transcribe-cpp 0.0.10__tar.gz → 0.1.0__tar.gz

{transcribe_cpp-0.0.10 → transcribe_cpp-0.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: transcribe-cpp
-Version: 0.0.10
+Version: 0.1.0
 Summary: Python bindings for transcribe.cpp
 Project-URL: Homepage, https://github.com/handy-computer/transcribe.cpp
 Project-URL: Repository, https://github.com/handy-computer/transcribe.cpp
@@ -21,9 +21,9 @@ Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9
-Requires-Dist: transcribe-cpp-native==0.0.10.*
+Requires-Dist: transcribe-cpp-native==0.1.0.*
 Provides-Extra: cu12
-Requires-Dist: transcribe-cpp-native-cu12==0.0.10.*; extra == 'cu12'
+Requires-Dist: transcribe-cpp-native-cu12==0.1.0.*; extra == 'cu12'
 Provides-Extra: test
 Requires-Dist: numpy; extra == 'test'
 Requires-Dist: pytest>=7; extra == 'test'

{transcribe_cpp-0.0.10 → transcribe_cpp-0.1.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "transcribe-cpp"
-version = "0.0.10"
+version = "0.1.0"
 description = "Python bindings for transcribe.cpp"
 readme = "README.md"
 # 3.8 is EOL (2024-10); 3.9 is the floor. The binding is ctypes-only, so there
@@ -34,13 +34,13 @@ classifiers = [
 # packaging fix still resolves); the import-time version/header-hash check in
 # _library.py is the runtime backstop. check_version_sync.py gates this pin
 # against include/transcribe.h.
-dependencies = ["transcribe-cpp-native==0.0.10.*"]
+dependencies = ["transcribe-cpp-native==0.1.0.*"]
 
 [project.optional-dependencies]
 # Opt-in accelerator providers — ADDITIVE: they install alongside the default
 # provider and the best one wins at runtime. Same base-version pin contract
 # as the hard dependency (gated by check_version_sync.py).
-cu12 = ["transcribe-cpp-native-cu12==0.0.10.*"]
+cu12 = ["transcribe-cpp-native-cu12==0.1.0.*"]
 # Test-only deps. Run with: uv run --extra test pytest (from bindings/python).
 # numpy is here so the numpy PCM-input tests run in every lane instead of
 # silently skipping wherever numpy happens to be absent.

{transcribe_cpp-0.0.10 → transcribe_cpp-0.1.0}/src/transcribe_cpp/__init__.py

@@ -46,7 +46,7 @@ from .errors import (
     raise_for_status,
 )
 
-__version__ = "0.0.10"
+__version__ = "0.1.0"
 
 # String-enum types, exported so callers (and type checkers) can name them.
 Backend = Literal["auto", "cpu", "metal", "vulkan", "cpu_accel", "cuda"]
@@ -474,6 +474,7 @@ class Capabilities:
     supports_streaming: bool
     supports_spec_decode: bool
     max_audio_ms: int
+    translate_target_languages: tuple[str, ...]
 
 
 @dataclass(frozen=True)
@@ -847,6 +848,10 @@ class Model:
         if caps.languages and caps.n_languages > 0:
             for i in range(caps.n_languages):
                 languages.append(_decode(caps.languages[i]))
+        translate_targets = []
+        if caps.translate_target_languages and caps.n_translate_target_languages > 0:
+            for i in range(caps.n_translate_target_languages):
+                translate_targets.append(_decode(caps.translate_target_languages[i]))
         return Capabilities(
             native_sample_rate=caps.native_sample_rate,
             languages=tuple(languages),
@@ -856,6 +861,7 @@ class Model:
             supports_streaming=bool(caps.supports_streaming),
             supports_spec_decode=bool(caps.supports_spec_decode),
             max_audio_ms=caps.max_audio_ms,
+            translate_target_languages=tuple(translate_targets),
         )
 
     def supports(self, feature: Feature) -> bool:

{transcribe_cpp-0.0.10 → transcribe_cpp-0.1.0}/src/transcribe_cpp/_generated.py

@@ -13,7 +13,7 @@ import ctypes as _c
 # Stable digest of the ABI surface below (structs, enums, macros, layout,
 # prototypes). A native provider package echoes this back so the API
 # package can reject an ABI-mismatched provider before dlopen.
-PUBLIC_HEADER_HASH = "ebe6a6816e34a24e"
+PUBLIC_HEADER_HASH = "86b16dd97ad1cb58"
 
