nemotron3 diarization port

bindings/python/src/transcribe_cpp/__init__.py

@@ -57,6 +57,8 @@
 Itn = Literal["default", "off", "on"]
 Diarize = Literal["default", "off", "on"]
 SortformerPreset = Literal["default", "very_high_latency", "high_latency", "low_latency"]
+Nemotron3DiarPreset = Literal["default", "very_high_latency", "low_latency",
+                              "very_low_latency", "ultra_low_latency"]
 CommitPolicy = Literal["auto", "on_finalize", "stable_prefix"]
 Feature = Literal[
     "initial_prompt", "temperature_fallback", "long_form",
@@ -85,9 +87,12 @@
     "ParakeetStreamOptions",
     "ParakeetBufferedStreamOptions",
     "SortformerStreamOptions",
+    "Nemotron3DiarRunOptions",
+    "Nemotron3DiarStreamOptions",
     "VoxtralRealtimeStreamOptions",
     "Backend",
     "SortformerPreset",
+    "Nemotron3DiarPreset",
     "KVType",
     "Task",
     "Timestamps",
@@ -870,6 +875,57 @@ def _apply(self, ext) -> None:
             ext.preset = self._presets[self.preset]
 
 
+class _Nemotron3DiarPresetOptions(FamilyExtension):
+    """Shared preset mapping for the two Nemotron-3-Diarization slots."""
+
+    _presets = {
+        "default": _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_DEFAULT,
+        "very_high_latency": _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_HIGH_LATENCY,
+        "low_latency": _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_LOW_LATENCY,
+        "very_low_latency": _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_LOW_LATENCY,
+        "ultra_low_latency": _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_ULTRA_LOW_LATENCY,
+    }
+
+    def __init__(self, *, preset: Nemotron3DiarPreset | None = None):
+        if preset is not None and preset not in self._presets:
+            raise ValueError(f"unknown nemotron3_diar preset {preset!r}; "
+                             f"expected one of {sorted(self._presets)}")
+        self.preset = preset
+
+    def _apply(self, ext) -> None:
+        if self.preset is not None:
+            ext.preset = self._presets[self.preset]
+
+
+class Nemotron3DiarRunOptions(_Nemotron3DiarPresetOptions):
+    """Nemotron-3-Diarization operating point for a whole-file run (run slot).
+
+    A diarizer: a run produces speaker segments (up to 8 speakers), no text.
+    ``preset`` picks the model card's latency / accuracy bundle:
+    ``"very_high_latency"`` (30.4 s, the offline point; also ``"default"``),
+    ``"low_latency"`` (1.04 s), ``"very_low_latency"`` (0.64 s),
+    ``"ultra_low_latency"`` (0.32 s). Smaller chunks cost more compute per
+    audio second."""
+
+    _slot = "run"
+    _kind = _generated.TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_RUN
+    _struct = _generated.transcribe_nemotron3_diar_run_ext
+    _init = "transcribe_nemotron3_diar_run_ext_init"
+
+
+class Nemotron3DiarStreamOptions(_Nemotron3DiarPresetOptions):
+    """Nemotron-3-Diarization operating point for push-audio streaming
+    (stream slot). After each feed the speaker segments cover all audio
+    processed so far (a turn still in progress is open-ended and may extend);
+    finalize equals a whole-file run at the same preset. Presets as for
+    :class:`Nemotron3DiarRunOptions`."""
+
+    _slot = "stream"
+    _kind = _generated.TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_STREAM
+    _struct = _generated.transcribe_nemotron3_diar_stream_ext
+    _init = "transcribe_nemotron3_diar_stream_ext_init"
+
+
 # --- high-level handles ---------------------------------------------------
 
 

bindings/python/src/transcribe_cpp/_generated.py

@@ -13,7 +13,7 @@
 # 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 = "7df72bf9e667b8c2"
+PUBLIC_HEADER_HASH = "ae25d09c2b7b325b"
 
 # === enum constants ===
 TRANSCRIBE_OK = 0
@@ -102,6 +102,11 @@
 TRANSCRIBE_STREAM_COMMIT_AUTO = 0
 TRANSCRIBE_STREAM_COMMIT_ON_FINALIZE = 1
 TRANSCRIBE_STREAM_COMMIT_STABLE_PREFIX = 2
+TRANSCRIBE_NEMOTRON3_DIAR_PRESET_DEFAULT = 0
+TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_HIGH_LATENCY = 1
+TRANSCRIBE_NEMOTRON3_DIAR_PRESET_LOW_LATENCY = 2
+TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_LOW_LATENCY = 3
+TRANSCRIBE_NEMOTRON3_DIAR_PRESET_ULTRA_LOW_LATENCY = 4
 TRANSCRIBE_SORTFORMER_PRESET_DEFAULT = 0
 TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY = 1
 TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY = 2
@@ -111,6 +116,8 @@
 
