Bairan.Zhang/AI_Printer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(voice): improve on-screen dialog text and wifi provisioning hints

Browse Source
This commit is contained in:
admin
2026-03-02 03:52:10 +08:00
parent 6d82ccae3e
commit 6780bb2789
7 changed files with 283 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
components/domain/include/domain.h
View File

@@ -115,6 +115,10 @@ esp_err_t screen_preview_show_png(const uint8_t *png,
uint32_t timeout_ms,
char *err,
size_t err_len);
esp_err_t screen_preview_show_text(const char *text,
uint32_t timeout_ms,
char *err,
size_t err_len);
// ---------- printer_protocol ----------
typedef enum {

13
components/domain/src/raster_tools.c
View File

@@ -125,6 +125,13 @@ static void set_black(uint8_t *buf, uint16_t width, uint16_t x, uint16_t y) {
buf[idx] |= (uint8_t)(1u << bit);
}
static void set_black_bold(uint8_t *buf, uint16_t width, uint16_t x, uint16_t y) {
set_black(buf, width, x, y);
if ((uint16_t)(x + 1) < width) {
set_black(buf, width, (uint16_t)(x + 1), y);
}
}
static size_t cn16_index_count(void) {
size_t bytes = (size_t)(_binary_cn16_index_bin_end - _binary_cn16_index_bin_start);
return bytes / CN16_INDEX_ENTRY_BYTES;
@@ -211,7 +218,7 @@ static void draw_char(uint8_t *buf,
uint16_t px = (uint16_t)(x + col * scale + sx);
uint16_t py = (uint16_t)(y + row * scale + sy);
if (py < height) {
set_black(buf, width, px, py);
set_black_bold(buf, width, px, py);
}
}
}
@@ -241,7 +248,7 @@ static void draw_cn16_glyph(uint8_t *buf,
uint16_t px = (uint16_t)(x + col * scale + sx);
uint16_t py = (uint16_t)(y + row * scale + sy);
if (py < height && px < width) {
set_black(buf, width, px, py);
set_black_bold(buf, width, px, py);
}
}
}
@@ -265,7 +272,7 @@ static void draw_missing_box(uint8_t *buf,
uint16_t px = (uint16_t)(x + c);
uint16_t py = (uint16_t)(y + r);
if (px < width && py < height) {
set_black(buf, width, px, py);
set_black_bold(buf, width, px, py);
}
}
}

57
components/domain/src/screen_preview.c
View File

@@ -1,6 +1,13 @@
#include "domain.h"
#include "platform.h"
#include <stdlib.h>
#include <string.h>
#define SCREEN_PREVIEW_TEXT_SCALE 2
#define SCREEN_PREVIEW_TEXT_LINE_SPACING 1
#define SCREEN_PREVIEW_TEXT_MAX_HEIGHT 3000
esp_err_t screen_preview_show_raster(const uint8_t *raster,
size_t raster_len,
uint16_t width,
@@ -20,3 +27,53 @@ esp_err_t screen_preview_show_png(const uint8_t *png,
{
return platform_display_show_png(png, png_len, timeout_ms, err, err_len);
}
esp_err_t screen_preview_show_text(const char *text,
uint32_t timeout_ms,
char *err,
size_t err_len)
{
if (text == NULL || text[0] == '\0') {
if (err != NULL && err_len > 0) {
strlcpy(err, "invalid text", err_len);
}
return ESP_ERR_INVALID_ARG;
}
uint16_t width = 0;
uint16_t height = 0;
uint8_t *raster = NULL;
size_t raster_len = 0;
char render_err[96] = {0};
esp_err_t render_rc = raster_tools_render_text_384(text,
SCREEN_PREVIEW_TEXT_SCALE,
SCREEN_PREVIEW_TEXT_LINE_SPACING,
SCREEN_PREVIEW_TEXT_MAX_HEIGHT,
&width,
&height,
&raster,
&raster_len,
render_err,
sizeof(render_err));
if (render_rc != ESP_OK) {
if (err != NULL && err_len > 0 && render_err[0] != '\0') {
strlcpy(err, render_err, err_len);
}
return render_rc;
}
char preview_err[96] = {0};
esp_err_t preview_rc = screen_preview_show_raster(raster,
raster_len,
width,
height,
timeout_ms,
preview_err,
sizeof(preview_err));
free(raster);
if (preview_rc != ESP_OK && err != NULL && err_len > 0 && preview_err[0] != '\0') {
strlcpy(err, preview_err, err_len);
}
return preview_rc;
}

