APPS/PAILot.git - git.mnsoft.org

..	..	@@ -47,14 +47,16 @@
47	47	void initState() {
48	48	super.initState();
49	49	WidgetsBinding.instance.addObserver(this);
50		- _loadLastSeq();
51		- _initConnection();
	50	+ _initAll();
52	51	_scrollController.addListener(_onScroll);
53	52	}
54	53
55		- Future<void> _loadLastSeq() async {
	54	+ Future<void> _initAll() async {
	55	+ // Load lastSeq BEFORE connecting so catch_up sends the right value
56	56	final prefs = await SharedPreferences.getInstance();
57	57	_lastSeq = prefs.getInt('lastSeq') ?? 0;
	58	+ if (!mounted) return;
	59	+ _initConnection();
58	60	}
59	61
60	62	void _saveLastSeq() {
..	..	@@ -538,25 +540,8 @@
538	540	textCaption = '';
539	541	}
540	542
541		- // Send all images together — first with caption, rest without
542		- for (var i = 0; i < encodedImages.length; i++) {
543		- final isFirst = i == 0;
544		- final msgCaption = isFirst ? textCaption : '';
545		-
546		- _ws?.send({
547		- 'type': 'image',
548		- 'imageBase64': encodedImages[i],
549		- 'mimeType': 'image/jpeg',
550		- 'caption': msgCaption,
551		- if (isFirst && voiceB64 != null) 'audioBase64': voiceB64,
552		- 'sessionId': ref.read(activeSessionIdProvider),
553		- // Signal how many images follow so receiving session can wait
554		- if (isFirst && encodedImages.length > 1)
555		- 'totalImages': encodedImages.length,
556		- });
557		- }
558		-
559		- // If voice caption, also send the voice message so it gets transcribed
	543	+ // Send voice FIRST so Whisper transcribes it and the [PAILot:voice] prefix
	544	+ // sets the reply channel. Images follow — Claude sees transcript + images together.
560	545	if (voiceB64 != null) {
561	546	final voiceMsg = Message.voice(
562	547	role: MessageRole.user,
..	..	@@ -569,6 +554,20 @@
569	554	'audioBase64': voiceB64,
570	555	'content': '',
571	556	'messageId': voiceMsg.id,
	557	+ 'sessionId': ref.read(activeSessionIdProvider),
	558	+ });
	559	+ }
	560	+
	561	+ // Send images — first with text caption (if any), rest without
	562	+ for (var i = 0; i < encodedImages.length; i++) {
	563	+ final isFirst = i == 0;
	564	+ final msgCaption = isFirst ? textCaption : '';
	565	+
	566	+ _ws?.send({
	567	+ 'type': 'image',
	568	+ 'imageBase64': encodedImages[i],
	569	+ 'mimeType': 'image/jpeg',
	570	+ 'caption': msgCaption,
572	571	'sessionId': ref.read(activeSessionIdProvider),
573	572	});
574	573	}
..	..	@@ -591,8 +590,9 @@
591	590	final captionController = TextEditingController();
592	591	String? voicePath;
593	592	bool isVoiceRecording = false;
	593	+ bool hasVoiceCaption = false;
594	594
595		- return showModalBottomSheet<String>(
	595	+ final result = await showModalBottomSheet<String>(
596	596	context: context,
597	597	isScrollControlled: true,
598	598	builder: (ctx) => StatefulBuilder(
..	..	@@ -611,58 +611,94 @@
611	611	style: Theme.of(ctx).textTheme.titleSmall,
612	612	),
613	613	const SizedBox(height: 12),
614		- TextField(
615		- controller: captionController,
616		- decoration: InputDecoration(
617		- hintText: 'Add a caption (optional)',
618		- border: const OutlineInputBorder(),
619		- suffixIcon: IconButton(
620		- icon: Icon(
621		- isVoiceRecording ? Icons.stop_circle : Icons.mic,
622		- color: isVoiceRecording ? Colors.red : null,
623		- ),
624		- onPressed: () async {
625		- if (isVoiceRecording) {
626		- final path = await AudioService.stopRecording();
627		- setSheetState(() => isVoiceRecording = false);
628		- if (path != null) {
629		- voicePath = path;
630		- captionController.text = '🎤 Voice caption recorded';
631		- }
632		- } else {
633		- final path = await AudioService.startRecording();
634		- if (path != null) {