 # === macro constants (integer object-like macros) ===
 TRANSCRIBE_EXT_KIND_MOONSHINE_STREAMING_STREAM = 1414746957
+TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_RUN = 1380201294
+TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_STREAM = 1396978510
 TRANSCRIBE_EXT_KIND_PARAKEET_BUFFERED_STREAM = 1396853584
 TRANSCRIBE_EXT_KIND_PARAKEET_STREAM = 1414744912
 TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM = 1414743635
@@ -150,6 +157,10 @@ class transcribe_speaker_segment(_c.Structure):
     pass
 class transcribe_moonshine_streaming_stream_ext(_c.Structure):
     pass
+class transcribe_nemotron3_diar_run_ext(_c.Structure):
+    pass
+class transcribe_nemotron3_diar_stream_ext(_c.Structure):
+    pass
 class transcribe_parakeet_stream_ext(_c.Structure):
     pass
 class transcribe_parakeet_buffered_stream_ext(_c.Structure):
@@ -179,6 +190,8 @@ class transcribe_whisper_chunk_trace(_c.Structure):
 transcribe_token._fields_ = [("struct_size", _c.c_uint64), ("id", _c.c_int), ("p", _c.c_float), ("t0_ms", _c.c_int64), ("t1_ms", _c.c_int64), ("seg_index", _c.c_int), ("word_index", _c.c_int), ("text", _c.c_char_p)]
 transcribe_speaker_segment._fields_ = [("struct_size", _c.c_uint64), ("t0_ms", _c.c_int64), ("t1_ms", _c.c_int64), ("speaker_id", _c.c_int32), ("p", _c.c_float)]
 transcribe_moonshine_streaming_stream_ext._fields_ = [("ext", transcribe_ext), ("min_decode_interval_ms", _c.c_int32)]
+transcribe_nemotron3_diar_run_ext._fields_ = [("ext", transcribe_ext), ("preset", _c.c_int)]
+transcribe_nemotron3_diar_stream_ext._fields_ = [("ext", transcribe_ext), ("preset", _c.c_int)]
 transcribe_parakeet_stream_ext._fields_ = [("ext", transcribe_ext), ("att_context_right", _c.c_int32)]
 transcribe_parakeet_buffered_stream_ext._fields_ = [("ext", transcribe_ext), ("left_ms", _c.c_int32), ("chunk_ms", _c.c_int32), ("right_ms", _c.c_int32)]
 transcribe_sortformer_stream_ext._fields_ = [("ext", transcribe_ext), ("preset", _c.c_int)]
@@ -224,6 +237,8 @@ class transcribe_whisper_chunk_trace(_c.Structure):
     'transcribe_token': {'size': 48, 'align': 8, 'offsets': {'struct_size': 0, 'id': 8, 'p': 12, 't0_ms': 16, 't1_ms': 24, 'seg_index': 32, 'word_index': 36, 'text': 40}},
     'transcribe_speaker_segment': {'size': 32, 'align': 8, 'offsets': {'struct_size': 0, 't0_ms': 8, 't1_ms': 16, 'speaker_id': 24, 'p': 28}},
     'transcribe_moonshine_streaming_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'min_decode_interval_ms': 16}},
+    'transcribe_nemotron3_diar_run_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'preset': 16}},
+    'transcribe_nemotron3_diar_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'preset': 16}},
     'transcribe_parakeet_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'att_context_right': 16}},
     'transcribe_parakeet_buffered_stream_ext': {'size': 32, 'align': 8, 'offsets': {'ext': 0, 'left_ms': 16, 'chunk_ms': 20, 'right_ms': 24}},
     'transcribe_sortformer_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'preset': 16}},
@@ -343,6 +358,10 @@ def configure(lib):
     lib.transcribe_n_tokens.argtypes = [_c.c_void_p]
     lib.transcribe_n_words.restype = _c.c_int
     lib.transcribe_n_words.argtypes = [_c.c_void_p]
+    lib.transcribe_nemotron3_diar_run_ext_init.restype = None
+    lib.transcribe_nemotron3_diar_run_ext_init.argtypes = [_c.POINTER(transcribe_nemotron3_diar_run_ext)]
+    lib.transcribe_nemotron3_diar_stream_ext_init.restype = None
+    lib.transcribe_nemotron3_diar_stream_ext_init.argtypes = [_c.POINTER(transcribe_nemotron3_diar_stream_ext)]
     lib.transcribe_open.restype = _c.c_int
     lib.transcribe_open.argtypes = [_c.c_char_p, _c.POINTER(transcribe_model_load_params), _c.POINTER(transcribe_session_params), _c.POINTER(_c.c_void_p)]
     lib.transcribe_parakeet_buffered_stream_ext_init.restype = None

