nexaai 1.0.21rc11__cp310-cp310-win_amd64.whl → 1.0.21rc12__cp310-cp310-win_amd64.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-rc11" 
+__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-1.0.21rc11.dist-info → nexaai-1.0.21rc12.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nexaai
-Version: 1.0.21rc11
+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-1.0.21rc11.dist-info → nexaai-1.0.21rc12.dist-info}/RECORD

@@ -1,7 +1,7 @@
 nexaai/__init__.py,sha256=Yt2YPVcRfwdA6DQoUR4ZI34CHBUYoY2cVsrG7Zp6IrI,2516
-nexaai/_stub.cp310-win_amd64.pyd,sha256=zYnHVcZRM3z1y3douQmEhWPskoy22QHQbu91CooUWII,10752
-nexaai/_version.py,sha256=BEUUhjlv6AJ5LT0jDEF4DBrSaAabJdkGe41TOx3IiqA,148
-nexaai/asr.py,sha256=eeFMFKsaEKiJ2PCGmgedRR4fTfrhaDcnZBuhCChMAio,2148
+nexaai/_stub.cp310-win_amd64.pyd,sha256=UeLgiRBaKGJM0d4yrTux6FbDFrUvfq8y7NiNUGPLo_U,10752
+nexaai/_version.py,sha256=bqNPJMf1z8jxbrr3Um66aksBDh9rjl-aoXigmhCOgog,148
+nexaai/asr.py,sha256=p_bTRqESa-4g6V3HGFzi8vqn8H4Q6yTLiLjL0yqVr3I,2362
 nexaai/base.py,sha256=qQBCiQVNzgpkQjZX9aiFDEdbAAe56TROKC3WnWra2Zg,1021
 nexaai/common.py,sha256=XRwjvRNxS2xRVb0CMR5NbPQ4E-PAewJ0m6VGbZOw19A,3565
 nexaai/cv.py,sha256=PhgqK5hO8FlXLR6QxE_pHtXjdtzLSEPzkCda-qrNpOs,3395
@@ -16,32 +16,33 @@ nexaai/tts.py,sha256=5lEzv50Q5iLTNHLg3kX-4J_mi1T_t-rVdZzOwd6sALY,2271
 nexaai/vlm.py,sha256=uglZ5g0VFizwIYOYjXg-LNrr5baFanTRlAppEkkF-Yw,4929
 nexaai/asr_impl/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nexaai/asr_impl/mlx_asr_impl.py,sha256=XwMX3LYMeulp8cDS0TCCYcjvttFHAyDWQ_oMvABwQmI,3349
-nexaai/asr_impl/pybind_asr_impl.py,sha256=20o5SOPzhF9x41ra8L_qIM7YxCkYeLb5csSrNde-dds,1560
-nexaai/binds/__init__.py,sha256=GA7iQMqrGSH1U4DSGzrjwt0rqw25F_9nSjNNhw0FSto,136
-nexaai/binds/common_bind.cp310-win_amd64.pyd,sha256=DysYAE5xLyUHK4CldEbOda5OubohPC67aYjuB5U-ssk,205824
-nexaai/binds/embedder_bind.cp310-win_amd64.pyd,sha256=_dI1ZWBxQ2n28FNdQ0ouCKKju8edbyEd2H9rVs9JrFc,183808
+nexaai/asr_impl/pybind_asr_impl.py,sha256=I-J_pp4Tdafa-ipp34q2dN4dbL905nlpBYVBGLQwLVI,4807
+nexaai/binds/__init__.py,sha256=cDlzRJF5pq0YmiwmUvdNO9CrvYrBdkYWR4xXrEbwLyU,161
+nexaai/binds/asr_bind.cp310-win_amd64.pyd,sha256=EI5IlfnS6t8o-EASZh8ibD4vZXQy4EHieAmDh56C8-A,192000
+nexaai/binds/common_bind.cp310-win_amd64.pyd,sha256=iMvrdOCoatEyYGCq68-fESc8rymVG9SXOuWZcSmzqnI,205824
+nexaai/binds/embedder_bind.cp310-win_amd64.pyd,sha256=ktHF49Qip1bW4wZIWTW_tE_FpNNPrl2o6U2pkKpQ3W0,183808
 nexaai/binds/libcrypto-3-x64.dll,sha256=X7hDuQPMn6eRe58dTGJxTmRPKFWgd5i6Gjoy9QPLmPg,7315968
