test: qualify deployed realtime voice STT Hermes and audio delivery

scripts/qualify-voice-live.py

@@ -0,0 +1,107 @@
+#!/usr/bin/env python3
+"""Explicit live fake-microphone qualification. Never physical microphone evidence.
+
+Reads ignored protected deployment credentials; emits only safe qualification facts.
+Requires local synthetic PCM fixture and deployed Hope. No deployment changes.
+"""
+import argparse
+import json
+import re
+import time
+from pathlib import Path
+from playwright.sync_api import sync_playwright, expect
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--url', default='https://192.168.86.2:9178')
+    parser.add_argument('--credentials', default='.runtime/deployment.json')
+    parser.add_argument('--wav', default='.runtime/voice-qualification/microphone.wav')
+    parser.add_argument('--timeout', type=int, default=180)
+    args = parser.parse_args()
+    values = json.loads(Path(args.credentials).read_text())
+    fixture = Path(args.wav).resolve(strict=True)
+    receipt = {'mode': 'synthetic_browser_microphone_not_physical', 'connected': False,
+               'user_transcript': False, 'assistant_transcript': False,
+               'audio_track': False, 'audio_bytes': 0, 'errors': 0}
+    with sync_playwright() as playwright:
+        browser = playwright.chromium.launch(headless=True, args=[
+            '--use-fake-ui-for-media-stream', '--use-fake-device-for-media-stream',
+            '--use-file-for-fake-audio-capture=' + str(fixture),
+            '--autoplay-policy=no-user-gesture-required'])
+        context = browser.new_context(ignore_https_errors=True, permissions=['microphone'],
+                                      viewport={'width': 1440, 'height': 1000})
+        page = context.new_page()
+        page.add_init_script("window.__qualificationVoiceID=null;window.addEventListener('hope:voice-state',event=>{if(event.detail?.conversation_id)window.__qualificationVoiceID=event.detail.conversation_id;});")
+        page.on('pageerror', lambda error: receipt.update(errors=receipt['errors'] + 1))
+        # Read-only receiver stats: exposes no tokens, credentials or private payloads.
+        page.add_init_script('''window.__voiceStats={bytes:0,packets:0};
+            const original=RTCPeerConnection.prototype.getStats;
+            const connections=[]; const Native=window.RTCPeerConnection;
+            window.RTCPeerConnection=new Proxy(Native,{construct(target,args){
+                const connection=Reflect.construct(target,args); connections.push(connection); return connection;}});
+            setInterval(async()=>{for(const connection of connections){try{
+                const report=await original.call(connection);
+                report.forEach(item=>{if(item.type==='inbound-rtp'&&(item.kind==='audio'||item.mediaType==='audio')){
+                    window.__voiceStats.bytes=Math.max(window.__voiceStats.bytes,item.bytesReceived||0);
+                    window.__voiceStats.packets=Math.max(window.__voiceStats.packets,item.packetsReceived||0);}});
+            }catch{}}},500);''')
+        try:
+            page.goto(args.url.rstrip('/') + '/', wait_until='networkidle')
+            expect(page.locator('#loginSubmit')).to_be_enabled(timeout=30000)
+            page.locator('#loginUsername').fill('admin')
+            page.locator('#loginPassword').fill(values['admin_password'])
+            page.locator('#loginSubmit').click()
+            expect(page.locator('#primaryInput')).to_be_enabled(timeout=30000)
+            # Prefer restored prominent voice entry, tolerate current compatibility ID.
+            for selector in ['#talkToHope', '#talkVoice', '#openVoice']:
+                if page.locator(selector).count() and page.locator(selector).is_visible():
+                    page.locator(selector).click()
+                    break
+            expect(page.locator('#voiceModeBtn')).to_be_visible(timeout=10000)
+            page.locator('#voiceModeBtn').click()
+            expect(page.locator('#startBtn')).to_be_enabled(timeout=10000)
+            page.locator('#startBtn').click()
+            deadline = time.monotonic() + args.timeout
+            while time.monotonic() < deadline:
+                status = page.locator('#status').inner_text()
+                body = page.locator('#messages').inner_text() if page.locator('#messages').count() else ''
+                receipt['connected'] |= 'Microphone on' in status or 'Secure room' in status
+                # Synthetic question and expected answer facts, not transcript dumps.
+                user_text = ' '.join(page.locator('#messages .message-row.user').all_inner_texts()).lower()
+                receipt['user_transcript'] |= bool(re.search(r'(?:two|2)\s*(?:plus|\+)\s*(?:two|2)', user_text))
+                assistant_text = ' '.join(page.locator('#messages .message-row.agent').all_inner_texts()).lower()
+                receipt['assistant_transcript'] |= 'four' in assistant_text or '4' in assistant_text
+                receipt['audio_track'] |= page.locator('audio[id^="audio-"]').count() > 0
+                stats = page.evaluate('window.__voiceStats')
+                receipt['audio_bytes'] = stats['bytes']
+                if receipt['assistant_transcript'] and receipt['audio_bytes'] > 0:
+                    break
+                page.wait_for_timeout(500)
+            receipt['audio_playing'] = page.evaluate('''Array.from(document.querySelectorAll('audio[id^="audio-"]')).some(el=>!el.paused&&el.srcObject&&el.srcObject.getAudioTracks().some(track=>track.readyState==='live'))''')
+            output = Path('.runtime/voice-qualification/live-receipt.json')
+            output.write_text(json.dumps(receipt, indent=2) + '\n')
+            print(json.dumps(receipt, sort_keys=True))
+            assert receipt['connected'], 'synthetic_voice_connection_failed'
+            assert receipt['user_transcript'], 'synthetic_voice_stt_not_observed'
+            assert receipt['assistant_transcript'], 'synthetic_voice_answer_not_observed'
+            assert receipt['audio_track'] and receipt['audio_bytes'] > 0, 'remote_tts_audio_not_observed'
+        finally:
+            if page.locator('#stopBtn').count() and page.locator('#stopBtn').is_visible() and page.locator('#stopBtn').is_enabled():
+                page.locator('#stopBtn').click()
+            # Delete only the ID captured for this newly created qualification
+            # voice conversation. Never enumerate or remove other account chats.
+            try:
+                cleanup = page.evaluate('''async()=>{let removed=false;const cid=window.__qualificationVoiceID;
+                    if(cid){try{const result=await HopeAuth.request('/api/conversations/'+encodeURIComponent(cid),{method:'DELETE'});removed=result.ok===true;}catch{}}
+                    try{await HopeAuth.request('/api/auth/logout',{method:'POST'});}catch{}
+                    return {created_conversation_seen:!!cid,removed};}''')
+                print(json.dumps({'qualification_cleanup': cleanup}, sort_keys=True))
+            except Exception:
+                print('qualification_cleanup_unavailable')
+            context.close()
+            browser.close()
+
+
+if __name__ == '__main__':
+    main()