增加了单击启动对话和长按期间对话，默认使用短按启动对话

components/control_plane/src/runtime/control_plane_key_input.c

@@ -52,6 +52,10 @@
 #define CONFIG_TQ_KEY_INPUT_POLL_MS 20
 #endif
 
+#if !defined(CONFIG_TQ_PTT_MODE_HOLD_TO_TALK) && !defined(CONFIG_TQ_PTT_MODE_CLICK_TO_TALK)
+#define CONFIG_TQ_PTT_MODE_CLICK_TO_TALK 1
+#endif
+
 #ifndef CONTROL_PLANE_PTT_INITIAL_INACTIVE_STABLE_MS
 #define CONTROL_PLANE_PTT_INITIAL_INACTIVE_STABLE_MS 120
 #endif
@@ -123,9 +127,15 @@ static const control_plane_key_config_t s_key_configs[] = {
         .long_press_ms = CONFIG_TQ_KEY_LONG_PRESS_MS,
         .double_click_ms = CONFIG_TQ_KEY_DOUBLE_CLICK_MS,
         .emit_press_release = true,
+#if CONFIG_TQ_PTT_MODE_CLICK_TO_TALK
+        .enable_short_press = true,
+        .enable_long_press = false,
+        .enable_double_click = true,
+#else
         .enable_short_press = false,
         .enable_long_press = false,
         .enable_double_click = false,
+#endif
     },
     {
         .key_id = CONTROL_PLANE_KEY_ID_SW2,

components/control_plane/src/runtime/control_plane_runtime_workers.c

@@ -80,6 +80,14 @@
 #define CONFIG_TQ_TASK_MONITOR_MAX_TASKS 64
 #endif
 
+#if !defined(CONFIG_TQ_PTT_MODE_HOLD_TO_TALK) && !defined(CONFIG_TQ_PTT_MODE_CLICK_TO_TALK)
+#define CONFIG_TQ_PTT_MODE_CLICK_TO_TALK 1
+#endif
+
+#ifndef CONFIG_TQ_PTT_CLICK_CAPTURE_TIMEOUT_MS
+#define CONFIG_TQ_PTT_CLICK_CAPTURE_TIMEOUT_MS 8000
+#endif
+
 #ifndef CONFIG_TQ_NETWORK_TASK_CORE
 #define CONFIG_TQ_NETWORK_TASK_CORE 0
 #endif
@@ -1189,6 +1197,18 @@ static void control_plane_drive_ptt_state_machine(TickType_t *last_ws_start_atte
     }
 
     TickType_t now = xTaskGetTickCount();
+#if CONFIG_TQ_PTT_MODE_CLICK_TO_TALK
+    TickType_t capture_timeout_ticks = pdMS_TO_TICKS(CONFIG_TQ_PTT_CLICK_CAPTURE_TIMEOUT_MS);
+    if (capture_timeout_ticks == 0) {
+        capture_timeout_ticks = 1;
+    }
+    if (ptt.press_tick != 0 &&
+        (now - ptt.press_tick) >= capture_timeout_ticks) {
+        control_plane_set_ptt_active(false);
+        return;
+    }
+#endif
+
     TickType_t denied_ticks = pdMS_TO_TICKS(CONTROL_PLANE_PTT_DENIED_TIMEOUT_MS);
     if (denied_ticks == 0) {
         denied_ticks = 1;

components/control_plane/src/runtime/control_plane_system_events.c

@@ -30,6 +30,10 @@
 #define CONFIG_TQ_SYSTEM_EVENT_TASK_STACK 4096
 #endif
 
+#if !defined(CONFIG_TQ_PTT_MODE_HOLD_TO_TALK) && !defined(CONFIG_TQ_PTT_MODE_CLICK_TO_TALK)
+#define CONFIG_TQ_PTT_MODE_CLICK_TO_TALK 1
+#endif
+
 #define CONTROL_PLANE_PTT_HOLD_STEP_MS 80
 