58
components/domain/src/system_runtime.c
View File

@@ -1,12 +1,69 @@
#include "domain.h"
#include "platform.h"
#include <stdio.h>
#include "esp_log.h"
static const char *TAG = "system_runtime";
#define SYSTEM_RUNTIME_WIFI_HINT_TIMEOUT_MS 220
static bool s_wifi_hint_shown;
static void system_runtime_show_wifi_hint(void) {
if (s_wifi_hint_shown) {
return;
}
s_wifi_hint_shown = true;
const char *ap_ssid = CONFIG_TQ_WIFI_PROV_SOFTAP_SSID;
if (ap_ssid == NULL || ap_ssid[0] == '\0') {
ap_ssid = "TalkingQ-Setup";
}
char hint[256] = {0};
snprintf(hint,
sizeof(hint),
"未检测到Wi-Fi配置\n连接热点: %s\n打开: http://192.168.4.1",
ap_ssid);
char preview_err[96] = {0};
esp_err_t preview_rc = screen_preview_show_text(hint,
SYSTEM_RUNTIME_WIFI_HINT_TIMEOUT_MS,
preview_err,
sizeof(preview_err));
if (preview_rc != ESP_OK && preview_rc != ESP_ERR_NOT_SUPPORTED) {
ESP_LOGW(TAG,
"wifi_hint_preview_failed: rc=0x%x msg=%s",
(unsigned)preview_rc,
preview_err[0] != '\0' ? preview_err : "show text failed");
}
}
static void system_runtime_try_show_wifi_hint_for_unprovisioned(void) {
if (!wifi_manager_has_saved_credentials()) {
system_runtime_show_wifi_hint();
}
}
static void system_runtime_wifi_event_handler(wifi_manager_event_t event, void *user_data) {
(void)user_data;
if (event == WIFI_MANAGER_EVENT_PROVISIONING_STARTED) {
system_runtime_show_wifi_hint();
}
}
esp_err_t system_runtime_start(void) {
esp_err_t err = platform_bootstrap_init();
if (err != ESP_OK) {
return err;
}
s_wifi_hint_shown = false;
(void)wifi_manager_set_event_callback(system_runtime_wifi_event_handler, NULL);
system_runtime_try_show_wifi_hint_for_unprovisioned();
err = wifi_manager_start();
if (err != ESP_OK) {
// Reset platform state so lifecycle retry runs from a clean baseline.
@@ -18,6 +75,7 @@ esp_err_t system_runtime_start(void) {
}
esp_err_t system_runtime_shutdown(void) {
(void)wifi_manager_set_event_callback(NULL, NULL);
return wifi_manager_stop();
}

