nexaai 1.0.21rc10__cp310-cp310-macosx_14_0_universal2.whl → 1.0.21rc12__cp310-cp310-macosx_14_0_universal2.whl

nexaai/_version.py

@@ -1,4 +1,4 @@
 # This file is generated by CMake from _version.py.in
 # Do not modify this file manually - it will be overwritten
 
-__version__ = "1.0.21-rc10" 
+__version__ = "1.0.21-rc12" 

nexaai/asr.py

@@ -3,7 +3,7 @@ from abc import abstractmethod
 from dataclasses import dataclass
 
 from nexaai.base import BaseModel
-from nexaai.common import PluginID
+from nexaai.common import PluginID, ModelConfig
 
 
 @dataclass
@@ -25,15 +25,17 @@ class ASRResult:
 class ASR(BaseModel):
     """Abstract base class for Automatic Speech Recognition models."""
 
-    def __init__(self):
+    def __init__(self, m_cfg: ModelConfig = ModelConfig()):
         """Initialize base ASR class."""
-        pass
+        self._m_cfg = m_cfg
 
     @classmethod
     def _load_from(cls,
                    model_path: str,
+                   model_name: Optional[str] = None,
                    tokenizer_path: Optional[str] = None,
                    language: Optional[str] = None,
+                   m_cfg: ModelConfig = ModelConfig(),
                    plugin_id: Union[PluginID, str] = PluginID.LLAMA_CPP,
                    device_id: Optional[str] = None,
                    **kwargs
@@ -44,10 +46,11 @@ class ASR(BaseModel):
         
         if plugin_value == "mlx":
             from nexaai.asr_impl.mlx_asr_impl import MLXASRImpl
-            return MLXASRImpl._load_from(model_path, tokenizer_path, language, plugin_id, device_id)
+            return MLXASRImpl._load_from(model_path, model_name, tokenizer_path, language, m_cfg, plugin_id, device_id)
         else:
             from nexaai.asr_impl.pybind_asr_impl import PyBindASRImpl
-            return PyBindASRImpl._load_from(model_path, tokenizer_path, language, plugin_id, device_id)
+            return PyBindASRImpl._load_from(model_path, model_name, tokenizer_path, language, m_cfg, plugin_id, device_id)
+
 
     @abstractmethod
     def transcribe(

nexaai/asr_impl/pybind_asr_impl.py

@@ -1,32 +1,78 @@
 from typing import List, Optional, Union
 
-from nexaai.common import PluginID
+from nexaai.common import PluginID, ModelConfig
 from nexaai.asr import ASR, ASRConfig, ASRResult
+from nexaai.binds import asr_bind, common_bind
+from nexaai.runtime import _ensure_runtime
 
 
 class PyBindASRImpl(ASR):
-    def __init__(self):
-        """Initialize PyBind ASR implementation."""
-        super().__init__()
-        # TODO: Add PyBind-specific initialization
+    def __init__(self, handle: any, m_cfg: ModelConfig = ModelConfig()):
+        """Private constructor, should not be called directly."""
+        super().__init__(m_cfg)
+        self._handle = handle  # This is a py::capsule
+        self._model_config = None
 
     @classmethod
     def _load_from(cls,
                    model_path: str,
+                   model_name: Optional[str] = None,
                    tokenizer_path: Optional[str] = None,
                    language: Optional[str] = None,
+                   m_cfg: ModelConfig = ModelConfig(),
                    plugin_id: Union[PluginID, str] = PluginID.LLAMA_CPP,
                    device_id: Optional[str] = None
         ) -> 'PyBindASRImpl':
         """Load ASR model from local path using PyBind backend."""
-        # TODO: Implement PyBind ASR loading
-        instance = cls()
-        return instance
+        _ensure_runtime()
+        
+        # Create model config
+        config = common_bind.ModelConfig()
+        
+        config.n_ctx = m_cfg.n_ctx
+        if m_cfg.n_threads is not None:
+            config.n_threads = m_cfg.n_threads
+        if m_cfg.n_threads_batch is not None:
+            config.n_threads_batch = m_cfg.n_threads_batch
+        if m_cfg.n_batch is not None:
+            config.n_batch = m_cfg.n_batch
+        if m_cfg.n_ubatch is not None:
+            config.n_ubatch = m_cfg.n_ubatch
+        if m_cfg.n_seq_max is not None:
+            config.n_seq_max = m_cfg.n_seq_max
+        config.n_gpu_layers = m_cfg.n_gpu_layers
+        
+        # handle chat template strings
+        if m_cfg.chat_template_path:
+            config.chat_template_path = m_cfg.chat_template_path
+        
+        if m_cfg.chat_template_content:
+            config.chat_template_content = m_cfg.chat_template_content
+        
+        # Convert plugin_id to string
+        plugin_id_str = plugin_id.value if isinstance(plugin_id, PluginID) else str(plugin_id)
+        
+        # Create ASR handle using the binding
+        handle = asr_bind.ml_asr_create(
+            model_path=model_path,
+            model_name=model_name,
+            tokenizer_path=tokenizer_path,
+            model_config=config,
+            language=language,
+            plugin_id=plugin_id_str,
+            device_id=device_id,
+            license_id=None,  # Optional
+            license_key=None  # Optional
+        )
+        
+        return cls(handle, m_cfg)
 
