Struct DiarizationResult 

pub struct DiarizationResult {
    pub segments: Vec<Segment>,
    pub turns: Vec<SpeakerTurn>,
    pub num_speakers: usize,
    pub schema_version: String,
    pub audio: AudioMeta,
    pub provenance: Provenance,
    pub speakers: Vec<SpeakerSummary>,
    pub exclusive_turns: Vec<SpeakerTurn>,
}

Result of offline diarization — the canonical v1 result type that every surface (RTTM/JSON/SRT/VTT/TXT, CLI, MCP) projects from.

The metadata fields (schema_version, audio, provenance, speakers, exclusive_turns) are additive and #[serde(default)], so older JSON without them still deserializes. Construct via DiarizationResult::new so speakers and schema_version are always populated.

Schema family stays diarization-result-v1: new fields are optional and omitted when empty, so consumers that ignore unknown keys remain compatible.

Fields

segments: Vec<Segment>

turns: Vec<SpeakerTurn>

num_speakers: usize

schema_version: String

Schema identifier for downstream consumers.

audio: AudioMeta

Audio metadata (duration, sample rate).

provenance: Provenance

How this result was produced.

speakers: Vec<SpeakerSummary>

Per-speaker rollup (sorted by id); exposes id AND the SPEAKER_NN label.

exclusive_turns: Vec<SpeakerTurn>

Single-speaker (exclusive) timeline derived from Self::turns.

At most one speaker is active at every frame — the ASR-reconciliation surface (overlap is collapsed by a deterministic per-frame argmax). Empty unless filled via DiarizationResult::with_exclusive. The overlap-aware Self::turns field is always the primary diarization output.

Implementations

impl DiarizationResult

pub fn new( segments: Vec<Segment>, turns: Vec<SpeakerTurn>, num_speakers: usize, ) -> Self

Build a v1 result from the core fields, computing the per-speaker speakers rollup from turns and stamping schema_version + provenance.version. Refine audio/provenance with DiarizationResult::with_audio / DiarizationResult::with_provenance.

pub fn with_audio(self, duration_secs: f64, sample_rate: u32) -> Self

Attach audio metadata (builder).

pub fn with_provenance(self, provenance: Provenance) -> Self

Attach producer provenance (builder). Keeps the existing version when the supplied provenance.version is empty.

pub fn with_exclusive(self) -> Self

Fill Self::exclusive_turns from the overlap-aware Self::turns.

Does not modify turns / segments. Safe to call multiple times.

pub fn with_speaker_embeddings( self, embeddings: &[(SpeakerId, Vec<f32>)], ) -> Self

Attach L2-normalized per-speaker embeddings onto Self::speakers.

Speakers with no matching entry keep embedding: None. Vectors are cloned and re-normalized so callers need not pre-normalize.

Trait Implementations

impl Clone for DiarizationResult

fn clone(&self) -> DiarizationResult

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DiarizationResult

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for DiarizationResult

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for DiarizationResult

fn eq(&self, other: &DiarizationResult) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for DiarizationResult

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for DiarizationResult