125
components/domain/src/voice_interaction_ws_protocol.c
View File

@@ -1,5 +1,7 @@
#include "voice_interaction_internal.h"
#include <ctype.h>
#include <stdio.h>
#include <string.h>
#define cJSON_malloc voice_malloc_prefer_psram
@@ -14,10 +16,99 @@ 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
#define VOICE_SCREEN_PREVIEW_TIMEOUT_MS 180
static portMUX_TYPE s_response_finalize_lock = portMUX_INITIALIZER_UNLOCKED;
static bool s_response_finalize_running;
static bool voice_is_space_char(char c) {
return isspace((unsigned char)c) != 0;
}
static size_t voice_find_first_sentence_end(const char *text, size_t len) {
if (text == NULL || len == 0) {
return 0;
}
for (size_t i = 0; i < len; ++i) {
unsigned char c = (unsigned char)text[i];
if (c == '.' || c == '!' || c == '?' || c == ':' || c == ';' ||
c == '\n' || c == '\r') {
return i + 1;
}
if (i + 2 < len) {
unsigned char b1 = (unsigned char)text[i];
unsigned char b2 = (unsigned char)text[i + 1];
unsigned char b3 = (unsigned char)text[i + 2];
// UTF-8: 。 (E3 80 82)
if (b1 == 0xE3 && b2 == 0x80 && b3 == 0x82) {
return i + 3;
}
// UTF-8: ！(EF BC 81) ？(EF BC 9F) ；(EF BC 9B) ：(EF BC 9A)
if (b1 == 0xEF && b2 == 0xBC &&
(b3 == 0x81 || b3 == 0x9F || b3 == 0x9B || b3 == 0x9A)) {
return i + 3;
}
}
}
return len;
}
static const char *voice_get_text_without_first_sentence(const char *text) {
if (text == NULL || text[0] == '\0') {
return NULL;
}
size_t len = strlen(text);
size_t begin = 0;
while (begin < len && voice_is_space_char(text[begin])) {
begin++;
}
if (begin >= len) {
return NULL;
}
size_t first_end = begin + voice_find_first_sentence_end(text + begin, len - begin);
while (first_end < len && voice_is_space_char(text[first_end])) {
first_end++;
}
if (first_end < len) {
return text + first_end;
}
// Fallback: keep answer body after marker prefix when punctuation split
// does not leave a second sentence.
const char *marker_begin = strstr(text + begin, "/*");
if (marker_begin == NULL) {
return NULL;
}
const char *marker_end = strstr(marker_begin + 2, "*/");
if (marker_end == NULL) {
return NULL;
}
const char *cursor = marker_end + 2;
while (*cursor != '\0') {
unsigned char ch = (unsigned char)*cursor;
if (!voice_is_space_char((char)ch) &&
ch != '.' && ch != '!' && ch != '?' &&
ch != ':' && ch != ';' && ch != ',') {
break;
}
cursor++;
}
if (*cursor == '\0') {
return NULL;
}
return cursor;
}
static esp_err_t voice_send_directive_with_retry(const char *action,
const char *directive,
bool with_dialog_id) {
@@ -136,6 +227,24 @@ static bool voice_get_json_bool(cJSON *obj, const char *name, bool *out_value) {
return true;
}
static void voice_try_preview_final_text(const char *text) {
if (text == NULL || text[0] == '\0') {
return;
}
char preview_err[96] = {0};
esp_err_t preview_rc = screen_preview_show_text(text,
VOICE_SCREEN_PREVIEW_TIMEOUT_MS,
preview_err,
sizeof(preview_err));
if (preview_rc != ESP_OK && preview_rc != ESP_ERR_NOT_SUPPORTED) {
ESP_LOGW(TAG,
"dialog_text_preview_failed: rc=0x%x msg=%s",
(unsigned)preview_rc,
preview_err[0] != '\0' ? preview_err : "show text failed");
}
}
static void voice_handle_output_event(cJSON *output) {
const char *event_name = voice_get_json_str(output, "event");
if (event_name == NULL) {
@@ -150,11 +259,14 @@ static void voice_handle_output_event(cJSON *output) {
bool finished = false;
bool has_finished = voice_get_json_bool(output, "finished", &finished);
const char *speech_final_text = NULL;
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, "RespondingContent") == 0) {
if (strcmp(event_name, "SpeechContent") == 0) {
speech_final_text = voice_get_json_str(output, "text");
} else if (strcmp(event_name, "RespondingContent") == 0) {
const char *final_text = voice_get_json_str(output, "text");
if (final_text == NULL || final_text[0] == '\0') {
final_text = voice_get_json_str(output, "spoken");
@@ -233,6 +345,17 @@ static void voice_handle_output_event(cJSON *output) {
voice_close_audio_with_drain(VOICE_AUDIO_DRAIN_WAIT_MS, "server_event_close");
}
if (speech_final_text != NULL && speech_final_text[0] != '\0') {
voice_try_preview_final_text(speech_final_text);
}
if (responding_final_text != NULL && responding_final_text[0] != '\0') {
const char *preview_text = voice_get_text_without_first_sentence(responding_final_text);
if (preview_text != NULL && preview_text[0] != '\0') {
voice_try_preview_final_text(preview_text);
}
}
if (should_check_marker_in_first_sentence) {
voice_marker_try_trigger_image_generation(dialog_id, responding_final_text);
}