feat: complete session — catalog, marketplace, WhatsApp, peer-to-peer, install scripts

Major features:
- Pixel-Perfect glassmorphism design (landing + POS + public catalog)
- OEM/Local catalog toggle with Nexpart taxonomy (14 groups, 108 subgroups, 558 part types)
- Marketplace B2B Phase 1 (bodegas, POs, status machine, WA+email notifications)
- Peer-to-peer inventory (multi-instance, LAN discovery)
- WhatsApp: photo→Vision AI, voice→Whisper, conversational quotations
- Smart unified search (VIN/plate/part_number/keyword auto-detect)
- Shop Supplies tab (vehicle-independent parts)
- Chatbot AI fallback chain (5 models) + response cache
- CSV inventory import tool + setup_instance.sh installer
- Tablet-responsive CSS + sidebar toggle
- Filters, export CSV, employee edit, business data save
- Quotation system (WA→POS) with auto-print on confirmation
- Live stats on landing page

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

pos/services/whatsapp_service.py

@@ -55,12 +55,63 @@ def logout():
 
 
 def process_incoming(webhook_data):
+    """Extract a normalized dict from a Baileys webhook payload.
+
+    Supports text messages, image messages, audio (voice notes), and video.
+    Media content comes pre-downloaded as base64 from the bridge so Python
+    doesn't have to re-authenticate with WhatsApp servers.
+
+    Returns:
+        dict with keys:
+            phone            — numeric phone (no JID suffix)
+            jid              — full remote JID (may be @s.whatsapp.net or @lid)
+            text             — text content (plain text or media caption)
+            from_me          — bool, True if we sent the message
+            message_id       — WhatsApp message ID
+            media_kind       — 'text' | 'image' | 'audio' | 'video'
+            media_base64     — base64 string if media, else None
+            media_mimetype   — e.g. 'image/jpeg', 'audio/ogg'
+            is_voice_note    — True for WhatsApp voice notes (audioMessage ptt)
+    """
     data = webhook_data.get('data', {})
     key = data.get('key', {})
     message = data.get('message', {})
+
+    # remoteJid can be phone@s.whatsapp.net or LID@lid
+    remote_jid = key.get('remoteJid', '')
+    phone = remote_jid.replace('@s.whatsapp.net', '').replace('@lid', '')
+
+    # The bridge now classifies and passes these extra fields. Fall back to
+    # the old parsing if they're missing (older bridge version).
+    media_kind = data.get('media_kind', 'text')
+    media_base64 = data.get('media_base64')
+    media_mimetype = data.get('media_mimetype')
+    media_caption = data.get('media_caption') or ''
+    is_voice_note = bool(data.get('media_ptt'))
+    push_name = data.get('push_name') or ''
+
+    # Text content:
+    #  - For 'text' messages → conversation or extendedTextMessage
+    #  - For 'image'/'video' → the caption (may be empty)
+    #  - For 'audio' → empty (filled in later by Whisper transcription)
+    if media_kind == 'text':
+        text = (
+            message.get('conversation', '')
+            or message.get('extendedTextMessage', {}).get('text', '')
+            or ''
+        )
+    else:
+        text = media_caption
+
     return {
-        'phone': key.get('remoteJid', '').replace('@s.whatsapp.net', ''),
-        'text': message.get('conversation', '') or message.get('extendedTextMessage', {}).get('text', ''),
+        'phone': phone,
+        'jid': remote_jid,
+        'text': text,
         'from_me': key.get('fromMe', False),
         'message_id': key.get('id', ''),
+        'media_kind': media_kind,
+        'media_base64': media_base64,
+        'media_mimetype': media_mimetype,
+        'is_voice_note': is_voice_note,
+        'push_name': push_name,
     }