Fix infinite text generation

Author: Mug

examples/low_level_api_chatllama_cpp.py

@@ -119,10 +119,9 @@ def generate(self):
 
 					# insert n_left/2 tokens at the start of embd from last_n_tokens
 					_insert = self.last_n_tokens[
-						-(int(n_left/2) - len(self.embd)):-len(self.embd)
+						self.n_ctx - int(n_left/2) - len(self.embd):-len(self.embd)
 					]
-					self.embd[:len(_insert)] = _insert
-					#TODO: Still untested
+					self.embd = _insert + self.embd
 
 				if (llama_cpp.llama_eval(
 					self.ctx, (llama_cpp.llama_token * len(self.embd))(*self.embd), len(self.embd), self.n_past, self.n_threads