feat(app): add toggle to enable/disable memory context in chat

.changeset/memory-context-toggle.md

@@ -0,0 +1,10 @@
+---
+"think-app": patch
+---
+
+feat(app): add toggle to enable/disable memory context in chat
+
+- Add "Memory on/off" toggle inside chat input container
+- Skip automatic RAG retrieval when toggle is off
+- Attached memories still work when toggle is off
+- Toggle defaults to on (current behavior preserved)

app/src/components/ChatInput.tsx

@@ -10,6 +10,7 @@ interface ChatInputProps {
   isLoading?: boolean;
   placeholder?: string;
   className?: string;
+  leftContent?: React.ReactNode;
 }
 
 export function ChatInput({
@@ -19,6 +20,7 @@ export function ChatInput({
   isLoading,
   placeholder = "Type your message...",
   className,
+  leftContent,
 }: ChatInputProps) {
   const inputRef = useRef<HTMLTextAreaElement>(null);
 
@@ -53,7 +55,7 @@ export function ChatInput({
   return (
     <div
       className={cn(
-        "relative flex items-end gap-2 p-2 rounded-2xl",
+        "relative rounded-2xl p-2",
         // Glassmorphism
         "bg-white/70 dark:bg-white/5 backdrop-blur-xl",
         "border border-white/60 dark:border-white/10",
@@ -62,33 +64,40 @@ export function ChatInput({
         className
       )}
     >
-      <textarea
-        ref={inputRef}
-        value={value}
-        onChange={(e) => onChange(e.target.value)}
-        onKeyDown={handleKeyDown}
-        placeholder={placeholder}
-        disabled={isLoading}
-        rows={1}
-        className={cn(
-          "flex-1 bg-transparent px-4 py-2 text-base min-h-[44px] max-h-[200px] resize-none",
-          "placeholder:text-muted-foreground/60",
-          "focus:outline-none",
-          "disabled:opacity-50"
-        )}
-      />
-      <Button
-        size="icon"
-        className="h-10 w-10 rounded-full shrink-0"
-        onClick={onSubmit}
-        disabled={isLoading || !value.trim()}
-      >
-        {isLoading ? (
-          <Loader2 className="h-4 w-4 animate-spin" />
-        ) : (
-          <Send className="h-4 w-4" />
-        )}
-      </Button>
+      <div className="flex items-end gap-2">
+        <textarea
+          ref={inputRef}
+          value={value}
+          onChange={(e) => onChange(e.target.value)}
+          onKeyDown={handleKeyDown}
+          placeholder={placeholder}
+          disabled={isLoading}
+          rows={1}
+          className={cn(
+            "flex-1 bg-transparent px-4 py-2 text-base min-h-[44px] max-h-[200px] resize-none",
+            "placeholder:text-muted-foreground/60",
+            "focus:outline-none",
+            "disabled:opacity-50"
+          )}
+        />
+        <Button
+          size="icon"
+          className="h-10 w-10 rounded-full shrink-0"
+          onClick={onSubmit}
+          disabled={isLoading || !value.trim()}
+        >
+          {isLoading ? (
+            <Loader2 className="h-4 w-4 animate-spin" />
+          ) : (
+            <Send className="h-4 w-4" />
+          )}
+        </Button>
+      </div>
+      {leftContent && (
+        <div className="px-2 pt-1">
+          {leftContent}
+        </div>
+      )}
     </div>
   );
 }

app/src/pages/ChatPage.tsx

@@ -10,14 +10,15 @@ import { AttachedMemoryChips } from "@/components/AttachedMemoryChips";
 import { useConversation } from "@/contexts/ConversationContext";
 import { useConversations } from "@/hooks/useConversations";
 import type { ChatMessage } from "@/types/chat";
-import { FileText, Pin, Trash2 } from "lucide-react";
+import { FileText, Pin, Trash2, Brain } from "lucide-react";
 import { Button } from "@/components/ui/button";
 import { cn } from "@/lib/utils";
 
 export default function ChatPage() {
   const [message, setMessage] = useState("");
   const [isLoading, setIsLoading] = useState(false);
   const [followupSuggestions, setFollowupSuggestions] = useState<string[]>([]);
+  const [useMemories, setUseMemories] = useState(true);
   const isStartingNewChatRef = useRef(false);
   const wantsNewChatRef = useRef(false);
   const [searchParams, setSearchParams] = useSearchParams();
@@ -145,6 +146,7 @@ export default function ChatPage() {
           message: userMessage.content,
           conversation_id: effectiveConversationId,
           attached_memory_ids: memoriesToSend.length > 0 ? memoriesToSend.map((m) => m.id) : undefined,
+          skip_memory_context: !useMemories,
         }),
       });
 
