Add get chat messages endpoint

bots/bot_controller/bot_controller.py

@@ -23,6 +23,8 @@
     BotMediaRequestMediaTypes,
     BotMediaRequestStates,
     BotStates,
+    ChatMessage,
+    ChatMessageToOptions,
     Credentials,
     MeetingTypes,
     Participant,
@@ -80,6 +82,7 @@ def get_google_meet_bot_adapter(self):
             wants_any_video_frames_callback=None,
             add_mixed_audio_chunk_callback=None,
             upsert_caption_callback=self.closed_caption_manager.upsert_caption,
+            upsert_chat_message_callback=self.on_new_chat_message,
             automatic_leave_configuration=self.automatic_leave_configuration,
             add_encoded_mp4_chunk_callback=None,
             recording_view=self.bot_in_db.recording_view(),
@@ -102,6 +105,7 @@ def get_teams_bot_adapter(self):
             wants_any_video_frames_callback=None,
             add_mixed_audio_chunk_callback=None,
             upsert_caption_callback=self.closed_caption_manager.upsert_caption,
+            upsert_chat_message_callback=self.on_new_chat_message,
             automatic_leave_configuration=self.automatic_leave_configuration,
             add_encoded_mp4_chunk_callback=None,
             recording_view=self.bot_in_db.recording_view(),
@@ -137,6 +141,7 @@ def get_zoom_bot_adapter(self):
             add_video_frame_callback=self.gstreamer_pipeline.on_new_video_frame,
             wants_any_video_frames_callback=self.gstreamer_pipeline.wants_any_video_frames,
             add_mixed_audio_chunk_callback=self.gstreamer_pipeline.on_mixed_audio_raw_data_received_callback,
+            upsert_chat_message_callback=self.on_new_chat_message,
             automatic_leave_configuration=self.automatic_leave_configuration,
             video_frame_size=self.bot_in_db.recording_dimensions(),
         )
@@ -740,6 +745,41 @@ def save_individual_audio_utterance(self, message):
         process_utterance.delay(utterance.id)
         return
 
+    def on_new_chat_message(self, chat_message):
+        GLib.idle_add(lambda: self.upsert_chat_message(chat_message))
+
+    def upsert_chat_message(self, chat_message):
+        logger.info(f"Upserting chat message: {chat_message}")
+
+        participant = self.adapter.get_participant(chat_message["participant_uuid"])
+
+        if participant is None:
+            logger.warning(f"Warning: No participant found for chat message: {chat_message}")
+            return
+
+        participant, _ = Participant.objects.get_or_create(
+            bot=self.bot_in_db,
+            uuid=participant["participant_uuid"],
+            defaults={
+                "user_uuid": participant["participant_user_uuid"],
+                "full_name": participant["participant_full_name"],
+            },
+        )
+
+        ChatMessage.objects.update_or_create(
+            bot=self.bot_in_db,
+            source_uuid=chat_message["message_uuid"],
+            defaults={
+                "timestamp": chat_message["timestamp"],
+                "to": ChatMessageToOptions.ONLY_BOT if chat_message.get("to_bot") else ChatMessageToOptions.EVERYONE,
+                "text": chat_message["text"],
+                "participant": participant,
+                "additional_data": chat_message.get("additional_data", {}),
+            },
+        )
+
+        return
+
     def on_message_from_adapter(self, message):
         GLib.idle_add(lambda: self.take_action_based_on_message_from_adapter(message))
 

bots/bot_pod_creator/bot_pod_creator.py

@@ -25,7 +25,8 @@ def __init__(self, namespace: str = "attendee"):
 
         # Parse instance from version (matches your pattern of {hash}-{timestamp})
         self.app_instance = f"{self.app_name}-{self.app_version.split('-')[-1]}"
-        self.image = f"nduncan{self.app_name}/{self.app_name}:{self.app_version}"
+        default_pod_image = f"nduncan{self.app_name}/{self.app_name}"
+        self.image = f"{os.getenv('BOT_POD_IMAGE', default_pod_image)}:{self.app_version}"
 
     def create_bot_pod(
         self,

bots/bots_api_urls.py

@@ -44,6 +44,11 @@
         bots_api_views.SpeechView.as_view(),
         name="bot-speech",
     ),
+    path(
+        "bots/<str:object_id>/chat_messages",
+        bots_api_views.ChatMessagesView.as_view(),
+        name="bot-chat-messages",
+    ),
     path(
         "bots/<str:object_id>/delete_data",
         bots_api_views.DeleteDataView.as_view(),

bots/bots_api_views.py

@@ -10,6 +10,8 @@
     extend_schema,
 )
 from rest_framework import status
+from rest_framework.generics import GenericAPIView
+from rest_framework.pagination import CursorPagination
 from rest_framework.response import Response
 from rest_framework.views import APIView
 
