nexaai 0.0.2__tar.gz → 0.0.3__tar.gz

{nexaai-0.0.2 → nexaai-0.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: nexaai
-Version: 0.0.2
+Version: 0.0.3
 Summary: Nexa AI SDK
 Author-Email: Nexa AI <octopus@nexa4ai.com>
 License: MIT

nexaai-0.0.3/nexa/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.0.3"

{nexaai-0.0.2 → nexaai-0.0.3}/nexa/gguf/nexa_inference_text.py

@@ -170,9 +170,11 @@ class NexaTextInference:
                         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:
@@ -180,9 +182,9 @@ class NexaTextInference:
             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"],

{nexaai-0.0.2 → nexaai-0.0.3}/nexa/gguf/nexa_inference_vlm.py

@@ -119,7 +119,7 @@ class NexaVLMInference:
                 )
                 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:

{nexaai-0.0.2 → nexaai-0.0.3}/nexa/gguf/nexa_inference_voice.py

@@ -5,11 +5,15 @@ import sys
 import time
 from pathlib import Path
 
-from nexa.constants import EXIT_REMINDER, NEXA_RUN_MODEL_MAP_VOICE, DEFAULT_VOICE_GEN_PARAMS
+from nexa.constants import (
+    DEFAULT_VOICE_GEN_PARAMS,
+    EXIT_REMINDER,
+    NEXA_RUN_MODEL_MAP_VOICE,
+)
 from nexa.general import pull_model
 from nexa.utils import nexa_prompt
 from faster_whisper import WhisperModel
-from nexaai.utils import nexa_prompt, SpinningCursorAnimation, suppress_stdout_stderr
+from nexa.utils import nexa_prompt, SpinningCursorAnimation, suppress_stdout_stderr
 
 logging.basicConfig(level=logging.INFO)
 
@@ -36,7 +40,7 @@ class NexaVoiceInference:
     def __init__(self, model_path, **kwargs):
         self.model_path = None
         self.downloaded_path = None
-        self.params = DEFAULT_VOICE_GEN_PARAMS        
+        self.params = DEFAULT_VOICE_GEN_PARAMS
         if model_path in NEXA_RUN_MODEL_MAP_VOICE:
             logging.debug(f"Found model {model_path} in public hub")
             self.model_path = NEXA_RUN_MODEL_MAP_VOICE.get(model_path)
@@ -54,7 +58,7 @@ class NexaVoiceInference:
                 exc_info=True,
             )
             exit(1)
-        
+
         self.params.update(kwargs)
         self.model = None
 
@@ -66,15 +70,16 @@ class NexaVoiceInference:
                 )
                 exit(1)
 
-
     @SpinningCursorAnimation()
     def _load_model(self):
+        from faster_whisper import WhisperModel
+
         logging.debug(f"Loading model from: {self.downloaded_path}")
         with suppress_stdout_stderr():
             self.model = WhisperModel(
                 self.downloaded_path,
-                device="cpu", 
-                compute_type=self.params["compute_type"]
+                device="cpu",
+                compute_type=self.params["compute_type"],
             )
         logging.debug("Model loaded successfully")
 
@@ -92,11 +97,11 @@ class NexaVoiceInference:
         logging.debug(f"Transcribing audio from: {audio_path}")
         try:
             segments, _ = self.model.transcribe(
-                audio_path, 
-                beam_size=self.params["beam_size"], 
-                language=self.params["language"], 
-                task=self.params["task"], 
-                temperature=self.params["temperature"], 
+                audio_path,
+                beam_size=self.params["beam_size"],
+                language=self.params["language"],
+                task=self.params["task"],
+                temperature=self.params["temperature"],
                 vad_filter=True,
             )
             transcription = "".join(segment.text for segment in segments)
@@ -105,7 +110,6 @@ class NexaVoiceInference:
         except Exception as e:
             logging.error(f"Error during transcription: {e}", exc_info=True)
 
-
     def _save_transcription(self, transcription):
         os.makedirs(self.params["output_dir"], exist_ok=True)
 

{nexaai-0.0.2 → nexaai-0.0.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "nexaai"
-version = "0.0.2"
+version = "0.0.3"
 description = "Nexa AI SDK"
 readme = "README.md"
 license = { text = "MIT" }

nexaai-0.0.2/nexa/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.0.0.dev"