. */
+static void enter_story_from_ring(void)
+{
+    s_story_ring_armed = false;
+    phone_ring_stop();
+    slic_ring_stop();
+    tones_stop();
+    dialer_reset();
+    s_offhook = true;
+#if CONFIG_PLIP_DIAL_DTMF
+    dtmf_start();
+#endif
+    audio_pa_set(true);
+    plip_gamebook_begin_random();
+    s_state = STATE_GAMEBOOK;
+    ESP_LOGI(TAG, "story-ring pickup -> GAMEBOOK (random story)");
+}
+
+#if CONFIG_PLIP_AUTO_RING
+/* Rings the phone at a random 15–30 min interval. Only rings when the line is
+ * truly idle (on-hook, no other call/ring armed) and a story pack is present.
+ * The 1-minute no-answer timeout is handled in conv_task. */
+static void story_ring_task(void *arg)
+{
+    (void)arg;
+    for (;;) {
+        uint32_t wait_ms = STORY_RING_MIN_MS + (esp_random() % STORY_RING_SPAN_MS);
+        ESP_LOGI(TAG, "auto story-ring: next attempt in %u min",
+                 (unsigned)(wait_ms / 60000));
+        vTaskDelay(pdMS_TO_TICKS(wait_ms));
+
+        if (s_state == STATE_IDLE && !s_offhook && !s_incoming_armed
+            && !s_story_ring_armed && plip_gamebook_available()) {
+            s_story_ring_armed = true;
+            s_story_ring_start_us = esp_timer_get_time();
+            phone_ring_start();
+            ESP_LOGI(TAG, "auto story-ring: ringing (up to %d s)",
+                     STORY_RING_TIMEOUT_MS / 1000);
+        } else {
+            ESP_LOGI(TAG, "auto story-ring: line busy/unavailable — skipped");
+        }
+    }
+}
+#endif /* CONFIG_PLIP_AUTO_RING */
+
 static void conv_task(void *arg)
 {
     (void)arg;
@@ -182,9 +247,23 @@ static void conv_task(void *arg)
             } else {
                 /* Off-hook from idle */
                 if (s_state == STATE_IDLE) {
-                    if (s_incoming_armed) {
+                    if (s_story_ring_armed) {
+                        /* Answered an auto story-ring → straight into a story. */
+                        enter_story_from_ring();
+                    } else 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();
@@ -208,6 +287,19 @@ static void conv_task(void *arg)
              * line directly — it reads the pickup cleanly mid-ring. */
             if (s_incoming_armed && slic_is_offhook()) {
                 enter_incoming_greet();
+            } else if (s_story_ring_armed) {
+                /* Same raw-SLIC poll for the auto story-ring, plus the
+                 * no-answer timeout: stop the bell after 1 min unanswered. */
+                if (slic_is_offhook()) {
+                    enter_story_from_ring();
+                } else if ((esp_timer_get_time() - s_story_ring_start_us) / 1000
+                           >= STORY_RING_TIMEOUT_MS) {
+                    phone_ring_stop();
+                    slic_ring_stop();
+                    s_story_ring_armed = false;
+                    ESP_LOGI(TAG, "auto story-ring: no answer after %d s — stop",
+                             STORY_RING_TIMEOUT_MS / 1000);
+                }
             }
             break;
 
@@ -224,6 +316,21 @@ 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();
@@ -454,6 +561,10 @@ void conversation_init(void)
     xTaskCreate(conv_task, "conv", 8192, NULL, 4, NULL);
 #else
     xTaskCreate(conv_task, "conv", 6144, NULL, 4, NULL);
+#endif
+#if CONFIG_PLIP_AUTO_RING
+    /* Auto story-ring scheduler: rings every 15–30 min when idle. */
+    xTaskCreate(story_ring_task, "storyring", 3072, NULL, 3, NULL);
 #endif
     ESP_LOGI(TAG, "conversation init");
 }

plip_voice/main/net.c

@@ -54,8 +54,9 @@
 
 #define SPIFFS_LABEL  "storage"
 #define SPIFFS_BASE   "/spiffs"
-#define MAX_BODY      (512 * 1024)   /* streamed in 2 KB chunks → RAM-safe; raised
-                                        so full-length say() hint WAVs fit on SD */
+#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;

plip_voice/main/plip_gamebook.c

@@ -0,0 +1,190 @@
+// 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 "esp_random.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_begin_random(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
+    int idx = (int)(esp_random() % (uint32_t)s_lib_n);
+    ESP_LOGI(TAG, "auto-start random story #%d/%d", idx, s_lib_n);
+    load_book(idx);        // straight into the story, no 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 +1,48 @@
+#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);
+
+// Démarre directement UNE histoire tirée au hasard (sans passer par le menu) —
+// utilisé quand le téléphone sonne tout seul et qu'on décroche.
+void plip_gamebook_begin_random(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

@@ -56,3 +56,6 @@ CONFIG_PLIP_HOOK_ACTIVE_HIGH=y
 # Long file names on the SD (voice sample pack uses >8.3 names)
 CONFIG_FATFS_LFN_HEAP=y
 CONFIG_FATFS_MAX_LFN=255
+
+# Speaker at max (earpiece)
+CONFIG_PLIP_SPEAKER_VOLUME=100