 nexaai/binds/libssl-3-x64.dll,sha256=GRw0cOaGUaR2QHRvBFuU_9S9iNgkyVgm0NIpZMnReAw,1313792
-nexaai/binds/llm_bind.cp310-win_amd64.pyd,sha256=3Lzi-W-Ym7UNMEr5jccvnFRU4noMoe2KyiZCILyv8rM,163840
-nexaai/binds/nexa_bridge.dll,sha256=e6AC6AU-NYKB73dWPfBeAFF9jF3cJfp5vJbE4zCjjJE,187904
-nexaai/binds/rerank_bind.cp310-win_amd64.pyd,sha256=1dinTXZQyUJ47ZEecS525T1N1IkAxVjhCHzPGcUYuJs,178176
-nexaai/binds/vlm_bind.cp310-win_amd64.pyd,sha256=2Tx87o4wNuIIZgxjm0jAwSniTuYkbRMBsQjeGaoSpbY,171520
-nexaai/binds/cpu_gpu/ggml-base.dll,sha256=z6-9_e6RL_E1hjgZwkMl3ludgQwqmL792iR5dvDAow8,532480
-nexaai/binds/cpu_gpu/ggml-cpu.dll,sha256=F1A_3KoEGgcZA_-vFiVjYFrB8MetMWqlkGaVEgB0xOc,672768
-nexaai/binds/cpu_gpu/ggml-cuda.dll,sha256=H_ajYEmGvW52GfmOCTDdJgaZVaAoixdysHQARqLXMpA,313528832
-nexaai/binds/cpu_gpu/ggml-vulkan.dll,sha256=w5URKVi4vv9POUvqmnuszvbYISQQPkISs2JInEDMHkQ,36627968
-nexaai/binds/cpu_gpu/ggml.dll,sha256=igSbybRUPftPVXV5qr5kPRT8OhUB5U5FDkuSz-3j9KQ,66560
-nexaai/binds/cpu_gpu/mtmd.dll,sha256=c8FxzwoR73mweoGLMpmORjlVPNiCtbb7Iyy0tSyDHnA,561152
-nexaai/binds/cpu_gpu/nexa_cpu_gpu.dll,sha256=Okp03moKh3DH_7JKokqyy3YilyafVr3nkGFkCux7at4,1611776
-nexaai/binds/cpu_gpu/nexa_plugin.dll,sha256=XiAqUyWLnVxntV1fzSW4uN1CuxqugeWo4A3EBD6GYNA,1424384
-nexaai/binds/nexaml/ggml-base.dll,sha256=z6-9_e6RL_E1hjgZwkMl3ludgQwqmL792iR5dvDAow8,532480
-nexaai/binds/nexaml/ggml-cpu.dll,sha256=F1A_3KoEGgcZA_-vFiVjYFrB8MetMWqlkGaVEgB0xOc,672768
-nexaai/binds/nexaml/ggml-cuda.dll,sha256=H_ajYEmGvW52GfmOCTDdJgaZVaAoixdysHQARqLXMpA,313528832
-nexaai/binds/nexaml/ggml-vulkan.dll,sha256=w5URKVi4vv9POUvqmnuszvbYISQQPkISs2JInEDMHkQ,36627968
-nexaai/binds/nexaml/ggml.dll,sha256=igSbybRUPftPVXV5qr5kPRT8OhUB5U5FDkuSz-3j9KQ,66560
+nexaai/binds/llm_bind.cp310-win_amd64.pyd,sha256=ig_LI15MUpQiSlEmz0AH8fVD9YxIaqtHxJC_5fEV9Vc,163840
+nexaai/binds/nexa_bridge.dll,sha256=2V-T2fLpn8AHJVPkTR425vGVNungIcHpALcoM5a8ifM,187904
+nexaai/binds/rerank_bind.cp310-win_amd64.pyd,sha256=tKPZpVnBJaTsVhXPJryFj41HceWKANlc-mZbHxbvUYQ,178176
+nexaai/binds/vlm_bind.cp310-win_amd64.pyd,sha256=Ah_XGenpplNq_etP0zykp0BIYheISeMAz9sMBrBbNRw,171520
+nexaai/binds/cpu_gpu/ggml-base.dll,sha256=dtjalnYgxV2ymW2gSvz-N2nL9BU3MwlfhJkGSkmfT1U,532480
+nexaai/binds/cpu_gpu/ggml-cpu.dll,sha256=D48mfGUk64IXSvupY0VM_uvSWtEh6T_B-DjItKo7d_c,672768
+nexaai/binds/cpu_gpu/ggml-cuda.dll,sha256=O-l3euXLbdEf43ArdsGkvnUH7J3r1rNwb_0laNL-q_I,313528832
+nexaai/binds/cpu_gpu/ggml-vulkan.dll,sha256=m1xIMYRWLsgQjCDe2emp7KD23K8k5vgFo2cTyBXJdGo,36627968