bindings/python/tests/test_family_ext.py

@@ -23,6 +23,8 @@
     t.ParakeetStreamOptions,
     t.ParakeetBufferedStreamOptions,
     t.SortformerStreamOptions,
+    t.Nemotron3DiarRunOptions,
+    t.Nemotron3DiarStreamOptions,
     t.VoxtralRealtimeStreamOptions,
 ]
 
@@ -87,6 +89,22 @@ def test_sortformer_unknown_preset_rejected():
         t.SortformerStreamOptions(preset="ultra_low_latency")  # type: ignore[arg-type]
 
 
+@pytest.mark.parametrize("cls,slot,kind", [
+    (t.Nemotron3DiarRunOptions, "run", _generated.TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_RUN),
+    (t.Nemotron3DiarStreamOptions, "stream", _generated.TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_STREAM),
+])
+def test_nemotron3_diar_preset_maps_to_enum_value(cls, slot, kind):
+    assert cls._slot == slot and cls._kind == kind
+    built = cls(preset="ultra_low_latency")._build()
+    assert built.preset == _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_ULTRA_LOW_LATENCY
+    assert cls()._build().preset == _generated.TRANSCRIBE_NEMOTRON3_DIAR_PRESET_DEFAULT
+
+
+def test_nemotron3_diar_unknown_preset_rejected():
+    with pytest.raises(ValueError, match="preset"):
+        t.Nemotron3DiarRunOptions(preset="high_latency")  # type: ignore[arg-type]
+
+
 # --- model-gated: resolve_family validation + a real extension run ----------
 
 

bindings/rust/sys/src/transcribe_sys.rs

@@ -1,15 +1,17 @@
 // @generated by `cargo xtask bindgen` from include/transcribe/extensions.h
 // DO NOT EDIT BY HAND. Regenerate: `cargo xtask bindgen`.
-// Pinned to include/transcribe.abihash = 7df72bf9e667b8c2
+// Pinned to include/transcribe.abihash = ae25d09c2b7b325b
 
 /// The public-ABI digest these bindings were generated against
 /// (sha256/16 over the normalized FFI surface). The load-time version
 /// gate and the CI drift check both anchor on this value.
-pub const PUBLIC_HEADER_HASH: &str = "7df72bf9e667b8c2";
+pub const PUBLIC_HEADER_HASH: &str = "ae25d09c2b7b325b";
 
