Follow up

INSTALL.md

@@ -4,7 +4,7 @@
 
 - [Docker](https://docs.docker.com/get-docker/) `19.0.3` or later.
 - [Docker Compose](https://docs.docker.com/compose/install/) (`docker compose` or `docker-compose`) `2.6.1` or later.
-- OpenAI API key (for text generation and embeddings). Get it from [here](https://platform.openai.com/api-keys).
+- OpenAI API key (for answer generation and embeddings). Get it from [here](https://platform.openai.com/api-keys).
 - Firecrawl API key (for website scraping). Get it from [here](https://www.firecrawl.dev/app/api-keys).
 
 ### Quick Install
@@ -131,4 +131,4 @@ Here's a detailed comparison between Gurubase Cloud and Self-hosted versions:
 | Knowledge Base Sources | ✅ Websites, YouTube, PDFs | ✅ Websites, YouTube, PDFs |
 | GitHub Codebase Indexing | ✅ Available | ✅ Available |
 | Website Widget | ✅ Available | ✅ Available |
-| Base LLM | ✅ OpenAI GPT-4o | ✅ OpenAI GPT-4o |
+| Base LLM | ✅ OpenAI GPT-4o | ✅ OpenAI GPT-4o |

src/gurubase-backend/backend/accounts/views.py

@@ -35,7 +35,7 @@ def get_auth_provider(connection_name):
 @auth_auth0
 def auth0_user_login(request):
     # Example Request Body:
-    # '{"auth0_id": "auth0|5f7c8ec7c33c6c004bbafe82", "email": "[email protected]", "name": "John Smith", "picture": "http://www.gravatar.com/avatar/?d=identicon", "full_request": {"connection": {"id": "con_fpe5kj482KO1eOzQ", "metadata": {}, "name": "Username-Password-Authentication", "strategy": "auth0"}, "request": {"geoip": {"cityName": "Bellevue", "continentCode": "NA", "countryCode": "US", "countryCode3": "USA", "countryName": "United States of America", "latitude": 47.61793, "longitude": -122.19584, "subdivisionCode": "WA", "subdivisionName": "Washington", "timeZone": "America/Los_Angeles"}, "hostname": "dev-o0e6bx1accmw5dg4.example.com", "ip": "13.33.86.47", "language": "en", "method": "POST", "user_agent": "curl/7.64.1"}, "tenant": {"id": "dev-o0e6bx1accmw5dg4"}, "transaction": {"acr_values": [], "id": "", "locale": "", "login_hint": "[email protected]", "prompt": ["none"], "protocol": "oauth2-access-token", "redirect_uri": "http://someuri.com", "requested_scopes": [], "response_mode": "form_post", "response_type": ["id_token"], "state": "AABBccddEEFFGGTTasrs", "ui_locales": []}, "user": {"app_metadata": {}, "created_at": "2024-11-01T12:33:27.606Z", "email": "[email protected]", "email_verified": true, "family_name": "Smith", "given_name": "John", "last_password_reset": "2024-11-01T12:33:27.606Z", "name": "John Smith", "nickname": "j+smith", "phoneNumber": "123-123-1234", "phone_number": "123-123-1234", "phone_verified": true, "picture": "http://www.gravatar.com/avatar/?d=identicon", "tenant": "dev-o0e6bx1accmw5dg4", "updated_at": "2024-11-01T12:33:27.606Z", "user_id": "auth0|5f7c8ec7c33c6c004bbafe82", "user_metadata": {}, "username": "j+smith"}, "configuration": {}, "secrets": {"AUTH0_MANAGEMENT_API_TOKEN": "ebdcad69aa2f6d6637761debf2f1bec02b023114bad350c3dce0769bd2ebb604", "BACKEND_URL": "https://fatih.ddosify.com/"}}}'
+    # '{"auth0_id": "auth0|5f7c8ec7c33c6c004bbafe82", "email": "[email protected]", "name": "John Smith", "picture": "http://www.gravatar.com/avatar/?d=identicon", "full_request": {"connection": {"id": "con_fpe5kj482KO1eOzQ", "metadata": {}, "name": "Username-Password-Authentication", "strategy": "auth0"}, "request": {"geoip": {"cityName": "Bellevue", "continentCode": "NA", "countryCode": "US", "countryCode3": "USA", "countryName": "United States of America", "latitude": 47.61793, "longitude": -122.19584, "subdivisionCode": "WA", "subdivisionName": "Washington", "timeZone": "America/Los_Angeles"}, "hostname": "dev-o0e6bx1accmw5dg4.example.com", "ip": "13.33.86.47", "language": "en", "method": "POST", "user_agent": "curl/7.64.1"}, "tenant": {"id": "dev-o0e6bx1accmw5dg4"}, "transaction": {"acr_values": [], "id": "", "locale": "", "login_hint": "[email protected]", "prompt": ["none"], "protocol": "oauth2-access-token", "redirect_uri": "http://someuri.com", "requested_scopes": [], "response_mode": "form_post", "response_type": ["id_token"], "state": "AABBccddEEFFGGTTasrs", "ui_locales": []}, "user": {"app_metadata": {}, "created_at": "2024-11-01T12:33:27.606Z", "email": "[email protected]", "email_verified": true, "family_name": "Smith", "given_name": "John", "last_password_reset": "2024-11-01T12:33:27.606Z", "name": "John Smith", "nickname": "j+smith", "phoneNumber": "123-123-1234", "phone_number": "123-123-1234", "phone_verified": true, "picture": "http://www.gravatar.com/avatar/?d=identicon", "tenant": "dev-o0e6bx1accmw5dg4", "updated_at": "2024-11-01T12:33:27.606Z", "user_id": "auth0|5f7c8ec7c33c6c004bbafe82", "user_metadata": {}, "username": "j+smith"}, "configuration": {}, "secrets": {"AUTH0_MANAGEMENT_API_TOKEN": "ebdcad69aa2f6d6637761debf2f1bec02b023114bad350c3dce0769bd2ebb604", "BACKEND_URL": ""}}}'
 