+nexaai/binds/cpu_gpu/ggml.dll,sha256=aD0G8mXEpawsIq2YiV1QwuNQD2LRhfWzt074x5DY9Yg,66560
+nexaai/binds/cpu_gpu/mtmd.dll,sha256=WSlY1hG8PWeE5qOwXqEKmvnEzc3Zwi3cZRKjwCTiWLY,561152
+nexaai/binds/cpu_gpu/nexa_cpu_gpu.dll,sha256=vXr2Fqe1hnD60j0Q3i0aNXvXMFEhzMTGXGoQbS0xtis,1611776
+nexaai/binds/cpu_gpu/nexa_plugin.dll,sha256=zooi3Iw1fRv4_QhbrYq1U9LOFDYTTqXQZOuVb9K0GJM,1424384
+nexaai/binds/nexaml/ggml-base.dll,sha256=dtjalnYgxV2ymW2gSvz-N2nL9BU3MwlfhJkGSkmfT1U,532480
+nexaai/binds/nexaml/ggml-cpu.dll,sha256=D48mfGUk64IXSvupY0VM_uvSWtEh6T_B-DjItKo7d_c,672768
+nexaai/binds/nexaml/ggml-cuda.dll,sha256=O-l3euXLbdEf43ArdsGkvnUH7J3r1rNwb_0laNL-q_I,313528832
+nexaai/binds/nexaml/ggml-vulkan.dll,sha256=m1xIMYRWLsgQjCDe2emp7KD23K8k5vgFo2cTyBXJdGo,36627968
+nexaai/binds/nexaml/ggml.dll,sha256=aD0G8mXEpawsIq2YiV1QwuNQD2LRhfWzt074x5DY9Yg,66560
 nexaai/binds/nexaml/nexa-mm-process.dll,sha256=DDD5qPbelZAIrRWGggOd3Pg9U_fh_ZEYIplwqiyD5LY,4643328
 nexaai/binds/nexaml/nexa-sampling.dll,sha256=Notkz287laSUG2_ED3oFXWVLJM3t7x_USkvX6wKVOLA,4265984
-nexaai/binds/nexaml/nexa_plugin.dll,sha256=X4uuyL2RXsack-5hCMR7LB7SYHkgrVNAPTEmRpDIeEA,604160
+nexaai/binds/nexaml/nexa_plugin.dll,sha256=5j922WSBy150W9OQndDLkVXgChM3EL1vFMqA16iRHKY,604160
 nexaai/binds/nexaml/nexaproc.dll,sha256=rRcdSfhkxnRwktXW4lp7Np8sNY_uefNXYwAmPp34AHg,2670080
 nexaai/binds/nexaml/qwen3-vl.dll,sha256=k0wPZNR6VejeR32KuFgCVyj7t7YaEmZOKhj9lwMw0-s,5874176
 nexaai/binds/nexaml/qwen3vl-vision.dll,sha256=5AX4MtAOWTcXSAkS2qqvl2pFti4HR55dtkCaOO03Evc,1063424
@@ -72,7 +73,7 @@ nexaai/utils/quantization_utils.py,sha256=FxnZ6-uAE_bl_vQ5jsRXlpU0NBn-U4Y8iN9_O6
 nexaai/vlm_impl/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nexaai/vlm_impl/mlx_vlm_impl.py,sha256=gI09QJcok68QG-36csXr3c2yPCZXTE0dlq1MUBEo_Ms,11091
 nexaai/vlm_impl/pybind_vlm_impl.py,sha256=w8gAZNW2lk3mCKolg4iKgD58RhHCEk4v9aYI1RwVGqc,10021
-nexaai-1.0.21rc11.dist-info/METADATA,sha256=T_LhH1n3qS9DnCW8fwK9jXt5fxVC6WbLjrYByMtwNas,1219
-nexaai-1.0.21rc11.dist-info/WHEEL,sha256=KUuBC6lxAbHCKilKua8R9W_TM71_-9Sg5uEP3uDWcoU,101
-nexaai-1.0.21rc11.dist-info/top_level.txt,sha256=LRE2YERlrZk2vfuygnSzsEeqSknnZbz3Z1MHyNmBU4w,7
-nexaai-1.0.21rc11.dist-info/RECORD,,
+nexaai-1.0.21rc12.dist-info/METADATA,sha256=BdNbXFtWgLQJ9VXQUafcGFUo3u12l1NHiTCNhdIivFU,1219
+nexaai-1.0.21rc12.dist-info/WHEEL,sha256=KUuBC6lxAbHCKilKua8R9W_TM71_-9Sg5uEP3uDWcoU,101
+nexaai-1.0.21rc12.dist-info/top_level.txt,sha256=LRE2YERlrZk2vfuygnSzsEeqSknnZbz3Z1MHyNmBU4w,7
+nexaai-1.0.21rc12.dist-info/RECORD,,