arona-agent 1.0.0

package/python/tts_say.py

@@ -0,0 +1,177 @@
+#!/usr/bin/env python3
+"""
+ARONA 非流式 TTS：一句话合成并播放（一次性进程，非流式 HTTP）。
+
+与 Node 侧（src/tts_stream.ts）通过 stdin/stdout JSON 行通信（每句一个进程）：
+  stdin  （单行）: {"text": "...", "voice": "..."}
+  stdout 事件（协议通道，禁止混入日志）：
+    {"event":"ready"}        进程就绪
+    {"event":"play_start"}   音频就绪、开始播放
+    {"event":"play_end"}     播放完毕
+    {"event":"error","message":"..."}
+
+流程：HTTP POST 非流式 TTS 端点 → 解析 output.audio.url → 下载音频 → 播放。
+整句一次合成，语气上下文完整，避免流式逐段截断（如"好好吃饭哦"被拆读）。
+播放：pyaudio 播 wav；pyaudio 不可用降级系统播放器（afplay / powershell / aplay / ffplay / paplay）。
+
+环境变量：
+  QWEN_TTS_API_KEY      - 百炼 API Key
+  QWEN_TTS_MODEL        - 模型名（默认 qwen-audio-3.0-tts-plus）
+  QWEN_WORKSPACE_ID     - 百炼业务空间 ID（可选；留空走全局域名 dashscope.aliyuncs.com）
+"""
+
+import io
+import json
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+import urllib.request
+
+from _i18n import t
+
+SAMPLE_RATE = 24000  # qwen-audio-3.0-tts 非流式 wav 标准采样率
+
+
+def emit(event, **kwargs):
+    """向 stdout 输出一行事件 JSON（协议通道，禁止日志混入）。"""
+    payload = {"event": event, **kwargs}
+    sys.stdout.write(json.dumps(payload, ensure_ascii=False) + "\n")
+    sys.stdout.flush()
+
+
+def _opener():
+    # 显式禁用代理：DashScope / OSS 国内服务直连，避免 Clash 系统代理干扰
+    return urllib.request.build_opener(urllib.request.ProxyHandler({}))
+
+
+def build_endpoint():
+    ws = os.environ.get("QWEN_WORKSPACE_ID", "")
+    if ws:
+        return f"https://{ws}.cn-beijing.maas.aliyuncs.com/api/v1/services/audio/tts/SpeechSynthesizer"
+    return "https://dashscope.aliyuncs.com/api/v1/services/audio/tts/SpeechSynthesizer"
+
+
+def synthesize(text, voice):
+    """非流式合成：返回完整 wav 音频字节。"""
+    api_key = os.environ.get("QWEN_TTS_API_KEY", "")
+    model = os.environ.get("QWEN_TTS_MODEL", "qwen-audio-3.0-tts-plus")
+    if not api_key:
+        raise RuntimeError(t("TTS: 未设置 QWEN_TTS_API_KEY", "TTS: QWEN_TTS_API_KEY not set"))
+
+    body = {
+        "model": model,
+        "input": {
+            "text": text,
+            "voice": voice,
+            "format": "wav",
+            "sample_rate": SAMPLE_RATE,
+        },
+    }
+    req = urllib.request.Request(
+        build_endpoint(),
+        data=json.dumps(body).encode("utf-8"),
+        headers={
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        },
+    )
+    with _opener().open(req, timeout=60) as resp:
+        raw = resp.read()
+    data = json.loads(raw)
+    url = (data.get("output", {}) or {}).get("audio", {}) or {}
+    url = url.get("url")
+    if not url:
+        raise RuntimeError(t(f"TTS 未返回音频 URL：{raw[:200]}", f"TTS no audio url: {raw[:200]}"))
+    with _opener().open(url, timeout=60) as resp:
+        return resp.read()
+
+
+def _play_command(wav_path):
+    """按平台选择降级播放命令（pyaudio 不可用时的非实时兜底）。"""
+    if sys.platform == "darwin":
+        return ["afplay", wav_path]
+    if sys.platform == "win32":
+        return ["powershell", "-NoProfile", "-NonInteractive",
+                "-Command", f"(New-Object Media.SoundPlayer '{wav_path}').PlaySync()"]
+    for cmd in ("aplay", "ffplay", "paplay"):
+        if shutil.which(cmd):
+            return [cmd, wav_path]
+    return ["aplay", wav_path]
+
+
+def _play_pyaudio(data):
+    import pyaudio
+    import wave
+    wf = wave.open(io.BytesIO(data), "rb")
+    p = pyaudio.PyAudio()
+    stream = p.open(
+        format=p.get_format_from_width(wf.getsampwidth()),
+        channels=wf.getnchannels(),
+        rate=wf.getframerate(),
+        output=True,
+    )
+    try:
+        chunk = 1024
+        frame = wf.readframes(chunk)
+        while frame:
+            stream.write(frame)
+            frame = wf.readframes(chunk)
+    finally:
+        stream.stop_stream()
+        stream.close()
+        p.terminate()
+        wf.close()
+
+
+def _play_fallback(data):
+    tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
+    try:
+        tmp.write(data)
+        tmp.close()
+        subprocess.run(_play_command(tmp.name), check=False)
+    finally:
+        try:
+            os.unlink(tmp.name)
+        except Exception:
+            pass
+
+
+def play_wav(data):
+    try:
+        _play_pyaudio(data)
+    except Exception:
+        _play_fallback(data)
+
+
+def main():
+    emit("ready")
+    line = sys.stdin.readline()
+    if not line:
+        return
+    try:
+        cmd = json.loads(line)
+    except Exception:
+        emit("error", message=t("TTS: 无效的 stdin JSON", "TTS: invalid stdin JSON"))
+        sys.stdout.flush()
+        return
+    text = (cmd.get("text") or "").strip()
+    voice = cmd.get("voice") or ""
+    if not text:
+        emit("error", message=t("TTS: 空文本", "TTS: empty text"))
+        sys.stdout.flush()
+        return
+    try:
+        audio = synthesize(text, voice)
+        emit("play_start")
+        play_wav(audio)
+        emit("play_end")
+        sys.stdout.flush()
+    except Exception as e:
+        emit("error", message=str(e))
+        sys.stdout.flush()
+
+
+if __name__ == "__main__":
+    main()