 # === enum constants ===
 TRANSCRIBE_OK = 0
@@ -153,7 +153,7 @@ transcribe_backend_device._fields_ = [("struct_size", _c.c_uint64), ("name", _c.
 transcribe_model_load_params._fields_ = [("struct_size", _c.c_uint64), ("backend", _c.c_int), ("gpu_device", _c.c_int)]
 transcribe_session_params._fields_ = [("struct_size", _c.c_uint64), ("n_threads", _c.c_int), ("kv_type", _c.c_int), ("n_ctx", _c.c_int32)]
 transcribe_run_params._fields_ = [("struct_size", _c.c_uint64), ("task", _c.c_int), ("timestamps", _c.c_int), ("pnc", _c.c_int), ("itn", _c.c_int), ("language", _c.c_char_p), ("target_language", _c.c_char_p), ("keep_special_tags", _c.c_bool), ("family", _c.POINTER(transcribe_ext)), ("spec_k_drafts", _c.c_int32)]
-transcribe_capabilities._fields_ = [("struct_size", _c.c_uint64), ("native_sample_rate", _c.c_int32), ("n_languages", _c.c_int), ("languages", _c.POINTER(_c.c_char_p)), ("max_timestamp_kind", _c.c_int), ("supports_language_detect", _c.c_bool), ("supports_translate", _c.c_bool), ("supports_streaming", _c.c_bool), ("supports_spec_decode", _c.c_bool), ("max_audio_ms", _c.c_int64)]
+transcribe_capabilities._fields_ = [("struct_size", _c.c_uint64), ("native_sample_rate", _c.c_int32), ("n_languages", _c.c_int), ("languages", _c.POINTER(_c.c_char_p)), ("max_timestamp_kind", _c.c_int), ("supports_language_detect", _c.c_bool), ("supports_translate", _c.c_bool), ("supports_streaming", _c.c_bool), ("supports_spec_decode", _c.c_bool), ("max_audio_ms", _c.c_int64), ("n_translate_target_languages", _c.c_int), ("translate_target_languages", _c.POINTER(_c.c_char_p))]
 transcribe_session_limits._fields_ = [("struct_size", _c.c_uint64), ("effective_n_ctx", _c.c_int32), ("effective_max_audio_ms", _c.c_int64), ("max_kv_bytes", _c.c_int64)]
 transcribe_stream_params._fields_ = [("struct_size", _c.c_uint64), ("family", _c.POINTER(transcribe_ext)), ("commit_policy", _c.c_int), ("stable_prefix_agreement_n", _c.c_uint32)]
 transcribe_stream_update._fields_ = [("struct_size", _c.c_uint64), ("result_changed", _c.c_bool), ("is_final", _c.c_bool), ("revision", _c.c_int32), ("input_received_ms", _c.c_int64), ("audio_committed_ms", _c.c_int64), ("buffered_ms", _c.c_int64), ("committed_changed", _c.c_bool), ("tentative_changed", _c.c_bool)]
@@ -195,7 +195,7 @@ STRUCT_LAYOUT = {
     'transcribe_model_load_params': {'size': 16, 'align': 8, 'offsets': {'struct_size': 0, 'backend': 8, 'gpu_device': 12}},
     'transcribe_session_params': {'size': 24, 'align': 8, 'offsets': {'struct_size': 0, 'n_threads': 8, 'kv_type': 12, 'n_ctx': 16}},
     'transcribe_run_params': {'size': 64, 'align': 8, 'offsets': {'struct_size': 0, 'task': 8, 'timestamps': 12, 'pnc': 16, 'itn': 20, 'language': 24, 'target_language': 32, 'keep_special_tags': 40, 'family': 48, 'spec_k_drafts': 56}},
-    'transcribe_capabilities': {'size': 40, 'align': 8, 'offsets': {'struct_size': 0, 'native_sample_rate': 8, 'n_languages': 12, 'languages': 16, 'max_timestamp_kind': 24, 'supports_language_detect': 28, 'supports_translate': 29, 'supports_streaming': 30, 'supports_spec_decode': 31, 'max_audio_ms': 32}},
+    'transcribe_capabilities': {'size': 56, 'align': 8, 'offsets': {'struct_size': 0, 'native_sample_rate': 8, 'n_languages': 12, 'languages': 16, 'max_timestamp_kind': 24, 'supports_language_detect': 28, 'supports_translate': 29, 'supports_streaming': 30, 'supports_spec_decode': 31, 'max_audio_ms': 32, 'n_translate_target_languages': 40, 'translate_target_languages': 48}},
     'transcribe_session_limits': {'size': 32, 'align': 8, 'offsets': {'struct_size': 0, 'effective_n_ctx': 8, 'effective_max_audio_ms': 16, 'max_kv_bytes': 24}},
     'transcribe_stream_params': {'size': 24, 'align': 8, 'offsets': {'struct_size': 0, 'family': 8, 'commit_policy': 16, 'stable_prefix_agreement_n': 20}},
     'transcribe_stream_update': {'size': 48, 'align': 8, 'offsets': {'struct_size': 0, 'result_changed': 8, 'is_final': 9, 'revision': 12, 'input_received_ms': 16, 'audio_committed_ms': 24, 'buffered_ms': 32, 'committed_changed': 40, 'tentative_changed': 41}},
@@ -297,6 +297,8 @@ def configure(lib):
     lib.transcribe_model_load_file.argtypes = [_c.c_char_p, _c.POINTER(transcribe_model_load_params), _c.POINTER(_c.c_void_p)]
     lib.transcribe_model_load_params_init.restype = None
     lib.transcribe_model_load_params_init.argtypes = [_c.POINTER(transcribe_model_load_params)]
+    lib.transcribe_model_meta_val_str.restype = _c.c_char_p
+    lib.transcribe_model_meta_val_str.argtypes = [_c.c_void_p, _c.c_char_p]
     lib.transcribe_model_supports.restype = _c.c_bool
     lib.transcribe_model_supports.argtypes = [_c.c_void_p, _c.c_int]
     lib.transcribe_model_variant_string.restype = _c.c_char_p