feat(voice): migrate to push2talk and serialize websocket text sends
This commit is contained in:
@@ -17,6 +17,7 @@ idf_component_register(
|
||||
"src/voice_interaction_ws_audio.c"
|
||||
"src/voice_interaction_marker.c"
|
||||
"src/voice_interaction_tasks.c"
|
||||
"src/voice_interaction_mic_key.c"
|
||||
"third_party/qrcodegen.c"
|
||||
INCLUDE_DIRS
|
||||
"include"
|
||||
|
||||
@@ -8,6 +8,7 @@
|
||||
#include "esp_opus_enc.h"
|
||||
#include "esp_websocket_client.h"
|
||||
#include "freertos/FreeRTOS.h"
|
||||
#include "freertos/queue.h"
|
||||
#include "freertos/semphr.h"
|
||||
#include "freertos/task.h"
|
||||
#include "domain.h"
|
||||
@@ -21,7 +22,7 @@
|
||||
|
||||
#define VOICE_SAMPLE_BYTES 2
|
||||
#define VOICE_STATUS_LOCK_TIMEOUT_MS 1000
|
||||
#define VOICE_WS_SEND_TIMEOUT_MS 1000
|
||||
#define VOICE_WS_SEND_TIMEOUT_MS 3000
|
||||
#define VOICE_WS_TASK_STACK 24576
|
||||
#define VOICE_UPLINK_TASK_STACK 32768
|
||||
#define VOICE_HEARTBEAT_TASK_STACK 4096
|
||||
@@ -31,15 +32,25 @@
|
||||
#define VOICE_AUDIO_DRAIN_QUIET_MS 180
|
||||
#define VOICE_AUDIO_DRAIN_WAIT_MS 2200
|
||||
#define VOICE_AUDIO_DRAIN_WAIT_ON_STOP_MS 3500
|
||||
#define VOICE_WS_TEXT_QUEUE_LEN 16
|
||||
#define VOICE_WS_TEXT_QUEUE_RX_TIMEOUT_MS 100
|
||||
|
||||
typedef struct {
|
||||
char *text;
|
||||
int len;
|
||||
} voice_ws_text_msg_t;
|
||||
|
||||
typedef struct {
|
||||
SemaphoreHandle_t lock;
|
||||
esp_websocket_client_handle_t ws;
|
||||
QueueHandle_t ws_text_queue;
|
||||
|
||||
TaskHandle_t uplink_task;
|
||||
TaskHandle_t heartbeat_task;
|
||||
TaskHandle_t ws_text_task;
|
||||
bool uplink_task_with_caps;
|
||||
bool heartbeat_task_with_caps;
|
||||
bool ws_text_task_with_caps;
|
||||
|
||||
void *opus_enc;
|
||||
void *opus_dec;
|
||||
@@ -63,7 +74,10 @@ typedef struct {
|
||||
bool ws_connected;
|
||||
bool started;
|
||||
bool tap_active;
|
||||
bool speech_active;
|
||||
bool stop_requested;
|
||||
bool ws_text_shutdown;
|
||||
bool ws_text_busy;
|
||||
bool ws_low_stack_warned;
|
||||
bool response_first_sentence_checked;
|
||||
|
||||
@@ -129,6 +143,10 @@ void voice_websocket_event_handler(void *handler_args,
|
||||
void *event_data);
|
||||
|
||||
void voice_interaction_ws_init_cjson_hooks(void);
|
||||
esp_err_t voice_interaction_try_send_speech(char *err, size_t err_len);
|
||||
|
||||
void voice_uplink_task(void *arg);
|
||||
void voice_heartbeat_task(void *arg);
|
||||
void voice_ws_text_task(void *arg);
|
||||
|
||||
esp_err_t voice_interaction_mic_key_init(void);
|
||||
|
||||
@@ -20,18 +20,44 @@ static uint32_t clamp_timeout_ms(uint32_t timeout_ms, uint32_t default_ms, uint3
|
||||
return value;
|
||||
}
|
||||
|
||||
static bool voice_wait_ws_text_idle(uint32_t timeout_ms) {
|
||||
int64_t deadline_ms = voice_now_ms() + (int64_t)timeout_ms;
|
||||
while (voice_now_ms() < deadline_ms) {
|
||||
bool busy = false;
|
||||
QueueHandle_t queue = NULL;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
busy = s_voice.ws_text_busy;
|
||||
queue = s_voice.ws_text_queue;
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
UBaseType_t pending = 0;
|
||||
if (queue != NULL) {
|
||||
pending = uxQueueMessagesWaiting(queue);
|
||||
}
|
||||
|
||||
if (!busy && pending == 0) {
|
||||
return true;
|
||||
}
|
||||
vTaskDelay(pdMS_TO_TICKS(20));
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
static bool voice_wait_worker_tasks_exit(uint32_t timeout_ms) {
|
||||
int64_t deadline_ms = voice_now_ms() + (int64_t)timeout_ms;
|
||||
while (voice_now_ms() < deadline_ms) {
|
||||
bool uplink_alive = false;
|
||||
bool heartbeat_alive = false;
|
||||
bool ws_text_alive = false;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
uplink_alive = (s_voice.uplink_task != NULL);
|
||||
heartbeat_alive = (s_voice.heartbeat_task != NULL);
|
||||
ws_text_alive = (s_voice.ws_text_task != NULL);
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
if (!uplink_alive && !heartbeat_alive) {
|
||||
if (!uplink_alive && !heartbeat_alive && !ws_text_alive) {
|
||||
return true;
|
||||
}
|
||||
vTaskDelay(pdMS_TO_TICKS(20));
|
||||
@@ -39,9 +65,10 @@ static bool voice_wait_worker_tasks_exit(uint32_t timeout_ms) {
|
||||
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
ESP_LOGW(TAG,
|
||||
"[stage] worker_task_exit_timeout: uplink_alive=%d heartbeat_alive=%d",
|
||||
"[stage] worker_task_exit_timeout: uplink_alive=%d heartbeat_alive=%d ws_text_alive=%d",
|
||||
s_voice.uplink_task != NULL,
|
||||
s_voice.heartbeat_task != NULL);
|
||||
s_voice.heartbeat_task != NULL,
|
||||
s_voice.ws_text_task != NULL);
|
||||
voice_unlock();
|
||||
}
|
||||
return false;
|
||||
@@ -76,6 +103,14 @@ esp_err_t voice_interaction_init(void) {
|
||||
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
|
||||
esp_err_t mic_key_err = voice_interaction_mic_key_init();
|
||||
if (mic_key_err != ESP_OK) {
|
||||
vSemaphoreDelete(s_voice.lock);
|
||||
memset(&s_voice, 0, sizeof(s_voice));
|
||||
return mic_key_err;
|
||||
}
|
||||
|
||||
s_voice.ready = true;
|
||||
return ESP_OK;
|
||||
}
|
||||
@@ -152,10 +187,14 @@ esp_err_t voice_interaction_start_with_timeout(uint32_t timeout_ms, char *err, s
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.stop_requested = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.ws_low_stack_warned = false;
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
s_voice.uplink_task_with_caps = false;
|
||||
s_voice.heartbeat_task_with_caps = false;
|
||||
s_voice.ws_text_task_with_caps = false;
|
||||
s_voice.ws_text_shutdown = false;
|
||||
s_voice.ws_text_busy = false;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
ESP_LOGI(TAG,
|
||||
"[stage] session_context_prepared: task_id=%s device_uuid=%s",
|
||||
@@ -172,9 +211,12 @@ esp_err_t voice_interaction_start_with_timeout(uint32_t timeout_ms, char *err, s
|
||||
.buffer_size = 2048,
|
||||
.network_timeout_ms = 10000,
|
||||
.reconnect_timeout_ms = 3000,
|
||||
.ping_interval_sec = 20,
|
||||
.ping_interval_sec = 55,
|
||||
.disable_auto_reconnect = false,
|
||||
.pingpong_timeout_sec = 10,
|
||||
// DashScope protocol already requires application-level HeartBeat keepalive.
