double the pseudo-streaming chunk size. nearly every reply takes more than 16 tokens.

matrix_pygmalion_bot/ai/koboldcpp.py

@@ -63,7 +63,7 @@ async def generate_sync(
     complete = False
     complete_reply = ""
     for i in range(TIMEOUT//DELAY):
-        input_data["max_length"] = 16 # pseudo streaming
+        input_data["max_length"] = 32 # pseudo streaming
         # Make the request
         try:
             r = requests.post(endpoint, json=input_data, headers=headers, timeout=600)