     request_body = request.data
 

src/gurubase-backend/backend/backend/settings.py

@@ -379,7 +379,7 @@
 
 FOLLOW_UP_QUESTION_LIMIT = config('FOLLOW_UP_QUESTION_LIMIT', default=100, cast=int)
 FOLLOW_UP_QUESTION_TIME_LIMIT_SECONDS = config('FOLLOW_UP_QUESTION_TIME_LIMIT_SECONDS', default=7200, cast=int) # 2 hours
-GENERATE_FOLLOW_UP_EXAMPLES = config('GENERATE_FOLLOW_UP_EXAMPLES', default=True, cast=bool)
+GENERATE_FOLLOW_UP_EXAMPLES = config('GENERATE_FOLLOW_UP_EXAMPLES', default=False, cast=bool)
 FOLLOW_UP_EXAMPLE_COUNT = config('FOLLOW_UP_EXAMPLE_COUNT', default=3, cast=int)
 
 BINGE_HISTORY_PAGE_SIZE = config('BINGE_HISTORY_PAGE_SIZE', default=30, cast=int)

src/gurubase-backend/backend/core/auth.py

@@ -212,3 +212,49 @@ def wrapper(request, *args, **kwargs):
         request.auth0_id = api_key_obj.user.auth0_id
         return view_func(request, *args, **kwargs)
     return wrapper
+
+
+def follow_up_examples_auth(view_func):
+    @wraps(view_func)
+    def wrapper(request, *args, **kwargs):
+        # First try JWT auth
+        auth_header = request.headers.get('Authorization', '')
+
+        if auth_header.startswith('Bearer '):
+            token = auth_header.split(' ')[1]
+            try:
+                # Fetch Auth0 public keys
+                jwks_url = f'{settings.AUTH0_DOMAIN}.well-known/jwks.json'
+                jwks_client = PyJWKClient(jwks_url)
+                signing_key = jwks_client.get_signing_key_from_jwt(token)
+
+                # Decode and validate the token
+                payload = jwt.decode(
+                    token,
+                    signing_key.key,
+                    algorithms=['RS256'],
+                    audience=settings.AUTH0_AUDIENCE,
+                    issuer=settings.AUTH0_DOMAIN
+                )
+
+                auth0_id = payload['sub']
+                if auth0_id:
+                    user = User.objects.filter(auth0_id=auth0_id).first()
+                    if user:
+                        request.auth0_id = auth0_id
+                        request.user = user
+            except Exception as e:
+                logger.debug(f"JWT validation failed: {str(e)}")
+                # Continue even if JWT auth fails
+                pass
+
+        # Check for widget ID
+        widget_id = request.headers.get('Authorization')
+        if widget_id:
+            widget_id_obj = WidgetId.validate_key(widget_id)
+            if widget_id_obj:
+                request.widget = True
+
+        # Allow the request to proceed regardless of auth status
+        return view_func(request, *args, **kwargs)
+    return wrapper

src/gurubase-backend/backend/core/tests/test_utils.py

@@ -88,7 +88,7 @@ def test_anonymous_user_widget_search(self):
             guru_type_object=self.guru_type,
             binge=None,
             slug="widget-question",
-            include_widget=True
+            only_widget=True
         )
         self.assertEqual(result, self.widget_question)
 
