| """Real WebSocket / ASR / diarization / TTS check with synthetic fixtures only.""" |
| import argparse,asyncio,json,struct,subprocess,time |
| from pathlib import Path |
| import aiohttp |
| import numpy as np |
| from speaker_v4.common import atomic |
| from .fixtures import ROOT |
|
|
| def audio(name): |
| p=subprocess.run(['ffmpeg','-nostdin','-v','error','-i',str(ROOT/(name+'.mp3')),'-ar','16000','-ac','1','-f','f32le','pipe:1'],capture_output=True,check=True) |
| x=np.frombuffer(p.stdout,'<f4').copy();rms=np.sqrt(np.mean(x*x));return x*(.055/max(float(rms),1e-6)) |
|
|
| async def main(url,mode='v4-long'): |
| intro=audio('christoph-intro');bg=audio('background');offset=1.28 |
| duration=offset+len(intro)/16000+5.12;n=round(duration*16000) |
| background=np.resize(bg,n)*.20;wave=background.copy();start=round(offset*16000);wave[start:start+len(intro)]+=intro |
| events=[];chunks=0;counts={};started=time.monotonic();first_endpoint=None;new_users=[];errors=[] |
| async with aiohttp.ClientSession() as client: |
| for _ in range(90): |
| try: |
| async with client.get(url.rstrip('/')+'/api/state') as status: |
| if (await status.json()).get('ready'):break |
| except aiohttp.ClientError:pass |
| await asyncio.sleep(.5) |
| async with client.ws_connect(url.rstrip('/')+'/ws') as ws: |
| async def receive(): |
| nonlocal chunks,first_endpoint |
| async for message in ws: |
| if message.type==aiohttp.WSMsgType.BINARY: |
| rid=struct.unpack_from('<i',message.data)[0] |
| if rid: |
| chunks+=1;counts[rid]=counts.get(rid,0)+(len(message.data)-4)//2 |
| await ws.send_str(f'played:{rid}:{counts[rid]}') |
| elif message.type==aiohttp.WSMsgType.TEXT: |
| item=json.loads(message.data);kind=item.get('type') |
| if kind in ('warning','error'):errors.append(item) |
| if kind=='identity' and item.get('decision',{}).get('new_user'):new_users.append(item['decision']) |
| if kind in ('endpoint','reply'):events.append({'at_s':time.monotonic()-started,**item}) |
| if kind=='endpoint' and first_endpoint is None:first_endpoint=time.monotonic()-started |
| task=asyncio.create_task(receive()) |
| try: |
| await ws.send_str('identity:mode:'+mode);await asyncio.sleep(.3) |
| started=time.monotonic() |
| for at in range(0,len(wave),320): |
| x=wave[at:at+320];pcm=np.clip(np.rint(x*32767),-32768,32767).astype('<i2') |
| if ws.closed:break |
| try:await ws.send_bytes(struct.pack('<ii',0,-1)+pcm.tobytes()) |
| except aiohttp.ClientConnectionError:break |
| target=started+(at+len(x))/16000;await asyncio.sleep(max(0,target-time.monotonic())) |
| for _ in range(300): |
| if chunks:break |
| await asyncio.sleep(.1) |
| report={'synthetic_fixtures_only':True,'mode':mode,'background_continues_after_target':True, |
| 'target_end_s':offset+len(intro)/16000,'first_endpoint_s':first_endpoint, |
| 'submit_after_target_end_s':None if first_endpoint is None else first_endpoint-(offset+len(intro)/16000), |
| 'new_users':new_users,'events':events,'reply_audio_chunks':chunks,'warnings':errors, |
| 'playback':'virtual immediate acknowledgments; not a physical microphone/DAC test'} |
| folder=Path('/mnt/nvme/robo-speaker-studies/speaker-live-results');folder.mkdir(exist_ok=True) |
| atomic(folder/(mode+'-live.json'),report);print(json.dumps(report,indent=2),flush=True) |
| assert first_endpoint is not None and chunks>0,'No spoken reply while background continued' |
| assert any(e.get('speaker_name')=='Christoph' for e in events if e.get('type')=='endpoint'),'Speaker name not bound after full introduction' |
| finally: |
| await ws.close();task.cancel();await asyncio.gather(task,return_exceptions=True) |
|
|
| if __name__=='__main__': |
| parser=argparse.ArgumentParser();parser.add_argument('--url',default='http://127.0.0.1:8793');parser.add_argument('--mode',default='v4-long');a=parser.parse_args() |
| asyncio.run(main(a.url,a.mode)) |
|
|