@@ -24,6 +26,7 @@
     BotMediaRequestMediaTypes,
     BotMediaRequestStates,
     BotStates,
+    ChatMessage,
     Credentials,
     MediaBlob,
     MeetingTypes,
@@ -33,6 +36,7 @@
 from .serializers import (
     BotImageSerializer,
     BotSerializer,
+    ChatMessageSerializer,
     CreateBotSerializer,
     RecordingSerializer,
     SpeechSerializer,
@@ -703,3 +707,92 @@ def get(self, request, object_id):
 
         except Bot.DoesNotExist:
             return Response({"error": "Bot not found"}, status=status.HTTP_404_NOT_FOUND)
+
+
+class ChatMessageCursorPagination(CursorPagination):
+    ordering = "created_at"
+    page_size = 25
+
+
+class ChatMessagesView(GenericAPIView):
+    authentication_classes = [ApiKeyAuthentication]
+    pagination_class = ChatMessageCursorPagination
+    serializer_class = ChatMessageSerializer
+
+    @extend_schema(
+        operation_id="Get Chat Messages",
+        summary="Get chat messages sent in the meeting",
+        description="If the meeting is still in progress, this returns the chat messages sent so far. Results are paginated using cursor pagination.",
+        responses={
+            200: OpenApiResponse(
+                response=ChatMessageSerializer(many=True),
+                description="List of chat messages",
+            ),
+            404: OpenApiResponse(description="Bot not found"),
+        },
+        parameters=[
+            *TokenHeaderParameter,
+            OpenApiParameter(
+                name="object_id",
+                type=str,
+                location=OpenApiParameter.PATH,
+                description="Bot ID",
+                examples=[OpenApiExample("Bot ID Example", value="bot_xxxxxxxxxxx")],
+            ),
+            OpenApiParameter(
+                name="updated_after",
+                type={"type": "string", "format": "ISO 8601 datetime"},
+                location=OpenApiParameter.QUERY,
+                description="Only return chat messages created after this time. Useful when polling for updates.",
+                required=False,
+                examples=[OpenApiExample("DateTime Example", value="2024-01-18T12:34:56Z")],
+            ),
+            OpenApiParameter(
+                name="cursor",
+                type=str,
+                location=OpenApiParameter.QUERY,
+                description="Cursor for pagination",
+                required=False,
+            ),
+        ],
+        tags=["Bots"],
+    )
+    def get(self, request, object_id):
+        try:
+            # Get the bot and verify it belongs to the project
+            bot = Bot.objects.get(object_id=object_id, project=request.auth.project)
+
+            # Get optional updated_after parameter
+            updated_after = request.query_params.get("updated_after")
+
+            # Query messages for this bot
+            messages_query = ChatMessage.objects.filter(bot=bot)
+
+            # Filter by updated_after if provided
+            if updated_after:
+                try:
+                    updated_after_datetime = parse_datetime(str(updated_after))
+                except Exception:
+                    updated_after_datetime = None
+
+                if not updated_after_datetime:
+                    return Response(
+                        {"error": "Invalid updated_after format. Use ISO 8601 format (e.g., 2024-01-18T12:34:56Z)"},
+                        status=status.HTTP_400_BAD_REQUEST,
+                    )
+                messages_query = messages_query.filter(created_at__gt=updated_after_datetime)
+
+            # Apply ordering - now using created_at for cursor pagination
+            messages = messages_query.order_by("created_at")
+
+            # Let the pagination class handle the rest
+            page = self.paginate_queryset(messages)
+            if page is not None:
+                serializer = self.get_serializer(page, many=True)
+                return self.get_paginated_response(serializer.data)
+
+            serializer = self.get_serializer(messages, many=True)
+            return Response(serializer.data)
+
+        except Bot.DoesNotExist:
+            return Response({"error": "Bot not found"}, status=status.HTTP_404_NOT_FOUND)

bots/google_meet_bot_adapter/google_meet_chromedriver_payload.js

@@ -442,6 +442,31 @@ const DEVICE_OUTPUT_TYPE = {
     VIDEO: 2
 }
 
+// Chat message manager
+class ChatMessageManager {
+    constructor(ws) {
+        this.ws = ws;
+    }
+
+    handleChatMessage(chatMessageRaw) {
+        try {
+            const chatMessage = chatMessageRaw.chatMessage;
+            console.log('handleChatMessage', chatMessage);
+
+            this.ws.sendJson({
+                type: 'ChatMessage',
+                message_uuid: chatMessage.messageId,
+                participant_uuid: chatMessage.deviceId,
+                timestamp: Math.floor(chatMessage.timestamp / 1000),
+                text: chatMessage.chatMessageContent.text,
+            });
+        }
+        catch (error) {
+            console.error('Error in handleChatMessage', error);
+        }
+    }
+}
+
 // User manager
 class UserManager {
     constructor(ws) {
@@ -1204,6 +1229,7 @@ const captionManager = new CaptionManager(ws);
 const videoTrackManager = new VideoTrackManager(ws);
 const styleManager = new StyleManager();
 const receiverManager = new ReceiverManager();
+const chatMessageManager = new ChatMessageManager(ws);
 let rtpReceiverInterceptor = null;
 if (window.initialData.sendPerParticipantAudio) {
     rtpReceiverInterceptor = new RTCRtpReceiverInterceptor((receiver, result, ...args) => {
@@ -1215,6 +1241,7 @@ window.videoTrackManager = videoTrackManager;
 window.userManager = userManager;
 window.styleManager = styleManager;
 window.receiverManager = receiverManager;
+window.chatMessageManager = chatMessageManager;
 // Create decoders for all message types
 const messageDecoders = {};
 messageTypes.forEach(type => {
@@ -1261,7 +1288,9 @@ const handleCollectionEvent = (event) => {
 