|
||||
// Some networks may drop/ignore ws-level PONG intermittently, causing false
|
||||
// disconnects if we enforce pingpong timeout aggressively.
|
||||
.disable_pingpong_discon = true,
|
||||
#if CONFIG_MBEDTLS_CERTIFICATE_BUNDLE
|
||||
.crt_bundle_attach = esp_crt_bundle_attach,
|
||||
#endif
|
||||
@@ -236,17 +278,53 @@ esp_err_t voice_interaction_start_with_timeout(uint32_t timeout_ms, char *err, s
|
||||
s_voice.session_active = true;
|
||||
s_voice.ws_connected = false;
|
||||
s_voice.started = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.ws_text_shutdown = false;
|
||||
s_voice.ws_text_busy = false;
|
||||
voice_unlock();
|
||||
voice_log_status_snapshot("session_marked_active");
|
||||
voice_log_heap_snapshot("before_create_uplink");
|
||||
|
||||
ESP_LOGI(TAG, "[stage] uplink_task_create_begin");
|
||||
BaseType_t ok = voice_create_task_prefer_psram(voice_uplink_task,
|
||||
"voice_uplink",
|
||||
VOICE_UPLINK_TASK_STACK,
|
||||
if (!voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
voice_fill_err(err, err_len, "voice lock timeout");
|
||||
ESP_LOGE(TAG, "[stage] session_start_abort: lock timeout before ws_text_task create");
|
||||
(void)voice_interaction_stop(NULL, 0);
|
||||
return ESP_ERR_TIMEOUT;
|
||||
}
|
||||
if (s_voice.ws_text_queue == NULL) {
|
||||
s_voice.ws_text_queue = xQueueCreate(VOICE_WS_TEXT_QUEUE_LEN, sizeof(voice_ws_text_msg_t));
|
||||
}
|
||||
QueueHandle_t ws_text_queue = s_voice.ws_text_queue;
|
||||
voice_unlock();
|
||||
if (ws_text_queue == NULL) {
|
||||
voice_fill_err(err, err_len, "create ws text queue failed");
|
||||
ESP_LOGE(TAG, "[stage] ws_text_queue_create_failed");
|
||||
(void)voice_interaction_stop(NULL, 0);
|
||||
return ESP_ERR_NO_MEM;
|
||||
}
|
||||
|
||||
ESP_LOGI(TAG, "[stage] ws_text_task_create_begin");
|
||||
BaseType_t ok = voice_create_task_prefer_psram(voice_ws_text_task,
|
||||
"voice_ws_text",
|
||||
VOICE_HEARTBEAT_TASK_STACK,
|
||||
5,
|
||||
&s_voice.uplink_task,
|
||||
&s_voice.uplink_task_with_caps);
|
||||
&s_voice.ws_text_task,
|
||||
&s_voice.ws_text_task_with_caps);
|
||||
if (ok != pdPASS) {
|
||||
voice_fill_err(err, err_len, "create ws text task failed");
|
||||
ESP_LOGE(TAG, "[stage] ws_text_task_create_failed");
|
||||
(void)voice_interaction_stop(NULL, 0);
|
||||
return ESP_ERR_NO_MEM;
|
||||
}
|
||||
ESP_LOGI(TAG, "[stage] ws_text_task_create_ok");
|
||||
|
||||
ESP_LOGI(TAG, "[stage] uplink_task_create_begin");
|
||||
ok = voice_create_task_prefer_psram(voice_uplink_task,
|
||||
"voice_uplink",
|
||||
VOICE_UPLINK_TASK_STACK,
|
||||
5,
|
||||
&s_voice.uplink_task,
|
||||
&s_voice.uplink_task_with_caps);
|
||||
if (ok != pdPASS) {
|
||||
voice_fill_err(err, err_len, "create uplink task failed");
|
||||
ESP_LOGE(TAG, "[stage] uplink_task_create_failed");
|
||||
@@ -346,22 +424,35 @@ esp_err_t voice_interaction_stop_with_timeout(uint32_t timeout_ms, char *err, si
|
||||
bool ws_connected = s_voice.ws_connected;
|
||||
TaskHandle_t uplink_task = s_voice.uplink_task;
|
||||
TaskHandle_t heartbeat_task = s_voice.heartbeat_task;
|
||||
TaskHandle_t ws_text_task = s_voice.ws_text_task;
|
||||
QueueHandle_t ws_text_queue = s_voice.ws_text_queue;
|
||||
esp_websocket_client_handle_t ws = s_voice.ws;
|
||||
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
voice_unlock();
|
||||
voice_log_status_snapshot("session_stop_requested");
|
||||
|
||||
if (ws_connected) {
|
||||
(void)voice_send_directive("finish-task", "Stop", true);
|
||||
if (!voice_wait_ws_text_idle(VOICE_WS_SEND_TIMEOUT_MS + 1200)) {
|
||||
ESP_LOGW(TAG, "[stage] ws_text_idle_timeout_before_shutdown");
|
||||
}
|
||||
}
|
||||
voice_close_audio_with_drain(VOICE_AUDIO_DRAIN_WAIT_ON_STOP_MS, "session_stop_close");
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.ws_text_shutdown = true;
|
||||
voice_unlock();
|
||||
}
|
||||
if (uplink_task != NULL) {
|
||||
(void)xTaskAbortDelay(uplink_task);
|
||||
}
|
||||
if (heartbeat_task != NULL) {
|
||||
(void)xTaskAbortDelay(heartbeat_task);
|
||||
}
|
||||
if (ws_text_task != NULL) {
|
||||
(void)xTaskAbortDelay(ws_text_task);
|
||||
}
|
||||
bool workers_exited = voice_wait_worker_tasks_exit(worker_exit_timeout_ms);
|
||||
if (!workers_exited) {
|
||||
voice_fill_err(err, err_len, "voice worker exit timeout");
|
||||
@@ -373,6 +464,11 @@ esp_err_t voice_interaction_stop_with_timeout(uint32_t timeout_ms, char *err, si
|
||||
return ESP_ERR_TIMEOUT;
|
||||
}
|
||||
|
||||
if (ws_text_queue != NULL) {
|
||||
vQueueDelete(ws_text_queue);
|
||||
ws_text_queue = NULL;
|
||||
}
|
||||
|
||||
if (ws != NULL) {
|
||||
ESP_LOGI(TAG, "[stage] ws_stop_destroy_begin");
|
||||
esp_websocket_unregister_events(ws,
|
||||
@@ -385,14 +481,20 @@ esp_err_t voice_interaction_stop_with_timeout(uint32_t timeout_ms, char *err, si
|
||||
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.ws = NULL;
|
||||
s_voice.ws_text_queue = NULL;