@@ -252,7 +254,7 @@ export default function ChatPage() {
       // Clear pending conversation ref
       pendingConversationRef.current = null;
     }
-  }, [addMessage, updateMessage, setCurrentConversationId, updateContextWindow, attachedMemories, clearAttachedMemories]);
+  }, [addMessage, updateMessage, setCurrentConversationId, updateContextWindow, attachedMemories, clearAttachedMemories, useMemories]);
 
   // Handler for manual chat input
   const handleChat = useCallback(() => {
@@ -381,6 +383,21 @@ export default function ChatPage() {
                       ? `Ask about ${attachedMemories[0].title}...`
                       : "Type your message..."
                   }
+                  leftContent={
+                    <button
+                      onClick={() => setUseMemories(!useMemories)}
+                      className={cn(
+                        "flex items-center gap-1.5 px-2.5 py-1 rounded-full text-xs font-medium transition-colors",
+                        useMemories
+                          ? "bg-muted/50 text-muted-foreground"
+                          : "bg-destructive/10 text-destructive"
+                      )}
+                      title={useMemories ? "Click to disable memory context" : "Click to enable memory context"}
+                    >
+                      <Brain className="h-3 w-3" />
+                      <span>{useMemories ? "Memory on" : "Memory off"}</span>
+                    </button>
+                  }
                 />
               </div>
               <ContextUsageIndicator

backend/app/routes/chat.py

@@ -52,12 +52,12 @@ async def chat_suggestions():
 
 
 async def _retrieve_context(
-    message: str, history: list[dict], attached_memory_ids: list[int] | None = None
+    message: str, history: list[dict], attached_memory_ids: list[int] | None = None, skip_rag: bool = False
 ) -> tuple[str, list[dict]]:
     """Retrieve relevant context and sources using RAG.
 
     If attached_memory_ids are provided, those memories are included as primary context.
-    RAG retrieval still runs to supplement with additional relevant memories.
+    RAG retrieval still runs to supplement with additional relevant memories (unless skip_rag=True).
 
     Returns:
         tuple[str, list[dict]]: (context string, list of source dicts)
@@ -83,8 +83,8 @@ async def _retrieve_context(
             attached_context = "## User's Selected Memory:\n" + format_memories_as_context(attached_memories)
             logger.info(f"Using {len(attached_memories)} attached memories as context")
 
-    # Skip RAG for very short messages (< 10 chars)
-    if len(message.strip()) < 10:
+    # Skip RAG for very short messages (< 10 chars) or when explicitly disabled
+    if len(message.strip()) < 10 or skip_rag:
         return attached_context, sources
 
     try:
@@ -192,7 +192,7 @@ async def chat(request: ChatRequest):
         ]
 
     # --- RAG: Retrieve relevant memories ---
-    context, sources = await _retrieve_context(request.message, history, request.attached_memory_ids)
+    context, sources = await _retrieve_context(request.message, history, request.attached_memory_ids, skip_rag=request.skip_memory_context)
 
     try:
         response = await ai_chat(request.message, context=context, history=history)
@@ -282,14 +282,14 @@ async def chat_stream(request: ChatRequest):
         ]
 
     # RAG: Retrieve relevant memories
-    context, sources = await _retrieve_context(request.message, history, request.attached_memory_ids)
+    context, sources = await _retrieve_context(request.message, history, request.attached_memory_ids, skip_rag=request.skip_memory_context)
 
     async def generate():
         full_response = ""
         usage_data = None
 
         # Send metadata first (conversation_id, sources)
-        yield f"data: {json.dumps({'type': 'meta', 'conversation_id': conversation_id, 'sources': sources, 'searched': True})}\n\n"
+        yield f"data: {json.dumps({'type': 'meta', 'conversation_id': conversation_id, 'sources': sources, 'searched': not request.skip_memory_context})}\n\n"
 
         try:
             async for token, usage in ai_chat_stream(request.message, context=context, history=history):

backend/app/schemas.py

@@ -42,6 +42,7 @@ class ChatRequest(BaseModel):
     conversation_id: int | None = None
     mode: str = "quick"
     attached_memory_ids: list[int] | None = None
+    skip_memory_context: bool = False
 
 
 # Conversation schemas