increase n_ctx to 8192

app.py

@@ -92,7 +92,7 @@ def update_llm(size, model_file, clip_file):
     if (model_cache['size'], model_cache['model_file'], model_cache['clip_file']) != (size, model_file, clip_file):
         mf, cf = ensure_weights(size, model_file, clip_file)
         handler = SmolVLM2ChatHandler(clip_model_path=cf, verbose=False)
-        llm = Llama(model_path=mf, chat_handler=handler, n_ctx=1024,
+        llm = Llama(model_path=mf, chat_handler=handler, n_ctx=8192,
                     verbose=False, n_threads=max(2, os.cpu_count()))
         model_cache.update({'size': size, 'model_file': mf, 'clip_file': cf, 'llm': llm})
     return None  # no UI output