|
||||
s_voice.uplink_task = NULL;
|
||||
s_voice.heartbeat_task = NULL;
|
||||
s_voice.ws_text_task = NULL;
|
||||
s_voice.uplink_task_with_caps = false;
|
||||
s_voice.heartbeat_task_with_caps = false;
|
||||
s_voice.ws_text_task_with_caps = false;
|
||||
s_voice.ws_connected = false;
|
||||
s_voice.started = false;
|
||||
s_voice.session_active = false;
|
||||
s_voice.stop_requested = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.ws_text_shutdown = false;
|
||||
s_voice.ws_text_busy = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.dialog_id[0] = '\0';
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
@@ -412,6 +514,60 @@ esp_err_t voice_interaction_stop(char *err, size_t err_len) {
|
||||
return voice_interaction_stop_with_timeout(1500, err, err_len);
|
||||
}
|
||||
|
||||
esp_err_t voice_interaction_try_send_speech(char *err, size_t err_len) {
|
||||
bool should_send = false;
|
||||
if (!voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
voice_fill_err(err, err_len, "voice lock timeout");
|
||||
return ESP_ERR_TIMEOUT;
|
||||
}
|
||||
|
||||
if (!s_voice.session_active || !s_voice.ws_connected || !s_voice.started) {
|
||||
voice_unlock();
|
||||
voice_fill_err(err, err_len, "voice session not ready");
|
||||
return ESP_ERR_INVALID_STATE;
|
||||
}
|
||||
|
||||
should_send = s_voice.tap_active &&
|
||||
!s_voice.speech_active &&
|
||||
s_voice.dialog_state == VOICE_DIALOG_STATE_LISTENING;
|
||||
voice_unlock();
|
||||
|
||||
if (!should_send) {
|
||||
return ESP_OK;
|
||||
}
|
||||
|
||||
esp_err_t send_err = voice_send_directive("continue-task", "SendSpeech", true);
|
||||
if (send_err != ESP_OK) {
|
||||
voice_fill_err(err, err_len, "send speech failed");
|
||||
ESP_LOGW(TAG, "[stage] send_speech_failed: rc=0x%x", (unsigned)send_err);
|
||||
return send_err;
|
||||
}
|
||||
|
||||
bool should_send_stop = false;
|
||||
if (!voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
(void)voice_send_directive("continue-task", "StopSpeech", true);
|
||||
voice_fill_err(err, err_len, "voice lock timeout");
|
||||
return ESP_ERR_TIMEOUT;
|
||||
}
|
||||
if (s_voice.session_active &&
|
||||
s_voice.ws_connected &&
|
||||
s_voice.started &&
|
||||
s_voice.tap_active &&
|
||||
s_voice.dialog_state == VOICE_DIALOG_STATE_LISTENING) {
|
||||
s_voice.speech_active = true;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
} else if (s_voice.session_active && s_voice.ws_connected && s_voice.started) {
|
||||
should_send_stop = true;
|
||||
}
|
||||
voice_unlock();
|
||||
|
||||
if (should_send_stop) {
|
||||
(void)voice_send_directive("continue-task", "StopSpeech", true);
|
||||
}
|
||||
|
||||
return ESP_OK;
|
||||
}
|
||||
|
||||
esp_err_t voice_interaction_tap_start(char *err, size_t err_len) {
|
||||
ESP_LOGI(TAG, "[stage] tap_start_enter");
|
||||
if (!voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
@@ -461,11 +617,33 @@ esp_err_t voice_interaction_tap_start(char *err, size_t err_len) {
|
||||
|
||||
bool listening = (s_voice.dialog_state == VOICE_DIALOG_STATE_LISTENING);
|
||||
s_voice.tap_active = true;
|
||||
s_voice.speech_active = false;
|
||||
voice_unlock();
|
||||
ESP_LOGI(TAG, "[stage] tap_start_marked_active: listening=%d", listening);
|
||||
|
||||
if (!listening) {
|
||||
(void)voice_send_directive("continue-task", "RequestToSpeak", true);
|
||||
esp_err_t req_rc = voice_send_directive("continue-task", "RequestToSpeak", true);
|
||||
if (req_rc != ESP_OK) {
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
voice_unlock();
|
||||
}
|
||||
voice_fill_err(err, err_len, "request to speak failed");
|
||||
ESP_LOGW(TAG, "[stage] tap_start_abort: request_to_speak failed rc=0x%x", (unsigned)req_rc);
|
||||
return req_rc;
|
||||
}
|
||||
} else {
|
||||
esp_err_t send_speech_err = voice_interaction_try_send_speech(err, err_len);
|
||||
if (send_speech_err != ESP_OK) {
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
voice_unlock();
|
||||
}
|
||||
voice_audio_close();
|
||||
return send_speech_err;
|
||||
}
|
||||
}
|
||||
|
||||
voice_log_status_snapshot("tap_start_exit_ok");
|
||||
@@ -491,12 +669,21 @@ esp_err_t voice_interaction_tap_cancel(char *err, size_t err_len) {
|
||||
return ESP_ERR_INVALID_STATE;
|
||||
}
|
||||
|
||||
bool should_send_stop = s_voice.speech_active;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
voice_unlock();
|
||||
voice_log_status_snapshot("tap_cancel_marked_inactive");
|
||||
esp_err_t cancel_err = voice_send_directive("continue-task", "CancelSpeech", true);
|
||||
voice_close_audio_with_drain(VOICE_AUDIO_DRAIN_WAIT_ON_STOP_MS, "tap_cancel_close");
|
||||
return cancel_err;
|
||||
|
||||
esp_err_t stop_err = ESP_OK;
|
||||
if (should_send_stop) {
|
||||
stop_err = voice_send_directive("continue-task", "StopSpeech", true);
|
||||
if (stop_err != ESP_OK) {
|
||||
// Fail-safe close to avoid leaving audio device open in abnormal flow.
