fix: stop transcript leakage and repetition loops in Generic chat template

Backend (inference-server/compat.rs):
- PreparedPrompt now carries extra_stop_sequences injected by template
- Generic template injects User:/System:/Assistant:/Human: stop sequences
  to prevent the model from continuing the transcript after its reply
- Increased repeat_penalty 1.15→1.3 and repeat_last_n 256→512 to
  break repetition loops more aggressively

Backend (inference-server/routes.rs):
- Merge template extra_stop_sequences into sampler before inference

Frontend (chat.ts):
- Added 'System:' and 'Human:' to CHAT_TRANSCRIPT_STOP_SEQUENCES
- Extended TRANSCRIPT_LEAKAGE_PATTERN to include 'Human:'

Frontend (ChatView.tsx):
- buildApiMessages now strips leaked transcript from assistant messages
  before sending history back to the model
- Reduced history window to 20 messages to limit context pollution

tauri-app/src/components/ChatView.tsx

@@ -205,9 +205,16 @@ export default function ChatView({
   const buildApiMessages = useCallback((cid: number) => {
     const history = db.query<{ role: string; content: string }>(
       "SELECT role, content FROM messages WHERE chat_id = ? AND content != '' AND content NOT LIKE '⚠️ Error:%' AND content != 'internal error' ORDER BY created_at DESC LIMIT ?",
-      [cid, 100],
+      [cid, 20],
     ).reverse();
-    return history;
+    // Strip any leaked reasoning/transcript from assistant messages before sending back to model
+    return history.map((msg) => {
+      if (msg.role !== "assistant") return msg;
+      // Remove content after transcript leakage markers
+      const leakageIdx = msg.content.search(/(?:^|\n)\s*(?:System|User|Assistant|Human):/i);
+      const cleaned = leakageIdx > 0 ? msg.content.slice(0, leakageIdx).trimEnd() : msg.content;
+      return { ...msg, content: cleaned };
+    }).filter((msg) => msg.content.trim() !== "");
   }, [db]);
 
   const send = useCallback(async (text: string, fromChatId?: number, replaceAssistantId?: number) => {

tauri-app/src/lib/chat.ts

@@ -7,7 +7,7 @@ export interface ParsedAssistantResponse {
 
 const THINK_OPEN = "<think>";
 const THINK_CLOSE = "</think>";
-const TRANSCRIPT_LEAKAGE_PATTERN = /(?:^|\n)\s*(?:System|User|Assistant):/i;
+const TRANSCRIPT_LEAKAGE_PATTERN = /(?:^|\n)\s*(?:System|User|Assistant|Human):/i;
 
 export const CHAT_TRANSCRIPT_STOP_SEQUENCES = [
   "\nSystem:",
@@ -16,6 +16,9 @@ export const CHAT_TRANSCRIPT_STOP_SEQUENCES = [
   "\n\nSystem:",
   "\n\nUser:",
   "\n\nAssistant:",
+  "System:",
+  "\nHuman:",
+  "\n\nHuman:",
 ];
 
 export function buildChatTitle(text: string): string {