     def eject(self):
-        """Destroy the model and free resources."""
-        # TODO: Implement PyBind ASR cleanup
-        pass
+        """Release the model from memory."""
+        # py::capsule handles cleanup automatically
+        if hasattr(self, '_handle') and self._handle is not None:
+            del self._handle
+            self._handle = None
 
     def transcribe(
         self,
@@ -35,10 +81,47 @@ class PyBindASRImpl(ASR):
         config: Optional[ASRConfig] = None,
     ) -> ASRResult:
         """Transcribe audio file to text."""
-        # TODO: Implement PyBind ASR transcription
-        raise NotImplementedError("PyBind ASR transcription not yet implemented")
+        if self._handle is None:
+            raise RuntimeError("ASR model not loaded. Call _load_from first.")
+        
+        # Convert ASRConfig to binding format if provided
+        asr_config = None
+        if config:
+            asr_config = asr_bind.ASRConfig()
+            asr_config.timestamps = config.timestamps
+            asr_config.beam_size = config.beam_size
+            asr_config.stream = config.stream
+        
+        # Perform transcription using the binding
+        result_dict = asr_bind.ml_asr_transcribe(
+            handle=self._handle,
+            audio_path=audio_path,
+            language=language,
+            config=asr_config
+        )
+        
+        # Convert result to ASRResult
+        transcript = result_dict.get("transcript", "")
+        confidence_scores = result_dict.get("confidence_scores")
+        timestamps = result_dict.get("timestamps")
+        
+        # Convert timestamps to the expected format
+        timestamp_pairs = []
+        if timestamps:
+            for start, end in timestamps:
+                timestamp_pairs.append((float(start), float(end)))
+        
+        return ASRResult(
+            transcript=transcript,
+            confidence_scores=confidence_scores or [],
+            timestamps=timestamp_pairs
+        )
 
     def list_supported_languages(self) -> List[str]:
         """List supported languages."""
-        # TODO: Implement PyBind ASR language listing
-        raise NotImplementedError("PyBind ASR language listing not yet implemented")
+        if self._handle is None:
+            raise RuntimeError("ASR model not loaded. Call _load_from first.")
+        
+        # Get supported languages using the binding
+        languages = asr_bind.ml_asr_list_supported_languages(handle=self._handle)
+        return languages

nexaai/binds/__init__.py

@@ -3,3 +3,4 @@ from .llm_bind import *
 from .embedder_bind import *
 from .vlm_bind import *
 from .rerank_bind import *
+from .asr_bind import *

{nexaai-1.0.21rc10.dist-info → nexaai-1.0.21rc12.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nexaai
-Version: 1.0.21rc10
+Version: 1.0.21rc12
 Summary: Python bindings for NexaSDK C-lib backend
 Author-email: "Nexa AI, Inc." <dev@nexa.ai>
 Project-URL: Homepage, https://github.com/NexaAI/nexasdk-bridge