|
||||
voice_close_audio_with_drain(VOICE_AUDIO_DRAIN_WAIT_MS, "tap_cancel_stop_failed_close");
|
||||
}
|
||||
}
|
||||
return stop_err;
|
||||
}
|
||||
|
||||
void voice_interaction_get_status(voice_interaction_status_t *out_status) {
|
||||
|
||||
@@ -54,12 +54,14 @@ void voice_log_status_snapshot(const char *stage) {
|
||||
}
|
||||
|
||||
ESP_LOGI(TAG,
|
||||
"[stage] %s: active=%d ws=%d started=%d tap=%d state=%s up=%" PRIu32 " down=%" PRIu32 " err=%s",
|
||||
"[stage] %s: active=%d ws=%d started=%d tap=%d speech=%d state=%s up=%" PRIu32
|
||||
" down=%" PRIu32 " err=%s",
|
||||
stage,
|
||||
s_voice.session_active,
|
||||
s_voice.ws_connected,
|
||||
s_voice.started,
|
||||
s_voice.tap_active,
|
||||
s_voice.speech_active,
|
||||
voice_interaction_dialog_state_str(s_voice.dialog_state),
|
||||
s_voice.upstream_packets,
|
||||
s_voice.downstream_packets,
|
||||
|
||||
227
components/domain/src/voice_interaction_mic_key.c
Normal file
227
components/domain/src/voice_interaction_mic_key.c
Normal file
@@ -0,0 +1,227 @@
|
||||
#include "voice_interaction_internal.h"
|
||||
|
||||
#include <inttypes.h>
|
||||
|
||||
#include "esp_log.h"
|
||||
#include "freertos/queue.h"
|
||||
|
||||
#define VOICE_MIC_KEY_TASK_STACK 6144
|
||||
#define VOICE_MIC_KEY_QUEUE_LEN 8
|
||||
#define VOICE_MIC_KEY_SESSION_READY_WAIT_MS 2000
|
||||
#define VOICE_MIC_KEY_SESSION_READY_POLL_MS 30
|
||||
|
||||
typedef enum {
|
||||
VOICE_MIC_KEY_CMD_PRESS_DOWN = 0,
|
||||
VOICE_MIC_KEY_CMD_PRESS_UP,
|
||||
} voice_mic_key_cmd_t;
|
||||
|
||||
static const char *TAG = "voice_mic_key";
|
||||
|
||||
static QueueHandle_t s_mic_key_queue;
|
||||
static TaskHandle_t s_mic_key_task;
|
||||
static bool s_mic_key_task_with_caps;
|
||||
static bool s_mic_key_ready;
|
||||
|
||||
static bool voice_mic_key_round_busy(const voice_interaction_status_t *voice,
|
||||
const printer_runtime_status_t *printer) {
|
||||
if (voice == NULL || printer == NULL) {
|
||||
return false;
|
||||
}
|
||||
|
||||
bool marker_busy = voice->marker_image_gen_running;
|
||||
bool print_busy = printer->busy || printer->queue_depth > 0;
|
||||
return marker_busy || print_busy;
|
||||
}
|
||||
|
||||
static bool voice_mic_key_session_ready(const voice_interaction_status_t *voice) {
|
||||
return voice != NULL && voice->session_active && voice->ws_connected && voice->started;
|
||||
}
|
||||
|
||||
static bool voice_mic_key_wait_session_ready(uint32_t timeout_ms) {
|
||||
int64_t deadline_ms = voice_now_ms() + (int64_t)timeout_ms;
|
||||
while (voice_now_ms() < deadline_ms) {
|
||||
voice_interaction_status_t voice = {0};
|
||||
voice_interaction_get_status(&voice);
|
||||
if (voice_mic_key_session_ready(&voice)) {
|
||||
return true;
|
||||
}
|
||||
vTaskDelay(pdMS_TO_TICKS(VOICE_MIC_KEY_SESSION_READY_POLL_MS));
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
static void voice_mic_key_handle_press_down(void) {
|
||||
voice_interaction_status_t voice = {0};
|
||||
voice_interaction_get_status(&voice);
|
||||
|
||||
if (voice.tap_active) {
|
||||
ESP_LOGI(TAG, "MIC_KEY press ignored: tap already active");
|
||||
return;
|
||||
}
|
||||
|
||||
printer_runtime_status_t printer = {0};
|
||||
printer_protocol_get_runtime_status(&printer);
|
||||
if (voice_mic_key_round_busy(&voice, &printer)) {
|
||||
ESP_LOGI(TAG,
|
||||
"MIC_KEY press ignored: state=%s marker_busy=%d print_busy=%d queue=%" PRIu32,
|
||||
voice_interaction_dialog_state_str(voice.dialog_state),
|
||||
voice.marker_image_gen_running,
|
||||
printer.busy,
|
||||
printer.queue_depth);
|
||||
return;
|
||||
}
|
||||
|
||||
if (!voice_mic_key_session_ready(&voice)) {
|
||||
char start_err[128] = {0};
|
||||
esp_err_t start_rc = voice_interaction_start_with_timeout(0, start_err, sizeof(start_err));
|
||||
if (start_rc != ESP_OK) {
|
||||
ESP_LOGW(TAG,
|
||||
"MIC_KEY auto session start failed: %s (%s)",
|
||||
esp_err_to_name(start_rc),
|
||||
start_err[0] != '\0' ? start_err : "-");
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
if (!voice_mic_key_wait_session_ready(VOICE_MIC_KEY_SESSION_READY_WAIT_MS)) {
|
||||
voice_interaction_get_status(&voice);
|
||||
ESP_LOGW(TAG,
|
||||
"MIC_KEY wait session ready timeout: active=%d ws=%d started=%d",
|
||||
voice.session_active,
|
||||
voice.ws_connected,
|
||||
voice.started);
|
||||
return;
|
||||
}
|
||||
|
||||
char tap_err[128] = {0};
|
||||
esp_err_t tap_rc = voice_interaction_tap_start(tap_err, sizeof(tap_err));
|
||||
if (tap_rc != ESP_OK) {
|
||||
ESP_LOGW(TAG,
|
||||
"MIC_KEY tap start failed: %s (%s)",
|
||||
esp_err_to_name(tap_rc),
|
||||
tap_err[0] != '\0' ? tap_err : "-");
|
||||
return;
|
||||
}
|
||||
|
||||
ESP_LOGI(TAG, "MIC_KEY tap started");
|
||||
}
|
||||
|
||||
static void voice_mic_key_handle_press_up(void) {
|
||||
voice_interaction_status_t voice = {0};
|
||||
voice_interaction_get_status(&voice);
|
||||
if (!voice.tap_active) {
|
||||