 /* automatically generated by rust-bindgen 0.72.1 */
 
 pub const TRANSCRIBE_EXT_KIND_MOONSHINE_STREAMING_STREAM: u32 = 1414746957;
+pub const TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_RUN: u32 = 1380201294;
+pub const TRANSCRIBE_EXT_KIND_NEMOTRON3_DIAR_STREAM: u32 = 1396978510;
 pub const TRANSCRIBE_EXT_KIND_PARAKEET_STREAM: u32 = 1414744912;
 pub const TRANSCRIBE_EXT_KIND_PARAKEET_BUFFERED_STREAM: u32 = 1396853584;
 pub const TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM: u32 = 1414743635;
@@ -1115,6 +1117,63 @@ unsafe extern "C" {
         ext: *mut transcribe_moonshine_streaming_stream_ext,
     );
 }
+impl transcribe_nemotron3_diar_preset {
+    pub const TRANSCRIBE_NEMOTRON3_DIAR_PRESET_DEFAULT: transcribe_nemotron3_diar_preset =
+        transcribe_nemotron3_diar_preset(0);
+    pub const TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_HIGH_LATENCY: transcribe_nemotron3_diar_preset =
+        transcribe_nemotron3_diar_preset(1);
+    pub const TRANSCRIBE_NEMOTRON3_DIAR_PRESET_LOW_LATENCY: transcribe_nemotron3_diar_preset =
+        transcribe_nemotron3_diar_preset(2);
+    pub const TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_LOW_LATENCY: transcribe_nemotron3_diar_preset =
+        transcribe_nemotron3_diar_preset(3);
+    pub const TRANSCRIBE_NEMOTRON3_DIAR_PRESET_ULTRA_LOW_LATENCY: transcribe_nemotron3_diar_preset =
+        transcribe_nemotron3_diar_preset(4);
+}
+#[repr(transparent)]
+#[derive(Debug, Copy, Clone, Hash, PartialEq, Eq)]
+pub struct transcribe_nemotron3_diar_preset(pub ::std::os::raw::c_uint);
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct transcribe_nemotron3_diar_run_ext {
+    pub ext: transcribe_ext,
+    pub preset: transcribe_nemotron3_diar_preset,
+}
+#[allow(clippy::unnecessary_operation, clippy::identity_op)]
+const _: () = {
+    ["Size of transcribe_nemotron3_diar_run_ext"]
+        [::std::mem::size_of::<transcribe_nemotron3_diar_run_ext>() - 24usize];
+    ["Alignment of transcribe_nemotron3_diar_run_ext"]
+        [::std::mem::align_of::<transcribe_nemotron3_diar_run_ext>() - 8usize];
+    ["Offset of field: transcribe_nemotron3_diar_run_ext::ext"]
+        [::std::mem::offset_of!(transcribe_nemotron3_diar_run_ext, ext) - 0usize];
+    ["Offset of field: transcribe_nemotron3_diar_run_ext::preset"]
+        [::std::mem::offset_of!(transcribe_nemotron3_diar_run_ext, preset) - 16usize];
+};
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct transcribe_nemotron3_diar_stream_ext {
+    pub ext: transcribe_ext,
+    pub preset: transcribe_nemotron3_diar_preset,
+}
+#[allow(clippy::unnecessary_operation, clippy::identity_op)]
+const _: () = {
+    ["Size of transcribe_nemotron3_diar_stream_ext"]
+        [::std::mem::size_of::<transcribe_nemotron3_diar_stream_ext>() - 24usize];
+    ["Alignment of transcribe_nemotron3_diar_stream_ext"]
+        [::std::mem::align_of::<transcribe_nemotron3_diar_stream_ext>() - 8usize];
+    ["Offset of field: transcribe_nemotron3_diar_stream_ext::ext"]
+        [::std::mem::offset_of!(transcribe_nemotron3_diar_stream_ext, ext) - 0usize];
+    ["Offset of field: transcribe_nemotron3_diar_stream_ext::preset"]
+        [::std::mem::offset_of!(transcribe_nemotron3_diar_stream_ext, preset) - 16usize];
+};
+unsafe extern "C" {
+    pub fn transcribe_nemotron3_diar_run_ext_init(ext: *mut transcribe_nemotron3_diar_run_ext);
+}
+unsafe extern "C" {
+    pub fn transcribe_nemotron3_diar_stream_ext_init(
+        ext: *mut transcribe_nemotron3_diar_stream_ext,
+    );
+}
 #[repr(C)]
 #[derive(Debug, Copy, Clone)]
 pub struct transcribe_parakeet_stream_ext {

bindings/rust/transcribe-cpp/src/family.rs

@@ -103,12 +103,58 @@ pub struct SortformerStreamOptions {
     pub preset: Option<SortformerPreset>,
 }
 
+/// Nemotron-3-Diarization operating point (model-card latency / accuracy
+/// bundle). `Default` keeps the GGUF-shipped configuration (= `VeryHighLatency`).
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+pub enum Nemotron3DiarPreset {
+    #[default]
+    Default,
+    /// 30.4 s (chunk 340 + lookahead 40 frames); the offline point.
+    VeryHighLatency,
+    /// 1.04 s (chunk 9 + lookahead 4).
+    LowLatency,
+    /// 0.64 s (chunk 6 + lookahead 2).
+    VeryLowLatency,
+    /// 0.32 s (chunk 3 + lookahead 1). Most compute per audio second.
+    UltraLowLatency,
+}
+
+impl Nemotron3DiarPreset {
+    fn to_sys(self) -> sys::transcribe_nemotron3_diar_preset {
+        use sys::transcribe_nemotron3_diar_preset as P;
+        match self {
+            Nemotron3DiarPreset::Default => P::TRANSCRIBE_NEMOTRON3_DIAR_PRESET_DEFAULT,
+            Nemotron3DiarPreset::VeryHighLatency => {
+                P::TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_HIGH_LATENCY
+            }
+            Nemotron3DiarPreset::LowLatency => P::TRANSCRIBE_NEMOTRON3_DIAR_PRESET_LOW_LATENCY,
+            Nemotron3DiarPreset::VeryLowLatency => {
+                P::TRANSCRIBE_NEMOTRON3_DIAR_PRESET_VERY_LOW_LATENCY
+            }
+            Nemotron3DiarPreset::UltraLowLatency => {
+                P::TRANSCRIBE_NEMOTRON3_DIAR_PRESET_ULTRA_LOW_LATENCY
+            }
+        }
+    }
+}
+
+/// Nemotron-3-Diarization knobs, used on both slots: as
+/// [`RunExtension::Nemotron3Diar`] for a whole-file run and as
+/// [`StreamExtension::Nemotron3Diar`] for push-audio streaming (segments cover
+/// all audio processed so far; finalize equals a whole-file run). A run
+/// produces speaker segments (up to 8 speakers), no text.
+#[derive(Debug, Clone, Default, PartialEq, Eq)]
+pub struct Nemotron3DiarOptions {
+    pub preset: Option<Nemotron3DiarPreset>,
+}
+
 /// A family extension for the run slot (offline `run`/`run_batch`).
 #[derive(Debug, Clone, PartialEq)]
 #[non_exhaustive]
 pub enum RunExtension {
     Whisper(WhisperRunOptions),
     Sortformer(SortformerStreamOptions),
+    Nemotron3Diar(Nemotron3DiarOptions),
 }
 
