| """Synthetic-only ASR -> full-utterance enrollment -> reply integration check. |
| |
| Pass a locally generated public synthetic voice saying Hey Robo in a sentence. |
| It uses the actual running EARS/TTS/Luna services, not a physical microphone. |
| No transcript, audio or embeddings are written by this test. |
| """ |
| import argparse |
| import asyncio |
| import json |
| from pathlib import Path |
| import struct |
| import subprocess |
| import time |
| import aiohttp |
| import numpy as np |
|
|
|
|
| async def run(args): |
| result=subprocess.run(['ffmpeg','-nostdin','-v','error','-i',args.audio,'-ar','16000','-ac','1','-f','f32le','pipe:1'],capture_output=True,check=True) |
| audio=np.frombuffer(result.stdout,'<f4') |
| audio=np.r_[np.zeros(10240,np.float32),audio,np.zeros(64000,np.float32)] |
| decisions=[];events={};enabled=asyncio.Event();enrolled=asyncio.Event();chunks=0;backchannels=0;counts={} |
| async with aiohttp.ClientSession() as client: |
| async with client.ws_connect(args.url.rstrip('/')+'/ws') as ws: |
| async def receive(): |
| nonlocal chunks,backchannels |
| async for message in ws: |
| if message.type==aiohttp.WSMsgType.BINARY: |
| rid=struct.unpack_from('<i',message.data)[0] |
| if rid:chunks+=1 |
| else:backchannels+=1;continue |
| 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');events[kind]=events.get(kind,0)+1 |
| if kind=='identity': |
| if item.get('enabled'):enabled.set() |
| if 'decision' in item: |
| decisions.append(item['decision']) |
| if item['decision'].get('new_user'):enrolled.set() |
| task=asyncio.create_task(receive()) |
| try: |
| await ws.send_str('identity:on');await asyncio.wait_for(enabled.wait(),10) |
| for i in range(0,len(audio),320): |
| x=audio[i:i+320] |
| pcm=np.clip(np.round(x*32767),-32768,32767).astype('<i2') |
| await ws.send_bytes(struct.pack('<ii',0,-1)+pcm.tobytes()) |
| await asyncio.sleep(.02) |
| await asyncio.wait_for(enrolled.wait(),12) |
| for _ in range(100): |
| if chunks:break |
| await asyncio.sleep(.1) |
| await ws.send_str('interrupt') |
| report={'teacher_filter_enabled':enabled.is_set(),'enrolled_from_asr_full_sentence':enrolled.is_set(), |
| 'identity_decisions':decisions,'events':events,'reply_audio_chunks':chunks, |
| 'backchannel_chunks':backchannels, |
| 'fixture':'public synthetic speech, not a private microphone recording', |
| 'playback':'virtual immediate acknowledgments, not physical sync evidence'} |
| print(json.dumps(report,indent=2));assert enrolled.is_set() and chunks>0 |
| if args.report:Path(args.report).write_text(json.dumps(report,indent=2)) |
| finally: |
| await ws.close();task.cancel();await asyncio.gather(task,return_exceptions=True) |
|
|
|
|
| if __name__=='__main__': |
| parser=argparse.ArgumentParser();parser.add_argument('--audio',required=True) |
| parser.add_argument('--url',default='http://127.0.0.1:8793');parser.add_argument('--report') |
| asyncio.run(run(parser.parse_args())) |
|
|