   const chatMessageWrapper = collectionEvent.body.userInfoListWrapperAndChatWrapperWrapper?.userInfoListWrapperAndChatWrapper?.chatMessageWrapper;
   if (chatMessageWrapper) {
-    console.log('chatMessageWrapper', chatMessageWrapper);
+    for (const chatMessage of chatMessageWrapper) {
+        window.chatMessageManager?.handleChatMessage(chatMessage);
+    }
   }
 
   //console.log('deviceOutputInfoList', JSON.stringify(collectionEvent.body.userInfoListWrapperAndChatWrapperWrapper?.deviceInfoWrapper?.deviceOutputInfoList));

bots/migrations/0030_chatmessage.py

@@ -0,0 +1,30 @@
+# Generated by Django 5.1.2 on 2025-05-23 04:55
+
+import django.db.models.deletion
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('bots', '0029_recording_transcription_failure_data_and_more'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='ChatMessage',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('text', models.TextField()),
+                ('created_at', models.DateTimeField(auto_now_add=True)),
+                ('updated_at', models.DateTimeField(auto_now=True)),
+                ('to', models.IntegerField(choices=[(1, 'only_bot'), (2, 'everyone')])),
+                ('timestamp', models.IntegerField()),
+                ('additional_data', models.JSONField(default=dict)),
+                ('object_id', models.CharField(editable=False, max_length=32, unique=True)),
+                ('source_uuid', models.CharField(max_length=255, null=True, unique=True)),
+                ('bot', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, related_name='chat_messages', to='bots.bot')),
+                ('participant', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, related_name='chat_messages', to='bots.participant')),
+            ],
+        ),
+    ]

bots/models.py

@@ -1486,3 +1486,31 @@ def add_to_response_body_list(self, response_body):
             self.response_body_list = [response_body]
         else:
             self.response_body_list.append(response_body)
+
+
+class ChatMessageToOptions(models.IntegerChoices):
+    ONLY_BOT = 1, "only_bot"
+    EVERYONE = 2, "everyone"
+
+
+class ChatMessage(models.Model):
+    bot = models.ForeignKey(Bot, on_delete=models.CASCADE, related_name="chat_messages")
+    text = models.TextField()
+    created_at = models.DateTimeField(auto_now_add=True)
+    updated_at = models.DateTimeField(auto_now=True)
+    participant = models.ForeignKey(Participant, on_delete=models.CASCADE, related_name="chat_messages")
+    to = models.IntegerField(choices=ChatMessageToOptions.choices, null=False)
+    timestamp = models.IntegerField()
+    additional_data = models.JSONField(null=False, default=dict)
+    object_id = models.CharField(max_length=32, unique=True, editable=False)
+
+    OBJECT_ID_PREFIX = "msg_"
+    object_id = models.CharField(max_length=32, unique=True, editable=False)
+    source_uuid = models.CharField(max_length=255, null=True, unique=True)
+
+    def save(self, *args, **kwargs):
+        if not self.object_id:
+            # Generate a random 16-character string
+            random_string = "".join(random.choices(string.ascii_letters + string.digits, k=16))
+            self.object_id = f"{self.OBJECT_ID_PREFIX}{random_string}"
+        super().save(*args, **kwargs)

bots/serializers.py

@@ -16,6 +16,7 @@
     BotEventSubTypes,
     BotEventTypes,
     BotStates,
+    ChatMessageToOptions,
     MediaBlob,
     MeetingTypes,
     Recording,
@@ -733,3 +734,17 @@ def validate_text_to_speech_settings(self, value):
             raise serializers.ValidationError(e.message)
 
         return value
+
+
+class ChatMessageSerializer(serializers.Serializer):
+    object_id = serializers.CharField()
+    text = serializers.CharField()
+    timestamp = serializers.IntegerField()
+    to = serializers.SerializerMethodField()
+    sender_name = serializers.CharField(source="participant.full_name")
+    sender_uuid = serializers.CharField(source="participant.uuid")
+    sender_user_uuid = serializers.CharField(source="participant.user_uuid", allow_null=True)
+    additional_data = serializers.JSONField()
+
+    def get_to(self, obj):
+        return ChatMessageToOptions.choices[obj.to - 1][1]