 #ifndef CONFIG_TQ_LOCAL_TASK_CORE
@@ -153,12 +157,32 @@ static void control_plane_dispatch_key_event(const control_plane_key_event_t *ev
 
     switch (event->type) {
     case CONTROL_PLANE_KEY_EVENT_PRESSED: {
+#if CONFIG_TQ_PTT_MODE_CLICK_TO_TALK
+        control_plane_multimodal_runtime_state_t state = {0};
+        if (control_plane_multimodal_get_runtime_state(&state) == ESP_OK &&
+            state.remote_responding) {
+            esp_err_t err = control_plane_multimodal_send_request_to_speak();
+            if (err != ESP_OK && err != ESP_ERR_INVALID_STATE) {
+                ESP_LOGW(TAG, "PTT barge-in request failed: %s", esp_err_to_name(err));
+            }
+        }
+#else
         control_plane_set_ptt_active(true);
+#endif
         break;
     }
 
     case CONTROL_PLANE_KEY_EVENT_RELEASED: {
+#if !CONFIG_TQ_PTT_MODE_CLICK_TO_TALK
         control_plane_set_ptt_active(false);
+#endif
+        break;
+    }
+
+    case CONTROL_PLANE_KEY_EVENT_SHORT_PRESS: {
+#if CONFIG_TQ_PTT_MODE_CLICK_TO_TALK
+        control_plane_set_ptt_active(true);
+#endif
         break;
     }
 

main/Kconfig.projbuild

@@ -278,6 +278,36 @@ menu "Example Configuration"
             help
                 Ignore raw GPIO chatter briefly after a double-click has been emitted.
 
+        choice TQ_PTT_INTERACTION_MODE
+            prompt "PTT interaction mode"
+            default TQ_PTT_MODE_CLICK_TO_TALK
+            depends on TQ_KEY_INPUT_ENABLE
+            help
+                Select how the PTT key controls multimodal speech input.
+
+            config TQ_PTT_MODE_CLICK_TO_TALK
+                bool "Click-to-talk"
+                help
+                    A short press opens one bounded speech-input window. Double-click
+                    remains available for the PTT double-click expression and does not
+                    start speech input.
+
+            config TQ_PTT_MODE_HOLD_TO_TALK
+                bool "Hold-to-talk"
+                help
+                    Preserve the original behavior: press and hold starts speech input,
+                    release stops speech input.
+        endchoice
+
+        config TQ_PTT_CLICK_CAPTURE_TIMEOUT_MS
+            int "PTT click-to-talk capture timeout (ms)"
+            default 8000
+            range 1000 60000
+            depends on TQ_KEY_INPUT_ENABLE && TQ_PTT_MODE_CLICK_TO_TALK
+            help
+                Maximum microphone capture window started by one PTT short press.
+                When it expires, uplink capture is stopped and StopSpeech is sent.
+
     endmenu
 
     menu "TQ Screen"

sdkconfig.defaults

@@ -56,6 +56,9 @@ CONFIG_TQ_KEY_DEBOUNCE_INTERVAL_MS=20
 CONFIG_TQ_KEY_LONG_PRESS_MS=800
 CONFIG_TQ_KEY_DOUBLE_CLICK_MS=450
 CONFIG_TQ_KEY_POST_DOUBLE_CLICK_SUPPRESS_MS=450
+CONFIG_TQ_PTT_MODE_CLICK_TO_TALK=y
+# CONFIG_TQ_PTT_MODE_HOLD_TO_TALK is not set
+CONFIG_TQ_PTT_CLICK_CAPTURE_TIMEOUT_MS=8000
 
 # Internal DMA-capable SRAM budget. TLS/auth/fortune flows, hardware AES DMA
 # bounce buffers, LCD DMA, and worker task stacks all need internal-only memory.

tools/png_to_rle.py

@@ -1,73 +1,9 @@
 #!/usr/bin/env python3
-"""Convert PNG images to the binary RGB565 RLE format used by lcd_direct."""
+"""Convert PNG images to standard TGA RLE image files."""
 