{nexaai-1.0.21rc10.dist-info → nexaai-1.0.21rc12.dist-info}/RECORD

@@ -1,7 +1,7 @@
 nexaai/__init__.py,sha256=gOd7sNsqEESopw_24xgnOSkIRENrk4Fa-RMtmVv62eA,2421
-nexaai/_stub.cpython-310-darwin.so,sha256=NdbTvDgu2I7P2YYMZofiGIT5QnxZMnCToP2qq_5Vg1s,66768
-nexaai/_version.py,sha256=6snif0L1bnVgpL0nY8W2-vX2Y1ZD7Us0axkHxinLkBw,144
-nexaai/asr.py,sha256=BglyvLM_SiLbQwKhihOuiyX2m1I9-Kq_gYcE7vpjBPY,2083
+nexaai/_stub.cpython-310-darwin.so,sha256=LWfONyxO5innYI-hcs69dkucsuoUQL0vumoCOKwOKVA,66768
+nexaai/_version.py,sha256=KA0_ep8LMLIu4v4BoECwDNI510xSHEzGEsb1IyR4iN8,144
+nexaai/asr.py,sha256=wqtq71cxIMGE4KvOIYZebHdWik8dy4LyKrDI98PDvzQ,2294
 nexaai/base.py,sha256=N8PRgDFA-XPku2vWnQIofQ7ipz3pPlO6f8YZGnuhquE,982
 nexaai/common.py,sha256=MRWZ6a7pnci_OUHxZRm3YqgKLAtZFD7b88STYDfeIF8,3460
 nexaai/cv.py,sha256=gpE3F__6bjh8OQKNJZs-QrBuCxqMj2eH-u6HR90vGZE,3302
@@ -16,11 +16,12 @@ nexaai/tts.py,sha256=jvgDZIyo47NBDny6z74IQT2SDDVo7Mpp-QZwl6YxARU,2196
 nexaai/vlm.py,sha256=LUrd1_SGHOsYpWyUymX93oEIsNJv7XzHIHo4hBZOhQA,4800
 nexaai/asr_impl/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nexaai/asr_impl/mlx_asr_impl.py,sha256=eosd8-TIWAOwV0HltmoFrLwzXHcU4jyxtncvuZE9pgA,3257
-nexaai/asr_impl/pybind_asr_impl.py,sha256=pE9Hb_hMi5yAc4MF83bLVOb8zDtreCkB3_u7XED9YpA,1516
-nexaai/binds/__init__.py,sha256=tI15Ghxgp_PZAdob7op8_K5Ia8DNiTvXiYpFZyB4oN0,131
+nexaai/asr_impl/pybind_asr_impl.py,sha256=FLOWIph37q_nIiNx8xYi-VnhQ6CrPuc4HFAJZQKc42w,4680
+nexaai/binds/__init__.py,sha256=2-Rr0NwyWygqwS8Xlxq0BJ2ltyID-WbGuzEYNlSanCI,155
+nexaai/binds/asr_bind.cpython-310-darwin.so,sha256=nr1BqGYRCooj2k5khaIvGG06_GrzudMXa9xa15UQ-88,200768
 nexaai/binds/common_bind.cpython-310-darwin.so,sha256=zxJuD0nSV--VZKxBfWZUavU7_bHj_JTi0FhkjvG4VJw,235264
 nexaai/binds/embedder_bind.cpython-310-darwin.so,sha256=nn3kINQUNyXIzmcU1olLg8RlkZYyIs_RtIlCk6OH1ds,202064
-nexaai/binds/libnexa_bridge.dylib,sha256=6bdfgd1xnQRJ9p43AdOqjb7OA0Ii1T1xD2wHTs5YMVY,272648
+nexaai/binds/libnexa_bridge.dylib,sha256=M3RoPffXWjlWwz67ZgglEoyTupLIDv1OM833ULNVoyY,272648
 nexaai/binds/llm_bind.cpython-310-darwin.so,sha256=YShsXbe_n2N05joMmTnZXaXh9gM9LGdcmuWUxUc5plI,182784
 nexaai/binds/rerank_bind.cpython-310-darwin.so,sha256=GJmffOLo9A48S_pMG2CtHyhbamtt97QikSSCXk3LqiM,183920
 nexaai/binds/vlm_bind.cpython-310-darwin.so,sha256=JEoQbyOkMXVK4WyBCqKBHbyXHtTiTWT18UkZMFvMh4k,182704
@@ -568,7 +569,7 @@ nexaai/utils/quantization_utils.py,sha256=FYcNSAKGlBqFDUTx3jSKOr2lnq4nyiyC0ZG8oS
 nexaai/vlm_impl/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nexaai/vlm_impl/mlx_vlm_impl.py,sha256=sgHqnX5OCSGLccCnTuRiktIbqThNn3AAIvYE2_Dy4TI,10833
 nexaai/vlm_impl/pybind_vlm_impl.py,sha256=stJKHdhYhBuWUQkky-nHgCv625qDB_1geI3v5BLNGpM,9765
-nexaai-1.0.21rc10.dist-info/METADATA,sha256=H3OlQuxzirdFvlMPT0cP8mSpVQ_pQ7UiOYe4E8OkraA,1188
-nexaai-1.0.21rc10.dist-info/WHEEL,sha256=T2p57lol9__xkoU6aJTyN1Pm43ZpRU3q6km7mIbrAMs,114
-nexaai-1.0.21rc10.dist-info/top_level.txt,sha256=LRE2YERlrZk2vfuygnSzsEeqSknnZbz3Z1MHyNmBU4w,7
-nexaai-1.0.21rc10.dist-info/RECORD,,
+nexaai-1.0.21rc12.dist-info/METADATA,sha256=c7oNlxJI7YWk4xaImnDQFtl4SyC5aVEN2Jlpu6PCAJ8,1188
+nexaai-1.0.21rc12.dist-info/WHEEL,sha256=T2p57lol9__xkoU6aJTyN1Pm43ZpRU3q6km7mIbrAMs,114
+nexaai-1.0.21rc12.dist-info/top_level.txt,sha256=LRE2YERlrZk2vfuygnSzsEeqSknnZbz3Z1MHyNmBU4w,7
+nexaai-1.0.21rc12.dist-info/RECORD,,