package/python/voice_clone.py

@@ -0,0 +1,98 @@
+#!/usr/bin/env python3
+"""
+ARONA Voice Cloning — uploads voice.mp3 to DashScope, creates a custom voice,
+and prints the voice_id as JSON on stdout. Deletes the uploaded file afterward.
+
+Environment variables:
+  QWEN_TTS_API_KEY     - DashScope API key
+  QWEN_TTS_MODEL       - target TTS model (default qwen-audio-3.0-tts-plus)
+  ARONA_VOICE_AUDIO    - path to the audio file (voice.mp3)
+  ARONA_VOICE_PREFIX   - voice ID prefix, lowercase alnum, <=10 chars (default "arona")
+
+Output (stdout): {"voice_id": "..."} on success, {"error": "..."} on failure.
+All logging goes to stderr so stdout stays clean JSON.
+"""
+
+import json
+import os
+import re
+import sys
+import time
+
+
+def fail(msg):
+    print(json.dumps({"error": msg}))
+    # 同步写 stderr：Node 侧 runPython 在非零退出码时只保留 stderr，stdout 会被丢弃，
+    # 不写 stderr 的话克隆失败的真实原因（如音频质量/网络）到不了调用方。
+    print(f"voice_clone error: {msg}", file=sys.stderr)
+    sys.exit(1)
+
+
+def main():
+    api_key = os.environ.get("QWEN_TTS_API_KEY", "")
+    audio_file = os.environ.get("ARONA_VOICE_AUDIO", "")
+    prefix = os.environ.get("ARONA_VOICE_PREFIX", "arona")
+    target_model = os.environ.get("QWEN_TTS_MODEL", "qwen-audio-3.0-tts-plus")
+
+    if not api_key:
+        fail("QWEN_TTS_API_KEY not set")
+    if not audio_file or not os.path.isfile(audio_file):
+        fail(f"Audio file not found: {audio_file}")
+    if not re.fullmatch(r"[a-z0-9]{1,10}", prefix):
+        fail(f"PREFIX must be lowercase alnum, 1-10 chars (got: {prefix})")
+
+    try:
+        import dashscope
+        from dashscope import Files
+        from dashscope.audio.tts_v2 import VoiceEnrollmentService
+    except ImportError:
+        fail("dashscope package not installed. Run: pip install dashscope")
+
+    dashscope.api_key = api_key
+
+    # 1. Upload audio to DashScope hosted OSS
+    try:
+        resp = Files.upload(file_path=audio_file, purpose="voice_clone")
+        file_id = resp.output["uploaded_files"][0]["file_id"]
+    except Exception as e:
+        fail(f"Upload failed: {e}")
+
+    # 2. Get the Alibaba Cloud internal URL (must use internal address)
+    try:
+        oss_url = Files.get(file_id).output["url"]
+    except Exception as e:
+        fail(f"Failed to get OSS URL: {e}")
+
+    # 3. Submit voice cloning
+    try:
+        svc = VoiceEnrollmentService()
+        voice_id = svc.create_voice(target_model=target_model, prefix=prefix, url=oss_url)
+    except Exception as e:
+        fail(f"create_voice failed: {e}")
+
+    # 4. Wait for voice to be ready (max 5 minutes)
+    try:
+        for _ in range(30):
+            info = svc.query_voice(voice_id=voice_id)
+            status = info.get("status")
+            if status == "OK":
+                break
+            if status == "UNDEPLOYED":
+                fail("Voice creation failed. Please check audio quality (10-20s clear voice recommended).")
+            time.sleep(10)
+        else:
+            fail("Timeout waiting for voice to be ready (5 minutes)")
+    except Exception as e:
+        fail(f"query_voice failed: {e}")
+
+    # 5. Delete the uploaded file (best-effort cleanup)
+    try:
+        Files.delete(file_id)
+    except Exception:
+        pass  # Cleanup is best-effort
+
+    print(json.dumps({"voice_id": voice_id}))
+
+
+if __name__ == "__main__":
+    main()

package/requirements.txt

@@ -0,0 +1,6 @@
+cua>=0.1.1,<0.2.0
+websockets>=12.0
+pyaudio>=0.2.14
+numpy>=1.26.0
+pynput>=1.8.0
+dashscope>=1.20.0