语音打印

This commit is contained in:
admin
2026-02-25 18:41:48 +08:00
parent 1cb1f5de64
commit 3f11e48eca
3 changed files with 259 additions and 4 deletions

View File

@@ -1,4 +1,14 @@
dependencies:
espressif/cmake_utilities:
component_hash: 351350613ceafba240b761b4ea991e0f231ac7a9f59a9ee901f751bddc0bb18f
dependencies:
- name: idf
require: private
version: '>=4.1'
source:
registry_url: https://components.espressif.com
type: service
version: 0.5.3
espressif/esp_audio_codec:
component_hash: af4663830bc73eca58febcc34a976d003465e78ed57458aa00a14b231e392ca1
dependencies:
@@ -19,6 +29,20 @@ dependencies:
registry_url: https://components.espressif.com/
type: service
version: 1.5.4
espressif/esp_lcd_gc9a01:
component_hash: e60bc2b5649c7679bd010797a875ae64b2a57cf9cdbd6a2bf7a0ade66913edc0
dependencies:
- name: espressif/cmake_utilities
registry_url: https://components.espressif.com
require: private
version: 0.*
- name: idf
require: private
version: '>=4.4'
source:
registry_url: https://components.espressif.com/
type: service
version: 2.0.4
espressif/esp_websocket_client:
component_hash: c5a067a9fddea370c478017e66fac302f4b79c3d4027e9bdd42a019786cceb92
dependencies:
@@ -60,9 +84,10 @@ dependencies:
direct_dependencies:
- espressif/esp_audio_codec
- espressif/esp_codec_dev
- espressif/esp_lcd_gc9a01
- espressif/esp_websocket_client
- espressif/libpng
- idf
manifest_hash: 3274b53ce89a58abbad0a03dfaff3410e35deb81305a21289606c4d2177a111b
manifest_hash: ee0dd489986ab012202f5537ca2249a0c4e3ca234a23194dddb81cc25075b54c
target: esp32s3
version: 2.0.0

