speechcore 0.1.0

Reusable Rust speech-to-text runtime with audio capture, VAD, backend selection, model provisioning, and transcript streaming.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79

use anyhow::Result;
use ndarray::{ArrayD, IxDyn};
use ort::session::{builder::GraphOptimizationLevel, Session};
use ort::value::Tensor;
use std::path::Path;
use std::sync::OnceLock;

static ORT_ENV_INITIALIZED: OnceLock<anyhow::Result<()>> = OnceLock::new();

/// Execution provider preference for ONNX sessions
#[derive(Debug, Clone, Default)]
pub enum ExecutionProviderPreference {
    /// Use CPU only
    #[default]
    CpuOnly,
    /// Prefer GPU (CUDA), fall back to CPU if unavailable
    PreferGpu,
}

#[derive(Debug, Clone)]
pub struct OnnxSessionOptions {
    pub intra_threads: usize,
    pub inter_threads: usize,
    pub execution_provider: ExecutionProviderPreference,
}

impl Default for OnnxSessionOptions {
    fn default() -> Self {
        Self {
            intra_threads: 1,
            inter_threads: 1,
            execution_provider: ExecutionProviderPreference::CpuOnly,
        }
    }
}

pub fn init_ort_environment() -> Result<()> {
    let init_result =
        ORT_ENV_INITIALIZED.get_or_init(|| ort::init().commit().map(|_| ()).map_err(|e| e.into()));
    init_result
        .as_ref()
        .map(|_| ())
        .map_err(|e| anyhow::anyhow!(e.to_string()))
}

pub fn load_session(path: impl AsRef<Path>, options: &OnnxSessionOptions) -> Result<Session> {
    init_ort_environment()?;

    let builder = Session::builder()?
        .with_optimization_level(GraphOptimizationLevel::Level3)?
        .with_intra_threads(options.intra_threads)?
        .with_inter_threads(options.inter_threads)?;

    // Configure execution providers based on preference
    #[cfg(not(feature = "ort-cuda"))]
    let builder = builder;

    #[cfg(feature = "ort-cuda")]
    let builder = {
        if matches!(
            options.execution_provider,
            ExecutionProviderPreference::PreferGpu
        ) {
            use ort::execution_providers::cuda::CUDAExecutionProvider;
            tracing::info!("ONNX session configured with CUDA execution provider");
            builder.with_execution_providers([CUDAExecutionProvider::default().build()])?
        } else {
            builder
        }
    };

    let session = builder.commit_from_file(path)?;
    Ok(session)
}

pub fn tensor_f32_from_slice(shape: &[usize], data: &[f32]) -> Result<Tensor<f32>> {
    let array = ArrayD::from_shape_vec(IxDyn(shape), data.to_vec())?;
    Ok(Tensor::from_array(array)?)
}