desklock/firmware/main/desklock_main.c

/*
 * DeskLock — living-room face and voice for the Tatlock butler.
 *
 * Boot sequence (planned):
 *   1. BSP init: display + LVGL, touch, audio codec (ES8311/ES7210)
 *   2. Wi-Fi up via ESP32-C6 (ESP-Hosted over SDIO)
 *   3. WebSocket connection to the DeskLock gateway
 *   4. Face state machine: idle / listening / thinking / speaking
 *
 * Current stage: bring-up — display + a placeholder face.
 */

#include <math.h>
#include <stdlib.h>

#include "esp_log.h"
#include "bsp/esp-bsp.h"
#include "esp_codec_dev.h"
#include "lvgl.h"

static const char *TAG = "desklock";

#define GONG_RATE 16000
#define GONG_SECONDS 5
#define GONG_SAMPLES (GONG_RATE * GONG_SECONDS)
#define GONG_VOLUME 58     /* codec volume: discreet, not loud */
#define GONG_PEAK 8000.0f  /* ~-12 dBFS headroom */

/* A distant temple gong in a stone space.
 *
 * Inharmonic partials with per-partial decay (lows ring for seconds, highs
 * die fast), a detuned fundamental pair for slow shimmer, a soft 45 ms
 * attack so the strike reads as far away, and three early reflections that
 * stand in for cathedral walls. Fundamental sits at 165 Hz — low enough to
 * feel voluminous, high enough for the bare 2 W speaker to reproduce.
 */
static const struct {
    float ratio;
    float amp;
    float tau; /* seconds to 1/e */
} gong_partials[] = {
    { 1.000f, 0.34f, 2.6f },
    { 1.004f, 0.22f, 2.2f }, /* beating partner -> shimmer */
    { 1.590f, 0.16f, 1.4f },
    { 2.140f, 0.11f, 0.9f },
    { 2.760f, 0.08f, 0.55f },
    { 3.570f, 0.05f, 0.35f },
    { 4.800f, 0.03f, 0.22f },
};

static const struct {
    int delay_ms;
    float gain;
} gong_echoes[] = { { 95, 0.45f }, { 210, 0.30f }, { 370, 0.18f } };

static void play_boot_gong(void)
{
    esp_codec_dev_handle_t spk = bsp_audio_codec_speaker_init();
    if (spk == NULL) {
        ESP_LOGW(TAG, "speaker init failed — no gong");
        return;
    }
    esp_codec_dev_sample_info_t fs = {
        .sample_rate = GONG_RATE,
        .channel = 1,
        .bits_per_sample = 16,
    };
    esp_codec_dev_set_out_vol(spk, GONG_VOLUME);
    esp_codec_dev_open(spk, &fs);

    float *dry = heap_caps_calloc(GONG_SAMPLES, sizeof(float), MALLOC_CAP_SPIRAM);
    int16_t *pcm = heap_caps_malloc(GONG_SAMPLES * sizeof(int16_t), MALLOC_CAP_SPIRAM);
    if (dry == NULL || pcm == NULL) {
        ESP_LOGW(TAG, "no memory for gong");
        goto out;
    }

    for (size_t p = 0; p < sizeof(gong_partials) / sizeof(gong_partials[0]); p++) {
        float step = 2.0f * (float)M_PI * 165.0f * gong_partials[p].ratio / GONG_RATE;
        float decay = expf(-1.0f / (gong_partials[p].tau * GONG_RATE));
        float env = gong_partials[p].amp;
        float phase = 0.0f;
        for (int i = 0; i < GONG_SAMPLES; i++) {
            phase += step;
            env *= decay;
            dry[i] += sinf(phase) * env;
        }
    }

    const int attack = GONG_RATE * 45 / 1000; /* distant, not percussive */
    for (int i = 0; i < attack; i++) {
        dry[i] *= (float)i / attack;
    }

    float peak = 1e-6f;
    for (int i = 0; i < GONG_SAMPLES; i++) {
        float s = dry[i];
        for (size_t e = 0; e < sizeof(gong_echoes) / sizeof(gong_echoes[0]); e++) {
            int j = i - GONG_RATE * gong_echoes[e].delay_ms / 1000;
            if (j >= 0) {
                s += dry[j] * gong_echoes[e].gain;
            }
        }
        const int fade = GONG_RATE * 2 / 5; /* last 400 ms to silence */
        if (i > GONG_SAMPLES - fade) {
            s *= (float)(GONG_SAMPLES - i) / fade;
        }
        dry[i] = s;
        if (fabsf(s) > peak) {
            peak = fabsf(s);
        }
    }
    for (int i = 0; i < GONG_SAMPLES; i++) {
        pcm[i] = (int16_t)(dry[i] * (GONG_PEAK / peak));
    }

    esp_codec_dev_write(spk, pcm, GONG_SAMPLES * sizeof(int16_t));
    ESP_LOGI(TAG, "boot gong played");

out:
    free(dry);
    free(pcm);
    esp_codec_dev_close(spk);
}

static void show_placeholder_face(void)
{
    bsp_display_lock(0);

    lv_obj_t *scr = lv_screen_active();
    lv_obj_set_style_bg_color(scr, lv_color_black(), 0);

    lv_obj_t *eyes = lv_label_create(scr);
    lv_label_set_text(eyes, "-   -");
    lv_obj_set_style_text_color(eyes, lv_color_hex(0xADFFC8), 0);
    lv_obj_set_style_text_font(eyes, &lv_font_montserrat_48, 0);
    lv_obj_align(eyes, LV_ALIGN_CENTER, 0, -80);

    lv_obj_t *mouth = lv_label_create(scr);
    lv_label_set_text(mouth, "\\_/");
    lv_obj_set_style_text_color(mouth, lv_color_hex(0xADFFC8), 0);
    lv_obj_set_style_text_font(mouth, &lv_font_montserrat_48, 0);
    lv_obj_align(mouth, LV_ALIGN_CENTER, 0, 40);

    lv_obj_t *tag = lv_label_create(scr);
    lv_label_set_text(tag, "DESKLOCK");
    lv_obj_set_style_text_color(tag, lv_color_hex(0x2F7A4B), 0);
    lv_obj_align(tag, LV_ALIGN_CENTER, 0, 160);

    bsp_display_unlock();
}

void app_main(void)
{
    ESP_LOGI(TAG, "DeskLock starting");

    bsp_display_start();
    bsp_display_backlight_on();
    show_placeholder_face();
    play_boot_gong();

    ESP_LOGI(TAG, "DeskLock up (placeholder face)");

    /* TODO(face): full LVGL face state machine per docs/architecture.md
     * TODO(net): Wi-Fi via ESP-Hosted (C6), WebSocket to gateway
     * TODO(voice): ES7210 capture upstream, ES8311 playback downstream
     */
}