return;
|
||||
}
|
||||
|
||||
char tap_err[128] = {0};
|
||||
esp_err_t tap_rc = voice_interaction_tap_cancel(tap_err, sizeof(tap_err));
|
||||
if (tap_rc != ESP_OK) {
|
||||
ESP_LOGW(TAG,
|
||||
"MIC_KEY tap stop failed: %s (%s)",
|
||||
esp_err_to_name(tap_rc),
|
||||
tap_err[0] != '\0' ? tap_err : "-");
|
||||
return;
|
||||
}
|
||||
|
||||
ESP_LOGI(TAG, "MIC_KEY tap stopped");
|
||||
}
|
||||
|
||||
static void voice_mic_key_task_main(void *arg) {
|
||||
(void)arg;
|
||||
|
||||
while (true) {
|
||||
voice_mic_key_cmd_t cmd = VOICE_MIC_KEY_CMD_PRESS_UP;
|
||||
if (xQueueReceive(s_mic_key_queue, &cmd, portMAX_DELAY) != pdTRUE) {
|
||||
continue;
|
||||
}
|
||||
|
||||
switch (cmd) {
|
||||
case VOICE_MIC_KEY_CMD_PRESS_DOWN:
|
||||
voice_mic_key_handle_press_down();
|
||||
break;
|
||||
case VOICE_MIC_KEY_CMD_PRESS_UP:
|
||||
voice_mic_key_handle_press_up();
|
||||
break;
|
||||
default:
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
static void voice_mic_key_platform_event_cb(platform_mic_key_event_t event, void *user_data) {
|
||||
(void)user_data;
|
||||
if (s_mic_key_queue == NULL) {
|
||||
return;
|
||||
}
|
||||
|
||||
voice_mic_key_cmd_t cmd = VOICE_MIC_KEY_CMD_PRESS_UP;
|
||||
switch (event) {
|
||||
case PLATFORM_MIC_KEY_EVENT_PRESS_DOWN:
|
||||
cmd = VOICE_MIC_KEY_CMD_PRESS_DOWN;
|
||||
break;
|
||||
case PLATFORM_MIC_KEY_EVENT_PRESS_UP:
|
||||
cmd = VOICE_MIC_KEY_CMD_PRESS_UP;
|
||||
break;
|
||||
default:
|
||||
return;
|
||||
}
|
||||
|
||||
BaseType_t queued = xQueueSend(s_mic_key_queue, &cmd, 0);
|
||||
if (queued == pdTRUE) {
|
||||
return;
|
||||
}
|
||||
|
||||
voice_mic_key_cmd_t dropped = VOICE_MIC_KEY_CMD_PRESS_UP;
|
||||
(void)xQueueReceive(s_mic_key_queue, &dropped, 0);
|
||||
(void)xQueueSend(s_mic_key_queue, &cmd, 0);
|
||||
}
|
||||
|
||||
esp_err_t voice_interaction_mic_key_init(void) {
|
||||
if (s_mic_key_ready) {
|
||||
return ESP_OK;
|
||||
}
|
||||
|
||||
esp_err_t err = platform_mic_key_init();
|
||||
if (err == ESP_ERR_NOT_SUPPORTED) {
|
||||
ESP_LOGW(TAG, "MIC_KEY trigger disabled by board config");
|
||||
s_mic_key_ready = true;
|
||||
return ESP_OK;
|
||||
}
|
||||
if (err != ESP_OK) {
|
||||
return err;
|
||||
}
|
||||
|
||||
s_mic_key_queue = xQueueCreate(VOICE_MIC_KEY_QUEUE_LEN, sizeof(voice_mic_key_cmd_t));
|
||||
if (s_mic_key_queue == NULL) {
|
||||
platform_mic_key_deinit();
|
||||
return ESP_ERR_NO_MEM;
|
||||
}
|
||||
|
||||
BaseType_t ok = voice_create_task_prefer_psram(voice_mic_key_task_main,
|
||||
"voice_mic_key",
|
||||
VOICE_MIC_KEY_TASK_STACK,
|
||||
4,
|
||||
&s_mic_key_task,
|
||||
&s_mic_key_task_with_caps);
|
||||
if (ok != pdPASS) {
|
||||
vQueueDelete(s_mic_key_queue);
|
||||
s_mic_key_queue = NULL;
|
||||
platform_mic_key_deinit();
|
||||
return ESP_ERR_NO_MEM;
|
||||
}
|
||||
|
||||
err = platform_mic_key_set_event_callback(voice_mic_key_platform_event_cb, NULL);
|
||||
if (err != ESP_OK) {
|
||||
voice_delete_task(s_mic_key_task, s_mic_key_task_with_caps);
|
||||
s_mic_key_task = NULL;
|
||||
s_mic_key_task_with_caps = false;
|
||||
vQueueDelete(s_mic_key_queue);
|
||||
s_mic_key_queue = NULL;
|
||||
platform_mic_key_deinit();
|
||||
return err;
|
||||
}
|
||||
|
||||
s_mic_key_ready = true;
|
||||
ESP_LOGI(TAG, "MIC_KEY trigger ready: press to speak, release to process");
|
||||
return ESP_OK;
|
||||
}
|
||||
@@ -27,6 +27,7 @@ void voice_uplink_task(void *arg) {
|
||||
bool connected = false;
|
||||
bool started = false;
|
||||
bool tap_active = false;
|
||||
bool speech_active = false;
|
||||
bool listening = false;
|
||||
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
@@ -34,6 +35,7 @@ void voice_uplink_task(void *arg) {
|
||||
connected = s_voice.ws_connected;
|
||||
started = s_voice.started;
|
||||
tap_active = s_voice.tap_active;
|
||||
speech_active = s_voice.speech_active;
|
||||
listening = (s_voice.dialog_state == VOICE_DIALOG_STATE_LISTENING);
|
||||
voice_unlock();
|
||||
}
|
||||
@@ -42,7 +44,7 @@ void voice_uplink_task(void *arg) {
|
||||
break;
|
||||
}
|
||||
|
||||
if (!(connected && started && listening && tap_active)) {
|
||||
if (!(connected && started && listening && tap_active && speech_active)) {
|
||||
vTaskDelay(pdMS_TO_TICKS(20));
|
||||
continue;
|
||||
}
|
||||
@@ -83,6 +85,7 @@ void voice_uplink_task(void *arg) {
|
||||
s_voice.ws_connected = false;
|
||||
s_voice.started = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
ESP_LOGW(TAG, "[stage] uplink_detected_ws_disconnected");
|
||||
|
||||
@@ -88,7 +88,7 @@ static char *voice_build_start_message(void) {
|
||||
cJSON *parameters = cJSON_AddObjectToObject(payload, "parameters");
|
||||
cJSON *upstream = cJSON_AddObjectToObject(parameters, "upstream");
|
||||
cJSON_AddStringToObject(upstream, "type", "AudioOnly");
|
||||
cJSON_AddStringToObject(upstream, "mode", "tap2talk");
|
||||
cJSON_AddStringToObject(upstream, "mode", "push2talk");
|
||||