635		- setSheetState(() => isVoiceRecording = true);
636		- }
637		- }
638		- },
	614	+ // Text caption input
	615	+ if (!isVoiceRecording && !hasVoiceCaption)
	616	+ TextField(
	617	+ controller: captionController,
	618	+ decoration: const InputDecoration(
	619	+ hintText: 'Add a text caption (optional)',
	620	+ border: OutlineInputBorder(),
	621	+ ),
	622	+ autofocus: true,
	623	+ maxLines: 3,
	624	+ ),
	625	+ // Voice recording indicator
	626	+ if (isVoiceRecording)
	627	+ Container(
	628	+ padding: const EdgeInsets.symmetric(vertical: 20),
	629	+ child: const Row(
	630	+ mainAxisAlignment: MainAxisAlignment.center,
	631	+ children: [
	632	+ Icon(Icons.fiber_manual_record, color: Colors.red, size: 16),
	633	+ SizedBox(width: 8),
	634	+ Text('Recording voice caption...', style: TextStyle(fontSize: 16)),
	635	+ ],
639	636	),
640	637	),
641		- autofocus: true,
642		- maxLines: 3,
643		- enabled: !isVoiceRecording,
644		- ),
	638	+ // Voice recorded confirmation
	639	+ if (hasVoiceCaption && !isVoiceRecording)
	640	+ Container(
	641	+ padding: const EdgeInsets.symmetric(vertical: 20),
	642	+ child: const Row(
	643	+ mainAxisAlignment: MainAxisAlignment.center,
	644	+ children: [
	645	+ Icon(Icons.check_circle, color: Colors.green, size: 20),
	646	+ SizedBox(width: 8),
	647	+ Text('Voice caption recorded', style: TextStyle(fontSize: 16)),
	648	+ ],
	649	+ ),
	650	+ ),
645	651	const SizedBox(height: 12),
	652	+ // Action row: mic/stop + cancel + send
646	653	Row(
647		- mainAxisAlignment: MainAxisAlignment.end,
648	654	children: [
	655	+ // Mic / Stop button — large and clear
	656	+ if (!hasVoiceCaption)
	657	+ IconButton.filled(
	658	+ onPressed: () async {
	659	+ if (isVoiceRecording) {
	660	+ final path = await AudioService.stopRecording();
	661	+ setSheetState(() {
	662	+ isVoiceRecording = false;
	663	+ if (path != null) {
	664	+ voicePath = path;
	665	+ hasVoiceCaption = true;
	666	+ }
	667	+ });
	668	+ } else {
	669	+ final path = await AudioService.startRecording();
	670	+ if (path != null) {
	671	+ setSheetState(() => isVoiceRecording = true);
	672	+ }
	673	+ }
	674	+ },
	675	+ icon: Icon(isVoiceRecording ? Icons.stop : Icons.mic),
	676	+ style: IconButton.styleFrom(
	677	+ backgroundColor: isVoiceRecording ? Colors.red : null,
	678	+ foregroundColor: isVoiceRecording ? Colors.white : null,
	679	+ ),
	680	+ ),
	681	+ const Spacer(),
649	682	TextButton(
650		- onPressed: () {
651		- if (isVoiceRecording) AudioService.cancelRecording();
652		- Navigator.pop(ctx);
	683	+ onPressed: () async {
	684	+ if (isVoiceRecording) {
	685	+ await AudioService.cancelRecording();
	686	+ }
	687	+ if (ctx.mounted) Navigator.pop(ctx);
653	688	},
654	689	child: const Text('Cancel'),
655	690	),
656	691	const SizedBox(width: 8),
657	692	FilledButton(
658		- onPressed: () {
659		- if (voicePath != null) {
660		- // Voice caption: send as voice message with images
661		- Navigator.pop(ctx, '__voice__:$voicePath');
662		- } else {
663		- Navigator.pop(ctx, captionController.text);
664		- }
665		- },
	693	+ onPressed: isVoiceRecording
	694	+ ? null // disable Send while recording
	695	+ : () {
	696	+ if (voicePath != null) {
	697	+ Navigator.pop(ctx, '__voice__:$voicePath');
	698	+ } else {
	699	+ Navigator.pop(ctx, captionController.text);
	700	+ }
	701	+ },
666	702	child: const Text('Send'),
667	703	),
668	704	],
..	..	@@ -673,6 +709,14 @@
673	709	),
674	710	),
675	711	);
	712	+
	713	+ // Safety net: clean up recording if sheet dismissed by swipe/tap outside
	714	+ if (isVoiceRecording) {
	715	+ await AudioService.cancelRecording();
	716	+ }
	717	+
	718	+ captionController.dispose();
	719	+ return result;
676	720	}
677	721
678	722	void _clearChat() {