""" 测试 TTS 音频流完整性和播放 覆盖所有通信模块：session管理、dialog_result、tts_audio_chunk、turn.complete """ import asyncio import json import sys import time from pathlib import Path import websockets import numpy as np import sounddevice as sd sys.path.insert(0, str(Path(__file__).parent)) from test_config import SERVER_URL, AUTH_TOKEN, DEVICE_ID from test_utils import TestResult async def test_tts_audio_stream(): """TTS 音频流完整测试""" print("=" * 60) print(" 测试 TTS 音频流完整性") print("=" * 60) print() results = [] # ========== 测试 1: 闲聊场景 TTS 音频流 ========== result = await test_scenario( "闲聊场景 TTS", "你好，今天天气怎么样？", "chitchat", check_audio_length=True, min_audio_length_s=1.0, # 至少 1 秒 ) results.append(result) result.print_report() # ========== 测试 2: 飞控指令场景 TTS 音频流 ========== result = await test_scenario( "飞控指令场景 TTS", "起飞然后在前方十米悬停", "flight_intent", check_audio_length=True, min_audio_length_s=1.0, ) results.append(result) result.print_report() # ========== 测试 3: 返航指令场景 TTS 音频流 ========== result = await test_scenario( "返航指令场景 TTS", "返航", "flight_intent", check_audio_length=True, min_audio_length_s=0.5, ) results.append(result) result.print_report() # ========== 测试 4: 降落指令场景 TTS 音频流 ========== result = await test_scenario( "降落指令场景 TTS", "降落", "flight_intent", check_audio_length=True, min_audio_length_s=0.5, ) results.append(result) result.print_report() # ========== 测试 5: 短文本 TTS 音频流 ========== result = await test_scenario( "短文本 TTS", "你好", "chitchat", check_audio_length=True, min_audio_length_s=0.3, ) results.append(result) result.print_report() # ========== 测试 6: 音频流播放验证 ========== result = await test_audio_playback( "音频流播放验证", "这是一段测试语音，用于验证音频流播放功能。" ) results.append(result) result.print_report() # ========== 汇总 ========== print("\n" + "=" * 60) print(" TTS 音频流测试汇总") print("=" * 60) all_passed = True for r in results: status = "✅ 通过" if r.success else "❌ 失败" print(f" {r.test_name:<30} {status}") if not r.success and r.errors: for err in r.errors: print(f" - {err}") print(f"\n 总计: {sum(1 for r in results if r.success)}/{len(results)} 通过") print("=" * 60) return all(r.success for r in results) async def test_scenario( name: str, text: str, expected_routing: str, check_audio_length: bool = True, min_audio_length_s: float = 0.5, ) -> TestResult: """测试一个场景的 TTS 音频流""" result = TestResult(f"TTS 音频流 - {name}") try: async with websockets.connect(SERVER_URL) as ws: # 1. 建立会话 session_start = { "type": "session.start", "proto_version": "1.0", "transport_profile": "text_uplink", "session_id": f"test-tts-{int(time.time())}", "auth_token": AUTH_TOKEN, "client": { "device_id": DEVICE_ID, "locale": "zh-CN", "capabilities": { "playback_sample_rate_hz": 24000, "prefer_tts_codec": "pcm_s16le" }, "protocol": {"dialog_result": "cloud_voice_dialog_v1"}, } } await ws.send(json.dumps(session_start, ensure_ascii=False)) # 接收 session.ready msg = await asyncio.wait_for(ws.recv(), timeout=10) data = json.loads(msg) if data.get("type") != "session.ready": result.add_error("未收到 session.ready") return result.finalize() # 2. 发送 turn.text turn_id = f"turn-{int(time.time())}" turn_text = { "type": "turn.text", "proto_version": "1.0", "transport_profile": "text_uplink", "turn_id": turn_id, "text": text, "is_final": True, "source": "device_stt" } await ws.send(json.dumps(turn_text, ensure_ascii=False)) # 3. 