@@ -98,7 +98,7 @@ def test_anonymous_user_widget_search(self):
             guru_type_object=self.guru_type,
             binge=None,
             slug="regular-question",
-            include_widget=True
+            only_widget=True
         )
         self.assertIsNone(result)
 

src/gurubase-backend/backend/core/utils.py

@@ -2553,14 +2553,14 @@ def check_binge_auth(binge, user):
         return True
     return binge.owner == user
 
-def search_question(user, guru_type_object, binge, slug=None, question=None, will_check_binge_auth=True, include_api=False, include_widget=False):
+def search_question(user, guru_type_object, binge, slug=None, question=None, will_check_binge_auth=True, include_api=False, only_widget=False):
     def get_source_conditions(user):
         """Helper function to get source conditions based on user"""
         if user is None:
             # For anonymous users
             # API requests are not allowed
             # Widget requests are allowed
-            if include_widget:
+            if only_widget:
                 return Q(source__in=[Question.Source.WIDGET_QUESTION.value])
             else:
                 return ~Q(source__in=[Question.Source.API.value, Question.Source.WIDGET_QUESTION.value])
@@ -2866,7 +2866,7 @@ def api_ask(question: str,
     is_widget = api_type == APIType.WIDGET
 
     include_api = api_type == APIType.API
-    include_widget = api_type == APIType.WIDGET
+    only_widget = api_type == APIType.WIDGET
 
     question_source = {
         APIType.WIDGET: Question.Source.WIDGET_QUESTION.value,
@@ -2884,7 +2884,7 @@ def api_ask(question: str,
             question,
             will_check_binge_auth=False,
             include_api=include_api,
-            include_widget=include_widget
+            only_widget=only_widget
         )
         if existing_question and not is_question_dirty(existing_question):
             logger.info(f"Found existing question {question} for guru type {guru_type.slug}")
@@ -2902,7 +2902,7 @@ def api_ask(question: str,
             question,
             will_check_binge_auth=False,
             include_api=include_api,
-            include_widget=include_widget
+            only_widget=only_widget
         )
         if existing_question and not is_question_dirty(existing_question):
             logger.info(f"Found existing question with slug for {question} in guru type {guru_type.slug}")

src/gurubase-backend/backend/core/views.py

@@ -13,7 +13,7 @@
 from core.requester import GeminiRequester, OpenAIRequester, RerankerRequester
 from core.data_sources import PDFStrategy, WebsiteStrategy, YouTubeStrategy, GitHubRepoStrategy
 from core.serializers import WidgetIdSerializer, BingeSerializer, DataSourceSerializer, GuruTypeSerializer, GuruTypeInternalSerializer, QuestionCopySerializer, FeaturedDataSourceSerializer
-from core.auth import auth, jwt_auth, combined_auth, stream_combined_auth, api_key_auth
+from core.auth import auth, follow_up_examples_auth, jwt_auth, combined_auth, stream_combined_auth, api_key_auth
 from core.gcp import replace_media_root_with_nginx_base_url
 from core.models import FeaturedDataSource, Question, ContentPageStatistics, QuestionValidityCheckPricing, Summarization, WidgetId, Binge, DataSource, GuruType
 from accounts.models import User
@@ -1058,7 +1058,7 @@ def export_questions(request):
 
 
 @api_view(['POST'])
-@combined_auth
+@follow_up_examples_auth
 def follow_up_examples(request, guru_type):
     user = request.user
 
@@ -1070,6 +1070,7 @@ def follow_up_examples(request, guru_type):
     binge_id = request.data.get('binge_id')
     question_slug = request.data.get('question_slug')
     question_text = request.data.get('question')
+    widget = request.widget if hasattr(request, 'widget') else False
 
     if not question_slug and not question_text:
         return Response({'msg': 'Question slug is required'}, status=status.HTTP_400_BAD_REQUEST)
@@ -1082,7 +1083,7 @@ def follow_up_examples(request, guru_type):
     else:
         binge = None
 
