fix: throttle streaming renders via rAF batching

- Batch Svelte reactivity triggers during streaming to ~60fps using
  requestAnimationFrame, reducing rendering overhead for fast models
- Throttle TTS content parsing to every 150ms instead of every token
- Remove per-token console.log, await tick(), and scrollToBottom calls
- Keep immediate rendering for done events and non-streaming events

Author: Algorithm5838

src/lib/components/chat/Chat.svelte

@@ -154,6 +154,10 @@
 
 	let taskIds = null;
 
+	// Streaming render state
+	let renderRafId = null;
+	let lastTtsProcessTime = 0;
+
 	// Chat Input
 	let prompt = '';
 	let chatFiles = [];
@@ -382,8 +386,6 @@
 	};
 
 	const chatEventHandler = async (event, cb) => {
-		console.log(event);
-
 		if (event.chat_id === $chatId) {
 			await tick();
 			let message = history.messages[event.message_id];
@@ -508,7 +510,24 @@
 					console.log('Unknown message type', data);
 				}
 
-				history.messages[event.message_id] = message;
+				// Batch reactivity updates during streaming to reduce rendering overhead.
+				// During active streaming, content changes accumulate on the message reference
+				// and a single render is scheduled per animation frame (~60fps).
+				// Non-streaming events (done, status changes, etc.) render immediately.
+				if (type === 'chat:completion' && !data?.done) {
+					if (!renderRafId) {
+						const msgId = event.message_id;
+						renderRafId = requestAnimationFrame(() => {
+							renderRafId = null;
+							history.messages[msgId] = history.messages[msgId];
+							if (autoScroll) {
+								scrollToBottom();
+							}
+						});
+					}
+				} else {
+					history.messages[event.message_id] = message;
+				}
 			}
 		}
 	};
@@ -1479,27 +1498,31 @@
 						navigator.vibrate(5);
 					}
 
-					// Emit chat event for TTS
-					const messageContentParts = getMessageContentParts(
-						removeAllDetails(message.content),
-						$config?.audio?.tts?.split_on ?? 'punctuation'
-					);
-					messageContentParts.pop();
+					// Throttle TTS processing to avoid re-parsing content on every token
+					const now = Date.now();
+					if (now - lastTtsProcessTime >= 150) {
+						lastTtsProcessTime = now;
 
-					// dispatch only last sentence and make sure it hasn't been dispatched before
-					if (
-						messageContentParts.length > 0 &&
-						messageContentParts[messageContentParts.length - 1] !== message.lastSentence
-					) {
-						message.lastSentence = messageContentParts[messageContentParts.length - 1];
-						eventTarget.dispatchEvent(
-							new CustomEvent('chat', {
-								detail: {
-									id: message.id,
-									content: messageContentParts[messageContentParts.length - 1]
-								}
-							})
+						const messageContentParts = getMessageContentParts(
+							removeAllDetails(message.content),
+							$config?.audio?.tts?.split_on ?? 'punctuation'
 						);
+						messageContentParts.pop();
+
+						if (
+							messageContentParts.length > 0 &&
+							messageContentParts[messageContentParts.length - 1] !== message.lastSentence
+						) {
+							message.lastSentence = messageContentParts[messageContentParts.length - 1];
+							eventTarget.dispatchEvent(
+								new CustomEvent('chat', {
+									detail: {
+										id: message.id,
+										content: messageContentParts[messageContentParts.length - 1]
+									}
+								})
+							);
+						}
 					}
 				}
 			}
@@ -1513,27 +1536,31 @@
 				navigator.vibrate(5);
 			}
 
-			// Emit chat event for TTS
-			const messageContentParts = getMessageContentParts(
-				removeAllDetails(message.content),
-				$config?.audio?.tts?.split_on ?? 'punctuation'
-			);
-			messageContentParts.pop();
+			// Throttle TTS processing to avoid re-parsing content on every token
+			const now = Date.now();
+			if (now - lastTtsProcessTime >= 150) {
+				lastTtsProcessTime = now;
 
-			// dispatch only last sentence and make sure it hasn't been dispatched before
-			if (
-				messageContentParts.length > 0 &&
-				messageContentParts[messageContentParts.length - 1] !== message.lastSentence
-			) {
-				message.lastSentence = messageContentParts[messageContentParts.length - 1];
-				eventTarget.dispatchEvent(
-					new CustomEvent('chat', {
-						detail: {
-							id: message.id,
-							content: messageContentParts[messageContentParts.length - 1]
-						}
-					})
+				const messageContentParts = getMessageContentParts(
+					removeAllDetails(message.content),
+					$config?.audio?.tts?.split_on ?? 'punctuation'
 				);
+				messageContentParts.pop();
+
+				if (
+					messageContentParts.length > 0 &&
+					messageContentParts[messageContentParts.length - 1] !== message.lastSentence
+				) {
+					message.lastSentence = messageContentParts[messageContentParts.length - 1];
+					eventTarget.dispatchEvent(
+						new CustomEvent('chat', {
+							detail: {
+								id: message.id,
+								content: messageContentParts[messageContentParts.length - 1]
+							}
+						})
+					);
+				}
 			}
 		}
 
@@ -1546,8 +1573,6 @@
 			message.usage = usage;
 		}
 
-		history.messages[message.id] = message;
-
 		if (done) {
 			message.done = true;
 
@@ -1596,13 +1621,6 @@
 				createMessagesList(history, message.id)
 			);
 		}
-
-		console.log(data);
-		await tick();
-
-		if (autoScroll) {
-			scrollToBottom();
-		}
 	};
 
 	//////////////////////////