View File

@@ -15,6 +15,8 @@
#include "esp_audio_enc.h"
#include "esp_log.h"
#include "image_generation.h"
#include "printer_protocol.h"
#include "raster_tools.h"
#include "voice_audio.h"
static const char *TAG = "voice_interaction";
@@ -23,12 +25,78 @@ static const char *TAG = "voice_interaction";
#define VOICE_IMAGE_MARKER_FIRST_SENTENCE_MAX_LEN 1024
#define VOICE_IMAGE_GEN_TASK_STACK 16384
#define VOICE_IMAGE_GEN_TASK_PRIO 3
#define VOICE_IMAGE_MARKER_REPEAT_SUPPRESS_MS 45000
#define VOICE_IMAGE_PRINT_THRESHOLD 160
#define VOICE_IMAGE_PRINT_MAX_HEIGHT 2200
#define VOICE_IMAGE_MARKER_TEMPLATE \
"%s黑白线稿风格白色背景主体居中边缘清晰细节简洁高对比避免大面积阴影和灰度渐变适配热敏打印机打印。"
#define VOICE_IMAGE_MARKER_TEMPLATE_FALLBACK_SUBJECT "简洁卡通图案"
typedef struct {
char prompt[VOICE_IMAGE_MARKER_PROMPT_MAX_LEN + 1];
char dialog_id[40];
} voice_image_generation_task_arg_t;
static portMUX_TYPE s_marker_image_gen_lock = portMUX_INITIALIZER_UNLOCKED;
static bool s_marker_image_gen_running;
static int64_t s_marker_last_success_ms;
static char s_marker_last_success_prompt[VOICE_IMAGE_MARKER_PROMPT_MAX_LEN + 1];
static char s_marker_last_success_dialog_id[40];
static bool voice_marker_image_generation_try_acquire(const char *prompt,
const char *dialog_id,
int64_t now_ms,
uint32_t *out_remain_ms) {
bool acquired = false;
uint32_t remain_ms = 0;
const char *safe_prompt = prompt != NULL ? prompt : "";
const char *safe_dialog = dialog_id != NULL ? dialog_id : "";
portENTER_CRITICAL(&s_marker_image_gen_lock);
if (s_marker_image_gen_running) {
acquired = false;
} else if (s_marker_last_success_ms > 0 &&
strcmp(safe_prompt, s_marker_last_success_prompt) == 0 &&
strcmp(safe_dialog, s_marker_last_success_dialog_id) == 0) {
int64_t elapsed_ms = now_ms - s_marker_last_success_ms;
if (elapsed_ms >= 0 && elapsed_ms < VOICE_IMAGE_MARKER_REPEAT_SUPPRESS_MS) {
remain_ms = (uint32_t)(VOICE_IMAGE_MARKER_REPEAT_SUPPRESS_MS - elapsed_ms);
acquired = false;
} else {
s_marker_image_gen_running = true;
acquired = true;
}
} else {
s_marker_image_gen_running = true;
acquired = true;
}
portEXIT_CRITICAL(&s_marker_image_gen_lock);
if (out_remain_ms != NULL) {
*out_remain_ms = remain_ms;
}
return acquired;
}
static void voice_marker_image_generation_release(void) {
portENTER_CRITICAL(&s_marker_image_gen_lock);
s_marker_image_gen_running = false;
portEXIT_CRITICAL(&s_marker_image_gen_lock);
}
static void voice_marker_image_generation_record_success(const char *prompt,
const char *dialog_id,
int64_t now_ms) {
const char *safe_prompt = prompt != NULL ? prompt : "";
const char *safe_dialog = dialog_id != NULL ? dialog_id : "";
portENTER_CRITICAL(&s_marker_image_gen_lock);
strlcpy(s_marker_last_success_prompt, safe_prompt, sizeof(s_marker_last_success_prompt));
strlcpy(s_marker_last_success_dialog_id, safe_dialog, sizeof(s_marker_last_success_dialog_id));
s_marker_last_success_ms = now_ms;
portEXIT_CRITICAL(&s_marker_image_gen_lock);
}
static void *voice_cjson_malloc_hook(size_t size) {
return voice_malloc_prefer_psram(size);
}
@@ -418,6 +486,111 @@ static bool voice_extract_marker_prompt(const char *text, char *out_prompt, size
return prompt_len > 0;
}
static void voice_build_marker_image_prompt(const char *subject,
char *out_prompt,
size_t out_len) {
if (out_prompt == NULL || out_len == 0) {
return;
}
const char *safe_subject = (subject != NULL && subject[0] != '\0') ?
subject :
VOICE_IMAGE_MARKER_TEMPLATE_FALLBACK_SUBJECT;
int written = snprintf(out_prompt,
out_len,
VOICE_IMAGE_MARKER_TEMPLATE,
safe_subject);
if (written < 0) {
out_prompt[0] = '\0';
return;
}
if ((size_t)written >= out_len) {
out_prompt[out_len - 1] = '\0';
}
}
static esp_err_t voice_submit_marker_generated_image(const voice_image_generation_task_arg_t *task_arg,
const image_generation_result_t *result,
char *err,
size_t err_len) {
if (task_arg == NULL || result == NULL || result->png == NULL || result->png_len == 0) {
voice_fill_err(err, err_len, "invalid image result");
return ESP_ERR_INVALID_ARG;
}
uint16_t raster_width = 0;
uint16_t raster_height = 0;
uint8_t *raster = NULL;
size_t raster_len = 0;
char decode_err[128] = {0};
ESP_LOGI(TAG,
"[stage] marker_image_print_decode_start: dialog_id=%s request_id=%s png_bytes=%u",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
result->request_id[0] != '\0' ? result->request_id : "-",
(unsigned)result->png_len);
esp_err_t decode_rc = raster_tools_convert_png_to_raster_384(result->png,
result->png_len,
true,
VOICE_IMAGE_PRINT_THRESHOLD,
false,
VOICE_IMAGE_PRINT_MAX_HEIGHT,
&raster_width,
&raster_height,
&raster,
&raster_len,
decode_err,
sizeof(decode_err));
if (decode_rc != ESP_OK) {
voice_fill_err(err,
err_len,
decode_err[0] != '\0' ? decode_err : "decode generated image failed");
ESP_LOGW(TAG,
"[stage] marker_image_print_decode_failed: dialog_id=%s rc=0x%x msg=%s",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
(unsigned)decode_rc,
err != NULL && err[0] != '\0' ? err : "decode generated image failed");
return decode_rc;
}
ESP_LOGI(TAG,
"[stage] marker_image_print_decode_done: dialog_id=%s raster_size=%ux%u raster_bytes=%u",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
(unsigned)raster_width,
(unsigned)raster_height,