-    if binge and not check_binge_auth(binge, user):
+    if binge and not widget and not check_binge_auth(binge, user):
         return Response({'msg': 'User does not have access to this binge'}, status=status.HTTP_401_UNAUTHORIZED)
 
     guru_type_object = get_guru_type_object(guru_type, only_active=True)
@@ -1092,7 +1093,9 @@ def follow_up_examples(request, guru_type):
         guru_type_object, 
         binge, 
         question_slug, 
-        question_text
+        question_text,
+        only_widget=widget,
+        will_check_binge_auth=not widget
     )
     if not last_question:
         return Response({'msg': 'Question does not exist'}, status=status.HTTP_400_BAD_REQUEST)
@@ -1111,7 +1114,19 @@ def follow_up_examples(request, guru_type):
     # Get relevant contexts from the last question
     contexts = []
     if last_question.processed_ctx_relevances and 'kept' in last_question.processed_ctx_relevances:
-        contexts = [x['context'] for x in last_question.processed_ctx_relevances['kept']]
+        for ctx in last_question.processed_ctx_relevances['kept']:
+            # Skip GitHub repo contexts
+            try:
+                # Extract metadata using regex pattern that matches any context number
+                context_parts = ctx['context'].split('\nContext ')
+                metadata_text = context_parts[1].split(' Text:')[0]
+                metadata_json = metadata_text.split('Metadata:\n')[1].replace("'", '"')
+                metadata = json.loads(metadata_json)
+                if metadata.get('type') == 'GITHUB_REPO':
+                    continue
+            except (json.JSONDecodeError, IndexError, KeyError):
+                pass  # If we can't parse metadata, include the context
+            contexts.append(ctx['context'])
 
     if not contexts:
         return Response([], status=status.HTTP_200_OK)
@@ -1324,7 +1339,7 @@ def ask_widget(request):
                 binge, 
                 parent_slug, 
                 will_check_binge_auth=False,
-                include_widget=True
+                only_widget=True
             )
         except Exception as e:
             return response_handler.handle_error_response("Parent question does not exist")

src/gurubase-frontend/src/components/Icons.js

@@ -254,14 +254,8 @@ export function LineMdTwitterXAlt(props) {
       xmlns="http://www.w3.org/2000/svg"
       {...props}>
       <path
-        d="M8.5 2h2.5L11 2h-2.5zM13 2h2.5L15.5 2h-2.5zM10.5 2h5v0h-5zM8.5 2h5v0h-5zM10 2h3.5L13.5 2h-3.5z"
-        fill="black">
-        <animate
-          attributeName="d"
-          dur="0.8s"
-          fill="freeze"
-          keyTimes="0;0.3;0.5;1"
-          values="M8.5 2h2.5L11 2h-2.5zM13 2h2.5L15.5 2h-2.5zM10.5 2h5v0h-5zM8.5 2h5v0h-5zM10 2h3.5L13.5 2h-3.5z;M8.5 2h2.5L11 22h-2.5zM13 2h2.5L15.5 22h-2.5zM10.5 2h5v2h-5zM8.5 20h5v2h-5zM10 2h3.5L13.5 22h-3.5z;M8.5 2h2.5L11 22h-2.5zM13 2h2.5L15.5 22h-2.5zM10.5 2h5v2h-5zM8.5 20h5v2h-5zM10 2h3.5L13.5 22h-3.5z"></animate>
+        d="M18.205 2.25h3.308l-7.227 8.26 8.502 11.24H16.13l-5.214-6.817L4.95 21.75H1.64l7.73-8.835L1.215 2.25H8.04l4.713 6.231zm-1.161 17.52h1.833L7.045 4.126H5.078z"
+        fill="currentColor">
       </path>
     </svg>
   );

src/gurubase-frontend/src/components/OtherGurus/MobileOtherGurus.js

@@ -47,7 +47,7 @@ const MobileOtherGurus = ({ allGuruTypes, isLongGuruName }) => {
         }}
         onClick={() => setIsPanelOpen(true)}>
         <div className="self-stretch my-auto">
-          {getGuruPromptMap(guruType, allGuruTypes)} Guru
+          {getGuruPromptMap(guruType, allGuruTypes)}
         </div>
         <Icon icon="tabler:chevron-down" />
       </div>