cJSON_AddStringToObject(upstream, "audio_format", "raw-opus");
|
||||
cJSON_AddNumberToObject(upstream, "sample_rate", CONFIG_TQ_VOICE_SAMPLE_RATE);
|
||||
|
||||
@@ -113,7 +113,7 @@ static char *voice_build_start_message(void) {
|
||||
return text;
|
||||
}
|
||||
|
||||
static esp_err_t voice_send_text(const char *text, int len) {
|
||||
static esp_err_t voice_send_text_immediate(const char *text, int len) {
|
||||
if (text == NULL || len <= 0) {
|
||||
return ESP_ERR_INVALID_ARG;
|
||||
}
|
||||
@@ -132,6 +132,7 @@ static esp_err_t voice_send_text(const char *text, int len) {
|
||||
s_voice.ws_connected = false;
|
||||
s_voice.started = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
ESP_LOGW(TAG, "[stage] ws_state_reconciled: transport disconnected");
|
||||
@@ -156,6 +157,43 @@ static esp_err_t voice_send_text(const char *text, int len) {
|
||||
return ESP_OK;
|
||||
}
|
||||
|
||||
static esp_err_t voice_send_text(const char *text, int len) {
|
||||
if (text == NULL || len <= 0) {
|
||||
return ESP_ERR_INVALID_ARG;
|
||||
}
|
||||
|
||||
QueueHandle_t queue = NULL;
|
||||
TaskHandle_t ws_text_task = NULL;
|
||||
bool session_active = false;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
queue = s_voice.ws_text_queue;
|
||||
ws_text_task = s_voice.ws_text_task;
|
||||
session_active = s_voice.session_active;
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
if (queue == NULL || ws_text_task == NULL || !session_active) {
|
||||
return voice_send_text_immediate(text, len);
|
||||
}
|
||||
|
||||
char *copy = (char *)voice_malloc_prefer_psram((size_t)len + 1);
|
||||
if (copy == NULL) {
|
||||
return ESP_ERR_NO_MEM;
|
||||
}
|
||||
memcpy(copy, text, (size_t)len);
|
||||
copy[len] = '\0';
|
||||
|
||||
voice_ws_text_msg_t msg = {
|
||||
.text = copy,
|
||||
.len = len,
|
||||
};
|
||||
if (xQueueSend(queue, &msg, pdMS_TO_TICKS(VOICE_WS_SEND_TIMEOUT_MS)) != pdTRUE) {
|
||||
free(copy);
|
||||
return ESP_ERR_TIMEOUT;
|
||||
}
|
||||
return ESP_OK;
|
||||
}
|
||||
|
||||
esp_err_t voice_send_directive(const char *action,
|
||||
const char *directive,
|
||||
bool with_dialog_id) {
|
||||
@@ -181,6 +219,68 @@ esp_err_t voice_send_directive(const char *action,
|
||||
return err;
|
||||
}
|
||||
|
||||
void voice_ws_text_task(void *arg) {
|
||||
(void)arg;
|
||||
|
||||
while (true) {
|
||||
bool shutdown = false;
|
||||
QueueHandle_t queue = NULL;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
shutdown = s_voice.ws_text_shutdown;
|
||||
queue = s_voice.ws_text_queue;
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
if (shutdown || queue == NULL) {
|
||||
break;
|
||||
}
|
||||
|
||||
voice_ws_text_msg_t msg = {0};
|
||||
if (xQueueReceive(queue, &msg, pdMS_TO_TICKS(VOICE_WS_TEXT_QUEUE_RX_TIMEOUT_MS)) != pdTRUE) {
|
||||
continue;
|
||||
}
|
||||
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.ws_text_busy = true;
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
esp_err_t err = voice_send_text_immediate(msg.text, msg.len);
|
||||
if (err != ESP_OK && voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
voice_set_last_error_locked("ws text send failed");
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
s_voice.ws_text_busy = false;
|
||||
voice_unlock();
|
||||
}
|
||||
free(msg.text);
|
||||
}
|
||||
|
||||
QueueHandle_t queue = NULL;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
queue = s_voice.ws_text_queue;
|
||||
s_voice.ws_text_busy = false;
|
||||
voice_unlock();
|
||||
}
|
||||
if (queue != NULL) {
|
||||
voice_ws_text_msg_t msg = {0};
|
||||
while (xQueueReceive(queue, &msg, 0) == pdTRUE) {
|
||||
free(msg.text);
|
||||
}
|
||||
}
|
||||
|
||||
bool delete_with_caps = false;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
delete_with_caps = s_voice.ws_text_task_with_caps;
|
||||
s_voice.ws_text_task = NULL;
|
||||
s_voice.ws_text_task_with_caps = false;
|
||||
voice_unlock();
|
||||
}
|
||||
voice_delete_self_task(delete_with_caps);
|
||||
}
|
||||
|
||||
static void voice_handle_text_chunk(esp_websocket_event_data_t *data) {
|
||||
if (data->payload_offset == 0) {
|
||||
free(s_voice.text_agg);
|
||||
@@ -310,6 +410,7 @@ void voice_websocket_event_handler(void *handler_args,
|
||||
s_voice.ws_connected = false;
|
||||
s_voice.started = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.last_event_ms = voice_now_ms();
|
||||
voice_unlock();
|
||||
|
||||
@@ -71,16 +71,24 @@ void voice_ws_audio_handle_downstream_packet(const uint8_t *data, size_t len) {
|
||||
VOICE_AUDIO_IO_TIMEOUT_MS,
|
||||
audio_err,
|
||||
sizeof(audio_err));
|
||||
if (write_err == ESP_OK && voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
int64_t now_ms = voice_now_ms();
|
||||
int64_t pcm_ms = ((int64_t)samples * 1000 +
|
||||
(int64_t)CONFIG_TQ_VOICE_SAMPLE_RATE - 1) /
|
||||
(int64_t)CONFIG_TQ_VOICE_SAMPLE_RATE;
|
||||
int64_t base_ms = s_voice.playback_deadline_ms > now_ms ?
