Implement cookie-less session management and user ID handling for Hugging Face Spaces compatibility

- Introduced cookie-less session management by storing user IDs in localStorage and transmitting them via HTTP headers and URL parameters.
- Updated FastAPI endpoints to retrieve user IDs from headers and query parameters, ensuring backward compatibility with cookie-based implementations.
- Enhanced the RAG pipeline to accept user IDs for tracking in Hugging Face deployments.
- Modified frontend to include user ID in all API requests using an axios interceptor, improving session management across the application.
- Updated documentation to reflect changes in session management and user ID handling.

HUGGINGFACE.md

@@ -16,6 +16,7 @@ This RAG (Retrieval-Augmented Generation) Chat application allows users to:
 - **Interactive Chat:** Ask questions about the uploaded documents
 - **Knowledge Quiz:** Test your understanding with automatically generated quizzes
 - **Thinking/Answer Format:** See the AI's reasoning process with expandable "thinking" sections
+- **Cookie-less Session Management:** Uses localStorage and HTTP headers for user sessions instead of cookies for better HuggingFace Spaces compatibility
 
 ## Deployment Instructions
 
@@ -63,6 +64,16 @@ You can customize your deployment by modifying:
 - `api/main.py` - Backend API endpoints
 - `app/frontend/src/` - Frontend React components
 
+## Notes on Session Management
+
+This application implements cookie-less session management specifically for Hugging Face Spaces:
+
+- User IDs are stored in the browser's localStorage instead of cookies
+- User identification is transmitted via HTTP headers and URL parameters
+- The application maintains backward compatibility with cookie-based implementations
+
+This approach ensures the application works smoothly in HuggingFace Spaces which has limitations with cookie handling.
+
 ## Troubleshooting
 
 If you encounter issues:

api/main.py

@@ -139,28 +139,22 @@ class GenerateQuizResponse(BaseModel):
 
 # Helper function to get or create a user ID
 def get_or_create_user_id(request: Request, response: Response) -> str:
-    # Try to get user ID from cookie
-    user_id = request.cookies.get("user_id")
+    # Try to get user ID from header first
+    user_id = request.headers.get("X-User-ID")
+    
+    # Then try to get from query parameter
+    if not user_id:
+        user_id = request.query_params.get("user_id")
     
     # If no user ID exists, create a new one
     if not user_id:
         user_id = str(uuid.uuid4())
-        # Set cookie with long expiration (1 year)
-        expires = int(time.time()) + 31536000  # 1 year in seconds
-        response.set_cookie(
-            key="user_id",
-            value=user_id,
-            expires=expires,
-            path="/",
-            httponly=True,
-            samesite="lax"
-        )
-        
         # Initialize with default prompts
-        user_prompts[user_id] = {
-            "system_template": DEFAULT_SYSTEM_TEMPLATE,
-            "user_template": DEFAULT_USER_TEMPLATE
-        }
+        if user_id not in user_prompts:
+            user_prompts[user_id] = {
+                "system_template": DEFAULT_SYSTEM_TEMPLATE,
+                "user_template": DEFAULT_USER_TEMPLATE
+            }
     
     return user_id
 
@@ -318,7 +312,7 @@ async def query(request: QueryRequest):
         )
     
     # Run the query
-    result = await retrieval_pipeline.arun_pipeline(request.query)
+    result = await retrieval_pipeline.arun_pipeline(request.query, user_id)
     
     # Process the result and return the response
     response_text = ""
@@ -348,11 +342,11 @@ async def stream_query(request: QueryRequest):
         )
     
     # Run the query
-    result = await retrieval_pipeline.arun_pipeline(request.query)
+    result = await retrieval_pipeline.arun_pipeline(request.query, user_id)
     
     async def generate():
         async for chunk in result["response"]:
-            yield f"data: {json.dumps({'chunk': chunk, 'session_id': session_id})}\n\n"
+            yield f"data: {json.dumps({'chunk': chunk, 'session_id': session_id, 'user_id': user_id})}\n\n"
     
     return StreamingResponse(
         generate(), 
@@ -391,14 +385,14 @@ async def stream_query_get(
         )
     
     # Run the query
-    result = await retrieval_pipeline.arun_pipeline(query)
+    result = await retrieval_pipeline.arun_pipeline(query, user_id)
     
     async def generate():
         async for chunk in result["response"]:
-            yield f"data: {json.dumps({'chunk': chunk, 'session_id': session_id})}\n\n"
+            yield f"data: {json.dumps({'chunk': chunk, 'session_id': session_id, 'user_id': user_id})}\n\n"
         
         # Send an event to signal completion
-        yield f"event: complete\ndata: {json.dumps({'session_id': session_id})}\n\n"
+        yield f"event: complete\ndata: {json.dumps({'session_id': session_id, 'user_id': user_id})}\n\n"
     
     return StreamingResponse(
         generate(), 
@@ -799,7 +793,7 @@ class RetrievalAugmentedQAPipeline:
         self.system_role_prompt = SystemRolePrompt(system_template)
         self.user_role_prompt = UserRolePrompt(user_template)
 
-    async def arun_pipeline(self, user_query: str):
+    async def arun_pipeline(self, user_query: str, user_id: str):
         context_list = self.vector_db_retriever.search_by_text(user_query, k=4)
 
         context_prompt = ""

api/services/pipeline.py

@@ -21,7 +21,17 @@ class RetrievalAugmentedQAPipeline:
         self.system_role_prompt = SystemRolePrompt(system_template)
         self.user_role_prompt = UserRolePrompt(user_template)
 