接收响应 dialog_result = None audio_chunks = [] audio_metadata = [] metrics = {} while True: msg = await asyncio.wait_for(ws.recv(), timeout=60) if isinstance(msg, bytes): # 二进制音频数据 audio_chunks.append(msg) else: data = json.loads(msg) msg_type = data.get("type") if msg_type == "dialog_result": dialog_result = data elif msg_type == "tts_audio_chunk": # 记录 TTS 元数据 audio_metadata.append({ "seq": data.get("seq"), "is_final": data.get("is_final"), "codec": data.get("codec"), "sample_rate_hz": data.get("sample_rate_hz"), }) elif msg_type == "turn.complete": metrics = data.get("metrics", {}) break # 4. 验证 dialog_result if dialog_result is None: result.add_error("未收到 dialog_result") return result.finalize() actual_routing = dialog_result.get("routing") if actual_routing != expected_routing: result.add_error( f"路由不匹配: 期望 {expected_routing}, 实际 {actual_routing}" ) return result.finalize() result.add_detail("routing", actual_routing) # 5. 验证音频块序列 if len(audio_chunks) == 0: result.add_error("未收到任何音频数据") return result.finalize() result.add_detail("audio_chunks_count", len(audio_chunks)) result.add_detail("audio_metadata_count", len(audio_metadata)) # 验证音频块序号连续性 for i, meta in enumerate(audio_metadata): if meta["seq"] != i: result.add_error(f"音频块序号不连续: 期望 {i}, 实际 {meta['seq']}") break if meta["codec"] != "pcm_s16le": result.add_error(f"音频编码格式错误: 期望 pcm_s16le, 实际 {meta['codec']}") break if meta["sample_rate_hz"] != 24000: result.add_error(f"音频采样率错误: 期望 24000, 实际 {meta['sample_rate_hz']}") break # 验证最后一个块标记为 final if audio_metadata: if not audio_metadata[-1]["is_final"]: result.add_error("最后一个音频块未标记为 is_final=True") # 6. 拼接音频数据 full_pcm = b"".join(audio_chunks) audio_data = np.frombuffer(full_pcm, dtype=np.int16) audio_length_s = len(audio_data) / 24000.0 result.add_detail("audio_length_s", f"{audio_length_s:.2f}") result.add_detail("audio_samples", len(audio_data)) # 7. 验证音频长度 if check_audio_length and audio_length_s < min_audio_length_s: result.add_error( f"音频长度过短: {audio_length_s:.2f}s < {min_audio_length_s:.2f}s" ) # 8. 验证音频数据有效性 if np.max(np.abs(audio_data)) == 0: result.add_error("音频数据全为零（静音）") # 9. 播放音频验证 print(f"\n 🔊 播放测试音频 ({audio_length_s:.2f}s)...") try: audio_float = audio_data.astype(np.float32) / 32768.0 sd.play(audio_float, samplerate=24000, blocking=True) print(" ✅ 音频播放成功") except Exception as e: result.add_error(f"音频播放失败: {e}") # 10. 验证性能指标 llm_ms = metrics.get("llm_ms", 0) or 0 tts_ms = metrics.get("tts_first_byte_ms", 0) or 0 result.add_detail("llm_latency_ms", llm_ms) result.add_detail("tts_latency_ms", tts_ms) if llm_ms == 0 and tts_ms == 0: result.add_error("性能指标为空") # 11. 结束会话 session_end = { "type": "session.end", "proto_version": "1.0", "session_id": session_start["session_id"] } await ws.send(json.dumps(session_end)) # 所有检查通过 result.mark_success() result.add_detail("test_text", text) result.add_detail("protocol_modules", [ "session.start/ready", "turn.text", "dialog_result", "tts_audio_chunk (text+binary)", "turn.complete", "session.end" ]) except Exception as e: result.add_error(f"测试异常: {e}") import traceback traceback.print_exc() return result.finalize() async def test_audio_playback(name: str, text: str) -> TestResult: """专门测试音频播放功能""" result = TestResult(f"TTS 音频播放 - {name}") try: async with websockets.connect(SERVER_URL) as ws: # 建立会话 session_start = { "type": "session.start", "proto_version": "1.0", "transport_profile": "text_uplink", "session_id": f"test-playback-{int(time.time())}", "auth_token": AUTH_TOKEN, "client": { "device_id": DEVICE_ID, "locale": "zh-CN", "capabilities": { "playback_sample_rate_hz": 24000, "prefer_tts_codec": "pcm_s16le" }, "protocol": {"dialog_result": "cloud_voice_dialog_v1"}, } } await ws.send(json.dumps(session_start, ensure_ascii=False)) await asyncio.wait_for(ws.recv(), timeout=10) # session.ready # 发送文本 turn_id = f"turn-{int(time.time())}" turn_text = { "type": "turn.text", "proto_version": "1.0", "transport_profile": "text_uplink", "turn_id": turn_id, "text": text, "is_final": True, "source": "device_stt" } await ws.send(json.dumps(turn_text, ensure_ascii=False)) # 接收并收集音频 audio_chunks = [] dialog_result = None metrics = {} while True: msg = await asyncio.wait_for(ws.recv(), timeout=60) if isinstance(msg, bytes): audio_chunks.append(msg) else: data = json.loads(msg) msg_type = data.get("type") if msg_type == "dialog_result": dialog_result = data elif msg_type == "turn.complete": metrics = data.get("metrics", {}) break # 播放音频 if audio_chunks: full_pcm = b"".join(audio_chunks) audio_data = np.frombuffer(full_pcm, dtype=np.int16) audio_length_s = len(audio_data) / 24000.0 print(f"\n 🎵 完整音频信息:") print(f" 长度: {audio_length_s:.2f}s") print(f" 采样数: {len(audio_data)}") print(f" 峰值: {np.max(np.abs(audio_data))}") print(f" 均值: {np.mean(np.abs(audio_data)):.1f}") # 播放 print(f"\n 🔊 播放音频...") t_start = time.time() audio_float = audio_data.astype(np.float32) / 32768.0 sd.play(audio_float, samplerate=24000, blocking=True) t_play = time.time() - t_start print(f" ✅ 播放完成 (耗时 {t_play:.2f}s)") # 验证实时率 realtime_factor = audio_length_s / t_play if t_play > 0 else 0 print(f" 📊 实时率: {realtime_factor:.2f}x") if realtime_factor < 1.0: result.add_warning( f"实时率低于 1.0: {realtime_factor:.2f}x " f"(可能导致播放卡顿)" ) result.mark_success() result.add_detail("audio_length_s", f"{audio_length_s:.2f}") result.add_detail("playback_time_s", f"{t_play:.2f}") result.add_detail("realtime_factor", f"{realtime_factor:.2f}") result.add_detail("metrics", metrics) else: result.add_error("未收到音频数据") # 结束会话 await ws.send(json.dumps({ "type": "session.end", "proto_version": "1.0", "session_id": session_start["session_id"] })) except Exception as e: result.add_error(f"测试异常: {e}") import traceback traceback.print_exc() return result.finalize() async def main(): print("\n" + "=" * 60) print(" TTS 音频流完整性测试") print(" 覆盖: session管理 + dialog_result + tts_audio_chunk + turn.complete") print("=" * 60) print() success = await test_tts_audio_stream() print("\n" + "=" * 60) if success: print(" ✅ 所有 TTS 音频流测试通过！") print(" 已验证:") print(" - session.start/ready 会话管理") print(" - turn.text 文本发送") print(" - dialog_result 结构化结果") print(" - tts_audio_chunk 流式音频（text 头 + binary 体）") print(" - turn.complete 完成通知") print(" - 音频数据完整性和长度") print(" - 音频流本地播放") else: print(" ❌ 部分 TTS 音频流测试失败！") print("=" * 60) sys.exit(0 if success else 1) if __name__ == "__main__": asyncio.run(main())