(unsigned)raster_len);
uint32_t job_id = 0;
char submit_err[128] = {0};
esp_err_t submit_rc = printer_protocol_submit_raster_job(raster,
raster_len,
raster_width,
raster_height,
NULL,
&job_id,
submit_err,
sizeof(submit_err));
if (submit_rc != ESP_OK) {
voice_fill_err(err,
err_len,
submit_err[0] != '\0' ? submit_err : "submit print job failed");
ESP_LOGW(TAG,
"[stage] marker_image_print_submit_failed: dialog_id=%s rc=0x%x msg=%s",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
(unsigned)submit_rc,
err != NULL && err[0] != '\0' ? err : "submit print job failed");
} else {
ESP_LOGI(TAG,
"[stage] marker_image_print_submit_ok: dialog_id=%s request_id=%s job_id=%u",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
result->request_id[0] != '\0' ? result->request_id : "-",
(unsigned)job_id);
}
free(raster);
return submit_rc;
}
static void voice_image_generation_task(void *arg) {
voice_image_generation_task_arg_t *task_arg = (voice_image_generation_task_arg_t *)arg;
if (task_arg == NULL) {
@@ -438,12 +611,17 @@ static void voice_image_generation_task(void *arg) {
image_generation_result_t result = {0};
image_generation_result_reset(&result);
bool status_poll_paused = false;
char gen_err[160] = {0};
ESP_LOGI(TAG,
"[stage] marker_image_gen_start: dialog_id=%s prompt_len=%u",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
(unsigned)strlen(task_arg->prompt));
printer_protocol_set_status_poll_paused(true);
status_poll_paused = true;
esp_err_t rc = image_generation_generate_png(&req, &result, gen_err, sizeof(gen_err));
printer_protocol_set_status_poll_paused(false);
status_poll_paused = false;
if (rc == ESP_OK) {
ESP_LOGI(TAG,
"[stage] marker_image_gen_done: dialog_id=%s request_id=%s size=%ux%u png_bytes=%u",
@@ -452,6 +630,23 @@ static void voice_image_generation_task(void *arg) {
(unsigned)result.width,
(unsigned)result.height,
(unsigned)result.png_len);
char submit_err[160] = {0};
esp_err_t submit_rc = voice_submit_marker_generated_image(task_arg,
&result,
submit_err,
sizeof(submit_err));
if (submit_rc == ESP_OK) {
voice_marker_image_generation_record_success(task_arg->prompt,
task_arg->dialog_id,
voice_now_ms());
} else {
ESP_LOGW(TAG,
"[stage] marker_image_print_failed: dialog_id=%s rc=0x%x msg=%s",
task_arg->dialog_id[0] != '\0' ? task_arg->dialog_id : "-",
(unsigned)submit_rc,
submit_err[0] != '\0' ? submit_err : "marker image print failed");
}
} else {
ESP_LOGW(TAG,
"[stage] marker_image_gen_failed: dialog_id=%s rc=0x%x msg=%s",
@@ -460,7 +655,11 @@ static void voice_image_generation_task(void *arg) {
gen_err[0] != '\0' ? gen_err : "image generation failed");
}
if (status_poll_paused) {
printer_protocol_set_status_poll_paused(false);
}
image_generation_result_free(&result);
voice_marker_image_generation_release();
image_generation_schedule_prewarm();
free(task_arg);
vTaskDelete(NULL);
@@ -479,10 +678,26 @@ static void voice_schedule_marker_image_generation(const char *prompt, const cha
return;
}
uint32_t repeat_remain_ms = 0;
if (!voice_marker_image_generation_try_acquire(prompt,
dialog_id,
voice_now_ms(),
&repeat_remain_ms)) {
if (repeat_remain_ms > 0) {
ESP_LOGI(TAG,
"[stage] marker_image_gen_skip: duplicate prompt suppressed, remain_ms=%u",
(unsigned)repeat_remain_ms);
} else {
ESP_LOGI(TAG, "[stage] marker_image_gen_skip: previous marker task running");
}
return;
}
voice_image_generation_task_arg_t *task_arg =
(voice_image_generation_task_arg_t *)calloc(1, sizeof(*task_arg));
if (task_arg == NULL) {
ESP_LOGW(TAG, "[stage] marker_image_gen_skip: no memory for task arg");
voice_marker_image_generation_release();
return;
}
@@ -499,6 +714,7 @@ static void voice_schedule_marker_image_generation(const char *prompt, const cha
NULL);
if (ok != pdPASS) {
ESP_LOGW(TAG, "[stage] marker_image_gen_skip: create task failed");
voice_marker_image_generation_release();
free(task_arg);
return;
}
@@ -519,15 +735,28 @@ static void voice_try_trigger_marker_image_generation(const char *dialog_id, con
return;
}
char prompt[VOICE_IMAGE_MARKER_PROMPT_MAX_LEN + 1] = {0};
if (!voice_extract_marker_prompt(first_sentence, prompt, sizeof(prompt))) {
char marker_subject[VOICE_IMAGE_MARKER_PROMPT_MAX_LEN + 1] = {0};
if (!voice_extract_marker_prompt(first_sentence, marker_subject, sizeof(marker_subject))) {
return;
}
char templated_prompt[VOICE_IMAGE_MARKER_PROMPT_MAX_LEN + 1] = {0};
voice_build_marker_image_prompt(marker_subject,
templated_prompt,
sizeof(templated_prompt));
if (templated_prompt[0] == '\0') {
ESP_LOGW(TAG, "[stage] marker_image_gen_skip: build prompt failed");
return;
}
ESP_LOGI(TAG,
"[stage] marker_image_gen_detected: first_sentence=%s",
first_sentence);
voice_schedule_marker_image_generation(prompt, dialog_id);
ESP_LOGI(TAG,
"[stage] marker_image_gen_subject: keyword=%s",
marker_subject);
voice_log_final_text("marker_image_gen_prompt_templated", templated_prompt);
voice_schedule_marker_image_generation(templated_prompt, dialog_id);
}
static void voice_handle_output_event(cJSON *output) {

View File

@@ -18,3 +18,4 @@ dependencies:
espressif/esp_codec_dev: ^1.5.4
espressif/esp_audio_codec: ^2.4.1
espressif/libpng: ^1.6.55
espressif/esp_lcd_gc9a01: ^2.0.4