|
||||
s_voice.playback_deadline_ms : now_ms;
|
||||
s_voice.playback_deadline_ms = base_ms + pcm_ms + VOICE_AUDIO_DRAIN_MARGIN_MS;
|
||||
s_voice.last_downstream_ms = now_ms;
|
||||
voice_unlock();
|
||||
if (write_err == ESP_OK) {
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
int64_t now_ms = voice_now_ms();
|
||||
int64_t pcm_ms = ((int64_t)samples * 1000 +
|
||||
(int64_t)CONFIG_TQ_VOICE_SAMPLE_RATE - 1) /
|
||||
(int64_t)CONFIG_TQ_VOICE_SAMPLE_RATE;
|
||||
int64_t base_ms = s_voice.playback_deadline_ms > now_ms ?
|
||||
s_voice.playback_deadline_ms : now_ms;
|
||||
s_voice.playback_deadline_ms = base_ms + pcm_ms + VOICE_AUDIO_DRAIN_MARGIN_MS;
|
||||
s_voice.last_downstream_ms = now_ms;
|
||||
voice_unlock();
|
||||
}
|
||||
} else {
|
||||
ESP_LOGW(TAG,
|
||||
"[stage] audio_downstream_write_failed: rc=0x%x err=%s samples=%u",
|
||||
(unsigned)write_err,
|
||||
audio_err[0] != '\0' ? audio_err : "-",
|
||||
(unsigned)samples);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -11,6 +11,111 @@
|
||||
|
||||
static const char *TAG = "voice_interaction";
|
||||
|
||||
#define VOICE_RESPONSE_FINALIZE_TASK_STACK 6144
|
||||
#define VOICE_RESPONSE_FINALIZE_TASK_RETRY_COUNT 3
|
||||
#define VOICE_RESPONSE_FINALIZE_TASK_RETRY_DELAY_MS 120
|
||||
|
||||
static portMUX_TYPE s_response_finalize_lock = portMUX_INITIALIZER_UNLOCKED;
|
||||
static bool s_response_finalize_running;
|
||||
|
||||
static esp_err_t voice_send_directive_with_retry(const char *action,
|
||||
const char *directive,
|
||||
bool with_dialog_id) {
|
||||
esp_err_t rc = ESP_FAIL;
|
||||
for (uint32_t i = 0; i < VOICE_RESPONSE_FINALIZE_TASK_RETRY_COUNT; ++i) {
|
||||
rc = voice_send_directive(action, directive, with_dialog_id);
|
||||
if (rc == ESP_OK || rc == ESP_ERR_INVALID_STATE) {
|
||||
return rc;
|
||||
}
|
||||
vTaskDelay(pdMS_TO_TICKS(VOICE_RESPONSE_FINALIZE_TASK_RETRY_DELAY_MS));
|
||||
}
|
||||
return rc;
|
||||
}
|
||||
|
||||
static uint32_t voice_calc_response_finalize_wait_ms(void) {
|
||||
uint32_t playback_wait_ms = VOICE_AUDIO_DRAIN_WAIT_MS;
|
||||
if (!voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
return playback_wait_ms;
|
||||
}
|
||||
|
||||
int64_t now_ms = voice_now_ms();
|
||||
int64_t deadline_ms = s_voice.playback_deadline_ms;
|
||||
voice_unlock();
|
||||
|
||||
if (deadline_ms <= now_ms) {
|
||||
return playback_wait_ms;
|
||||
}
|
||||
|
||||
int64_t remain_ms = (deadline_ms - now_ms) + VOICE_AUDIO_DRAIN_QUIET_MS + 120;
|
||||
if (remain_ms <= (int64_t)playback_wait_ms) {
|
||||
return playback_wait_ms;
|
||||
}
|
||||
|
||||
return (remain_ms > 12000) ? 12000 : (uint32_t)remain_ms;
|
||||
}
|
||||
|
||||
static void voice_response_finalize_task(void *arg) {
|
||||
(void)arg;
|
||||
|
||||
uint32_t playback_wait_ms = voice_calc_response_finalize_wait_ms();
|
||||
ESP_LOGI(TAG,
|
||||
"[stage] responding_ended_local_playback_wait: timeout_ms=%u",
|
||||
(unsigned)playback_wait_ms);
|
||||
voice_wait_audio_playback_done(playback_wait_ms, "responding_ended_local_playback");
|
||||
|
||||
esp_err_t local_end_rc =
|
||||
voice_send_directive_with_retry("continue-task", "LocalRespondingEnded", true);
|
||||
if (local_end_rc != ESP_OK) {
|
||||
ESP_LOGW(TAG,
|
||||
"[stage] responding_ended_local_end_failed: rc=0x%x",
|
||||
(unsigned)local_end_rc);
|
||||
}
|
||||
|
||||
esp_err_t stop_rc =
|
||||
voice_send_directive_with_retry("finish-task", "Stop", true);
|
||||
if (stop_rc != ESP_OK) {
|
||||
ESP_LOGW(TAG,
|
||||
"[stage] responding_ended_stop_failed: rc=0x%x",
|
||||
(unsigned)stop_rc);
|
||||
}
|
||||
|
||||
portENTER_CRITICAL(&s_response_finalize_lock);
|
||||
s_response_finalize_running = false;
|
||||
portEXIT_CRITICAL(&s_response_finalize_lock);
|
||||
|
||||
vTaskDelete(NULL);
|
||||
}
|
||||
|
||||
static void voice_schedule_response_finalize(void) {
|
||||
bool should_start = false;
|
||||
portENTER_CRITICAL(&s_response_finalize_lock);
|
||||
if (!s_response_finalize_running) {
|
||||
s_response_finalize_running = true;
|
||||
should_start = true;
|
||||
}
|
||||
portEXIT_CRITICAL(&s_response_finalize_lock);
|
||||
|
||||
if (!should_start) {
|
||||
ESP_LOGI(TAG, "[stage] responding_ended_finalize_skip: task already running");
|
||||
return;
|
||||
}
|
||||
|
||||
BaseType_t ok = xTaskCreate(voice_response_finalize_task,
|
||||
"voice_rsp_fin",
|
||||
VOICE_RESPONSE_FINALIZE_TASK_STACK,
|
||||
NULL,
|
||||
4,
|
||||
NULL);
|
||||
if (ok != pdPASS) {
|
||||
portENTER_CRITICAL(&s_response_finalize_lock);
|
||||
s_response_finalize_running = false;
|
||||
portEXIT_CRITICAL(&s_response_finalize_lock);
|
||||
ESP_LOGW(TAG, "[stage] responding_ended_finalize_start_failed");
|
||||
} else {
|
||||
ESP_LOGI(TAG, "[stage] responding_ended_finalize_task_started");
|
||||
}
|
||||
}
|
||||
|