 import argparse
-import struct
 from pathlib import Path
 
-try:
-    from PIL import Image
-except ImportError as exc:
-    raise SystemExit("Pillow is required. Install it with: python -m pip install Pillow") from exc
-
-
-MAX_RUN_COUNT = 0xFFFF
-
-
-def parse_rgb(value):
-    text = value.strip()
-    if text.startswith("#"):
-        text = text[1:]
-    if len(text) != 6:
-        raise argparse.ArgumentTypeError("expected RRGGBB or #RRGGBB")
-    try:
-        return tuple(int(text[index:index + 2], 16) for index in (0, 2, 4))
-    except ValueError as exc:
-        raise argparse.ArgumentTypeError("background must be hexadecimal RGB") from exc
-
-
-def rgb888_to_rgb565(red, green, blue):
-    return ((red & 0xF8) << 8) | ((green & 0xFC) << 3) | (blue >> 3)
-
-
-def image_to_runs(path, background):
-    with Image.open(path) as image:
-        rgba = image.convert("RGBA")
-        width, height = rgba.size
-        bg_red, bg_green, bg_blue = background
-
-        runs = []
-        current_color = None
-        current_count = 0
-
-        for red, green, blue, alpha in rgba.getdata():
-            if alpha < 255:
-                red = ((red * alpha) + (bg_red * (255 - alpha))) // 255
-                green = ((green * alpha) + (bg_green * (255 - alpha))) // 255
-                blue = ((blue * alpha) + (bg_blue * (255 - alpha))) // 255
-
-            color = rgb888_to_rgb565(red, green, blue)
-            if color == current_color and current_count < MAX_RUN_COUNT:
-                current_count += 1
-                continue
-
-            if current_color is not None:
-                runs.append((current_color, current_count))
-            current_color = color
-            current_count = 1
-
-        if current_color is not None:
-            runs.append((current_color, current_count))
-
-    return width, height, runs
-
-
-def write_rle(path, runs):
-    path.parent.mkdir(parents=True, exist_ok=True)
-    with path.open("wb") as output:
-        for color, count in runs:
-            output.write(struct.pack("<HH", color, count))
-
 
 def iter_inputs(input_path, recursive):
     if input_path.is_file():
@@ -91,16 +27,26 @@ def output_path_for(input_file, input_root, output):
     return (output / relative).with_suffix(".rle")
 
 
-def convert_file(input_file, input_root, output, background):
+def convert_file(input_file, input_root, output):
+    try:
+        from PIL import Image
+    except ImportError as exc:
+        raise SystemExit(
+            "Pillow is required. Install it with: python -m pip install Pillow"
+        ) from exc
+
     out_path = output_path_for(input_file, input_root, output)
-    width, height, runs = image_to_runs(input_file, background)
+    out_path.parent.mkdir(parents=True, exist_ok=True)
-    write_rle(out_path, runs)
+    with Image.open(input_file) as image:
-    print(f"{input_file} -> {out_path} ({width}x{height}, {len(runs)} runs)")
+        converted = image.convert("RGBA")
+        width, height = converted.size
+        converted.save(out_path, format="TGA", compression="tga_rle")
+    print(f"{input_file} -> {out_path} ({width}x{height})")
 
 
 def main():
     parser = argparse.ArgumentParser(
-        description="Convert PNG files to lcd_direct RGB565 RLE binary assets.",
+        description="Convert PNG files to standard TGA RLE files using Pillow.",
     )
     parser.add_argument("input", type=Path, help="PNG file or directory containing PNG files.")
     parser.add_argument(
@@ -115,12 +61,6 @@ def main():
         action="store_true",
         help="Recursively convert PNG files when input is a directory.",
     )
-    parser.add_argument(
-        "--background",
-        type=parse_rgb,
-        default=(0, 0, 0),
-        help="RGB background for transparent pixels, as RRGGBB or #RRGGBB. Default: 000000.",
-    )
     args = parser.parse_args()
 
     input_path = args.input.resolve()
@@ -136,7 +76,7 @@ def main():
         raise SystemExit("single-file output must use a .rle suffix")
 
     for input_file in inputs:
-        convert_file(input_file, input_path, output, args.background)
+        convert_file(input_file, input_path, output)
 
 
 if __name__ == "__main__":