Merge pull request #21 from NexaAI/david/bugfix

David/bugfix

nexa/gguf/nexa_inference_text.py

@@ -170,19 +170,21 @@ def run(self):
                         generated_text += delta
 
                 if self.chat_format:
-                    self.conversation_history.append(
-                        {"role": "assistant", "content": generated_text}
-                    )
+                    if len(self.conversation_history) >= 2:
+                        self.conversation_history = self.conversation_history[2:]
+
+                    self.conversation_history.append({"role": "user", "content": user_input})
+                    self.conversation_history.append({"role": "assistant", "content": generated_text})
             except KeyboardInterrupt:
                 pass
             except Exception as e:
                 logging.error(f"Error during generation: {e}", exc_info=True)
             print("\n")
 
     def _chat(self, user_input: str) -> Iterator:
-        self.conversation_history.append({"role": "user", "content": user_input})
+        current_messages = self.conversation_history + [{"role": "user", "content": user_input}]
         return self.model.create_chat_completion(
-            messages=self.conversation_history,
+            messages=current_messages,
             temperature=self.params["temperature"],
             max_tokens=self.params["max_new_tokens"],
             top_k=self.params["top_k"],

nexa/gguf/nexa_inference_vlm.py

@@ -119,7 +119,7 @@ def __init__(self, model_path, stop_words=None, **kwargs):
                 )
                 exit(1)
         else:
-            logging.error("Using model from hub is not supported yet.")
+            logging.error("VLM user model from hub is not supported yet.")
             exit(1)
 
         if self.downloaded_path is None:

nexa/gguf/nexa_inference_voice.py

@@ -11,7 +11,9 @@
     NEXA_RUN_MODEL_MAP_VOICE,
 )
 from nexa.general import pull_model
-from nexa.utils import SpinningCursorAnimation, nexa_prompt, suppress_stdout_stderr
+from nexa.utils import nexa_prompt
+from faster_whisper import WhisperModel
+from nexa.utils import nexa_prompt, SpinningCursorAnimation, suppress_stdout_stderr
 
 logging.basicConfig(level=logging.INFO)