||||
static const char *voice_get_json_str(cJSON *obj, const char *name) {
|
||||
if (obj == NULL || name == NULL) {
|
||||
return NULL;
|
||||
@@ -73,6 +178,7 @@ static void voice_handle_output_event(cJSON *output) {
|
||||
bool has_finished = voice_get_json_bool(output, "finished", &finished);
|
||||
const char *responding_final_text = NULL;
|
||||
bool should_check_marker_in_first_sentence = false;
|
||||
bool should_try_send_speech = false;
|
||||
if (has_finished && finished) {
|
||||
if (strcmp(event_name, "SpeechContent") == 0) {
|
||||
voice_log_final_text("asr_final_text", voice_get_json_str(output, "text"));
|
||||
@@ -114,13 +220,16 @@ static void voice_handle_output_event(cJSON *output) {
|
||||
if (prev_state != VOICE_DIALOG_STATE_LISTENING) {
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
}
|
||||
should_try_send_speech = s_voice.tap_active && !s_voice.speech_active;
|
||||
} else if (strcmp(state, "Thinking") == 0) {
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_THINKING;
|
||||
s_voice.speech_active = false;
|
||||
if (prev_state != VOICE_DIALOG_STATE_THINKING) {
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
}
|
||||
} else if (strcmp(state, "Responding") == 0) {
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_RESPONDING;
|
||||
s_voice.speech_active = false;
|
||||
if (prev_state != VOICE_DIALOG_STATE_RESPONDING) {
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
}
|
||||
@@ -131,14 +240,14 @@ static void voice_handle_output_event(cJSON *output) {
|
||||
state != NULL ? state : "-",
|
||||
voice_interaction_dialog_state_str(s_voice.dialog_state));
|
||||
} else if (strcmp(event_name, "SpeechEnded") == 0) {
|
||||
s_voice.tap_active = false;
|
||||
ESP_LOGI(TAG, "[stage] speech_ended: mark single_round_input_done");
|
||||
s_voice.speech_active = false;
|
||||
ESP_LOGI(TAG, "[stage] speech_ended: stop uplink");
|
||||
} else if (strcmp(event_name, "RespondingEnded") == 0) {
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
ESP_LOGI(TAG, "[stage] responding_ended: wait_local_playback_before_ack");
|
||||
ESP_LOGI(TAG, "[stage] responding_ended: schedule_local_playback_finalize");
|
||||
} else if (strcmp(event_name, "Stopped") == 0) {
|
||||
s_voice.started = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
should_close_audio = true;
|
||||
@@ -151,6 +260,7 @@ static void voice_handle_output_event(cJSON *output) {
|
||||
}
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.response_first_sentence_checked = false;
|
||||
should_close_audio = true;
|
||||
}
|
||||
@@ -165,31 +275,21 @@ static void voice_handle_output_event(cJSON *output) {
|
||||
voice_marker_try_trigger_image_generation(dialog_id, responding_final_text);
|
||||
}
|
||||
|
||||
if (should_try_send_speech) {
|
||||
char speech_err[96] = {0};
|
||||
esp_err_t rc = voice_interaction_try_send_speech(speech_err, sizeof(speech_err));
|
||||
if (rc != ESP_OK && rc != ESP_ERR_INVALID_STATE) {
|
||||
ESP_LOGW(TAG,
|
||||
"[stage] auto_send_speech_failed: rc=0x%x err=%s",
|
||||
(unsigned)rc,
|
||||
speech_err[0] != '\0' ? speech_err : "-");
|
||||
}
|
||||
}
|
||||
|
||||
if (strcmp(event_name, "RespondingStarted") == 0) {
|
||||
(void)voice_send_directive("continue-task", "LocalRespondingStarted", true);
|
||||
} else if (strcmp(event_name, "RespondingEnded") == 0) {
|
||||
uint32_t playback_wait_ms = VOICE_AUDIO_DRAIN_WAIT_MS;
|
||||
if (voice_lock(VOICE_STATUS_LOCK_TIMEOUT_MS)) {
|
||||
int64_t now_ms = voice_now_ms();
|
||||
int64_t deadline_ms = s_voice.playback_deadline_ms;
|
||||
voice_unlock();
|
||||
|
||||
if (deadline_ms > now_ms) {
|
||||
int64_t remain_ms = (deadline_ms - now_ms) + VOICE_AUDIO_DRAIN_QUIET_MS + 120;
|
||||
if (remain_ms > (int64_t)playback_wait_ms) {
|
||||
if (remain_ms > 12000) {
|
||||
playback_wait_ms = 12000;
|
||||
} else {
|
||||
playback_wait_ms = (uint32_t)remain_ms;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
ESP_LOGI(TAG, "[stage] responding_ended_local_playback_wait: timeout_ms=%u", (unsigned)playback_wait_ms);
|
||||
voice_wait_audio_playback_done(playback_wait_ms,
|
||||
"responding_ended_local_playback");
|
||||
(void)voice_send_directive("continue-task", "LocalRespondingEnded", true);
|
||||
(void)voice_send_directive("finish-task", "Stop", true);
|
||||
voice_schedule_response_finalize();
|
||||
}
|
||||
}
|
||||
|
||||
@@ -243,6 +343,7 @@ void voice_ws_protocol_handle_text_message(const char *text, size_t len) {
|
||||
voice_set_last_error_locked(error_msg != NULL ? error_msg : "task failed");
|
||||
s_voice.started = false;
|
||||
s_voice.tap_active = false;
|
||||
s_voice.speech_active = false;
|
||||
s_voice.dialog_state = VOICE_DIALOG_STATE_IDLE;
|
||||
voice_unlock();
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user