 /// A family extension for the stream slot.
@@ -119,6 +165,7 @@ pub enum StreamExtension {
     ParakeetBuffered(ParakeetBufferedStreamOptions),
     MoonshineStreaming(MoonshineStreamingOptions),
     VoxtralRealtime(VoxtralRealtimeStreamOptions),
+    Nemotron3Diar(Nemotron3DiarOptions),
 }
 
 /// Owns a materialized run-slot C extension struct (and any strings it points
@@ -131,6 +178,7 @@ pub(crate) enum RunExtRaw {
         _prompt: Option<CString>,
     },
     Sortformer(Box<sys::transcribe_sortformer_stream_ext>),
+    Nemotron3Diar(Box<sys::transcribe_nemotron3_diar_run_ext>),
 }
 
 impl RunExtRaw {
@@ -143,6 +191,10 @@ impl RunExtRaw {
             RunExtRaw::Sortformer(e) => {
                 (&**e) as *const sys::transcribe_sortformer_stream_ext as *const sys::transcribe_ext
             }
+            RunExtRaw::Nemotron3Diar(e) => {
+                (&**e) as *const sys::transcribe_nemotron3_diar_run_ext
+                    as *const sys::transcribe_ext
+            }
         }
     }
 }
@@ -182,6 +234,12 @@ impl RunExtension {
                 set(&mut ext.preset, o.preset.map(SortformerPreset::to_sys));
                 Ok(RunExtRaw::Sortformer(Box::new(ext)))
             }
+            RunExtension::Nemotron3Diar(o) => {
+                let mut ext: sys::transcribe_nemotron3_diar_run_ext = unsafe { std::mem::zeroed() };
+                unsafe { sys::transcribe_nemotron3_diar_run_ext_init(&mut ext) };
+                set(&mut ext.preset, o.preset.map(Nemotron3DiarPreset::to_sys));
+                Ok(RunExtRaw::Nemotron3Diar(Box::new(ext)))
+            }
         }
     }
 }
@@ -192,6 +250,7 @@ pub(crate) enum StreamExtRaw {
     ParakeetBuffered(Box<sys::transcribe_parakeet_buffered_stream_ext>),
     MoonshineStreaming(Box<sys::transcribe_moonshine_streaming_stream_ext>),
     VoxtralRealtime(Box<sys::transcribe_voxtral_realtime_stream_ext>),
+    Nemotron3Diar(Box<sys::transcribe_nemotron3_diar_stream_ext>),
 }
 
 impl StreamExtRaw {
@@ -212,6 +271,10 @@ impl StreamExtRaw {
                 (&**e) as *const sys::transcribe_voxtral_realtime_stream_ext
                     as *const sys::transcribe_ext
             }
+            StreamExtRaw::Nemotron3Diar(e) => {
+                (&**e) as *const sys::transcribe_nemotron3_diar_stream_ext
+                    as *const sys::transcribe_ext
+            }
         }
     }
 }
@@ -249,6 +312,13 @@ impl StreamExtension {
                 set(&mut e.min_decode_interval_ms, o.min_decode_interval_ms);
                 StreamExtRaw::VoxtralRealtime(Box::new(e))
             }
+            StreamExtension::Nemotron3Diar(o) => {
+                let mut e: sys::transcribe_nemotron3_diar_stream_ext =
+                    unsafe { std::mem::zeroed() };
+                unsafe { sys::transcribe_nemotron3_diar_stream_ext_init(&mut e) };
+                set(&mut e.preset, o.preset.map(Nemotron3DiarPreset::to_sys));
+                StreamExtRaw::Nemotron3Diar(Box::new(e))
+            }
         }
     }
 }