-    async def arun_pipeline(self, user_query: str):
+    async def arun_pipeline(self, user_query: str, user_id: str = None):
+        """
+        Run the RAG pipeline to answer a user query
+        
+        Args:
+            user_query: The user's question
+            user_id: Optional user ID for tracking in HuggingFace deployment
+            
+        Returns:
+            Dictionary containing response generator and context
+        """
         context_list = self.vector_db_retriever.search_by_text(user_query, k=4)
 
         context_prompt = ""
@@ -37,4 +47,10 @@ class RetrievalAugmentedQAPipeline:
             async for chunk in self.llm.astream([formatted_system_prompt, formatted_user_prompt]):
                 yield chunk
 
-        return {"response": generate_response(), "context": context_list} 
+        result = {"response": generate_response(), "context": context_list}
+        
+        # Include user_id in result if provided (for HuggingFace deployment)
+        if user_id:
+            result["user_id"] = user_id
+            
+        return result 

api/utils/user.py

@@ -5,7 +5,7 @@ from .prompts import user_prompts, DEFAULT_SYSTEM_TEMPLATE, DEFAULT_USER_TEMPLAT
 
 def get_or_create_user_id(request: Request, response: Response) -> str:
     """
-    Get or create a user ID
+    Get or create a user ID without using cookies to support HuggingFace deployments
     
     Args:
         request: FastAPI request object
@@ -14,27 +14,21 @@ def get_or_create_user_id(request: Request, response: Response) -> str:
     Returns:
         User ID (either existing or newly created)
     """
-    # Try to get user ID from cookie
-    user_id = request.cookies.get("user_id")
+    # Try to get user ID from header first
+    user_id = request.headers.get("X-User-ID")
+    
+    # Then try to get from query parameter
+    if not user_id:
+        user_id = request.query_params.get("user_id")
     
     # If no user ID exists, create a new one
     if not user_id:
         user_id = str(uuid.uuid4())
-        # Set cookie with long expiration (1 year)
-        expires = int(time.time()) + 31536000  # 1 year in seconds
-        response.set_cookie(
-            key="user_id",
-            value=user_id,
-            expires=expires,
-            path="/",
-            httponly=True,
-            samesite="lax"
-        )
-        
         # Initialize with default prompts
-        user_prompts[user_id] = {
-            "system_template": DEFAULT_SYSTEM_TEMPLATE,
-            "user_template": DEFAULT_USER_TEMPLATE
-        }
+        if user_id not in user_prompts:
+            user_prompts[user_id] = {
+                "system_template": DEFAULT_SYSTEM_TEMPLATE,
+                "user_template": DEFAULT_USER_TEMPLATE
+            }
     
     return user_id 

app/frontend/src/App.js

@@ -8,9 +8,11 @@ import { ThemeProvider } from './components/ui/theme-provider';
 import { ThemeToggle } from './components/ui/theme-toggle';
 import { SettingsDialog } from './components/ui/settings-dialog';
 import { getVersionString, fetchApiVersion } from './utils/version';
-import { identifyUser } from './utils/user';
+import { identifyUser, setupUserIdInterceptor } from './utils/user';
 import PromptEditor from './components/ui/PromptEditor';
-import CookieConsent from './components/ui/CookieConsent';
+
+// Setup the axios interceptor to include user ID in all requests
+setupUserIdInterceptor();
 
 function App() {
   const [sessionId, setSessionId] = useState('');
@@ -219,8 +221,6 @@ function App() {
             </div>
           </div>
         </footer>
-        
-        <CookieConsent />
       </div>
     </ThemeProvider>
   );

app/frontend/src/utils/user.js

@@ -11,11 +11,19 @@ let cachedUserId = null;
  */
 export const identifyUser = async () => {
   try {
-    // Return cached user ID if available
+    // First check localStorage for existing user ID
+    const storedUserId = localStorage.getItem('user_id');
+    
+    // Return cached or stored user ID if available
     if (cachedUserId) {
       return cachedUserId;
     }
     
+    if (storedUserId) {
+      cachedUserId = storedUserId;
+      return storedUserId;
+    }
+    
     // Get user ID from the server
     const response = await axios.get('/identify');
     const userId = response.data.user_id;
@@ -23,6 +31,9 @@ export const identifyUser = async () => {
     // Cache the user ID for future use
     cachedUserId = userId;
     
+    // Store in localStorage for persistence across sessions
+    localStorage.setItem('user_id', userId);
+    
     return userId;
   } catch (error) {
     console.error('Error identifying user:', error);
@@ -36,10 +47,39 @@ export const identifyUser = async () => {
  * @returns {string|null} The cached user ID
  */
 export const getCachedUserId = () => {
+  if (!cachedUserId) {
+    // Try to get from localStorage if not in memory
+    cachedUserId = localStorage.getItem('user_id');
+  }
   return cachedUserId;
 };
 
+/**
+ * Configure axios to automatically include user ID in requests
+ */
+export const setupUserIdInterceptor = () => {
+  axios.interceptors.request.use(function (config) {
+    const userId = getCachedUserId();
+    if (userId) {
+      // Add user ID as a header
+      config.headers['X-User-ID'] = userId;
+      
+      // Add user ID as a URL parameter for GET requests
+      if (config.method === 'get') {
+        config.params = config.params || {};
+        if (!config.params.user_id) {
+          config.params.user_id = userId;
+        }
+      }
+    }
+    return config;
+  }, function (error) {
+    return Promise.reject(error);
+  });
+};
+
 export default {
   identifyUser,
-  getCachedUserId
+  getCachedUserId,
+  setupUserIdInterceptor
 };