aurum_core/
lib.rs

1//! # aurum-core
2//!
3//! Reusable **on-device speech I/O** library (experimental API).
4//!
5//! - **STT** — local whisper.cpp by default; optional OpenRouter
6//! - **TTS** — local ONNX KittenTTS (cargo feature `tts`, default on)
7//! - **Cleanup** — rules or optional LLM post-edit
8//!
9//! Tagline: *Speech both ways. On-device by default.*
10//!
11//! The API may change without notice until a stable `0.1.0`.
12//!
13//! ## STT example
14//!
15//! ```rust,no_run
16//! use aurum_core::audio::{AudioInput, WHISPER_SAMPLE_RATE};
17//! use aurum_core::pcm::PcmBuffer;
18//! use aurum_core::providers::{LocalWhisperProvider, TranscriptionOptions};
19//! use std::path::PathBuf;
20//!
21//! # async fn demo() -> aurum_core::error::Result<()> {
22//! let provider = LocalWhisperProvider::new(PathBuf::from("/tmp/aurum-cache"))
23//!     .with_progress(false)
24//!     .with_local_only(false);
25//! provider.preload("tiny-q5_1").await?;
26//!
27//! let mut buf = PcmBuffer::dictation();
28//! buf.push(&[0.0f32; 1600])?;
29//! let result = provider
30//!     .transcribe_pcm(
31//!         buf.samples().as_slice(),
32//!         &TranscriptionOptions {
33//!             model: "tiny-q5_1".into(),
34//!             language: "en".into(),
35//!             timestamps: false,
36//!             cancel: None,
37//!         },
38//!     )
39//!     .await?;
40//! let _ = AudioInput::from_pcm_slice(buf.samples().as_slice(), WHISPER_SAMPLE_RATE)?;
41//! println!("{}", result.text);
42//! aurum_core::providers::local::clear_context_cache();
43//! # Ok(())
44//! # }
45//! ```
46
47pub mod audio;
48pub mod bench;
49pub mod cache;
50pub mod cancel;
51pub mod capabilities;
52pub mod cleanup;
53pub mod config;
54pub mod doctor;
55pub mod download;
56pub mod dto;
57pub mod error;
58pub mod eval;
59pub mod model;
60pub mod observability;
61pub mod output;
62pub mod partial;
63pub mod pcm;
64pub mod postprocess;
65pub mod providers;
66pub mod remote;
67pub mod runtime;
68#[cfg(feature = "tts")]
69pub mod tts;
70pub mod window;
71
72pub use audio::{load_audio, AudioInput, WHISPER_SAMPLE_RATE};
73pub use cancel::CancelFlag;
74pub use capabilities::{
75    preflight_cleanup, preflight_stt, preflight_tts, ProviderCapabilities,
76    CAPABILITY_SCHEMA_VERSION,
77};
78pub use cleanup::{
79    apply_cleanup, apply_cleanup_with_segments, cleanup_text, CleanupProviderKind, CleanupReport,
80    CleanupResult, CleanupStyle, OpenRouterCleanup, RulesCleanup, SegmentCleanupPolicy,
81    TextCleanup,
82};
83pub use config::{Config, EffectiveConfigDiagnostic};
84pub use doctor::{run_doctor, DoctorCheck, DoctorReport, DoctorSeverity, DOCTOR_SCHEMA_VERSION};
85pub use dto::{ErrorDto, SttResultDto, ERROR_SCHEMA_VERSION, STT_RESULT_SCHEMA_VERSION};
86pub use error::{AurumError, ErrorCategory, Result, TranscriptionError};
87pub use eval::{
88    build_report, char_error_rate, score_stt, smoke_corpus, word_error_rate, EvalCorpus,
89    EvalReport, SttFixture, SttScore,
90};
91pub use model::{list_models, DownloadProgress, EnsureModelOptions, ModelInfo, ModelStatus};
92pub use observability::{
93    process_metrics, DiagnosticBundle, Metrics, MetricsSnapshot, SpanTimer, METRICS_SCHEMA_VERSION,
94};
95pub use output::{
96    commit_text, format_result, write_result, write_result_to_path, CommitMode, OutputFormat,
97    OutputTransaction, SymlinkPolicy, DEFAULT_MAX_OUTPUT_BYTES,
98};
99pub use partial::{PartialSession, PartialSessionConfig, PartialUpdate};
100pub use pcm::PcmBuffer;
101pub use postprocess::{normalize_result_with_report, NormalizationReport};
102pub use providers::{
103    LocalWhisperProvider, OpenRouterProvider, OpenRouterSttMode, Segment, TranscriptionOptions,
104    TranscriptionProvider, TranscriptionResult,
105};
106pub use remote::{HardenedHttpClient, RemotePolicy};
107pub use runtime::{
108    GovernorConfig, Lifecycle, LifecycleState, OpContext, PermitKind, ResourceGovernor,
109};
110pub use window::{PartialClock, PartialWindowPolicy};
111
112#[cfg(feature = "tts")]
113pub use tts::{
114    format_adapters as format_tts_adapters, format_custom_list as format_tts_custom_list,
115    format_inspect as format_tts_inspect, format_model_list as format_tts_model_list,
116    format_voice_list as format_tts_voice_list, inspect_pack as inspect_tts_pack,
117    list_adapters as list_tts_adapters, list_models as list_tts_models,
118    list_voices as list_tts_voices, propose_add_local as propose_tts_add_local,
119    resolve_voice_for_model, run_kitten_catalogue_conformance, run_pack_conformance,
120    verify_pack as verify_tts_pack, write_add_manifest as write_tts_add_manifest,
121    write_wav_i16_mono_atomic, write_wav_i16_mono_transaction, BackendKind as TtsBackendKind,
122    LocalTtsProvider, SynthesisOptions, SynthesisProvider, SynthesisResult, TrustMode,
123    DEFAULT_TTS_MODEL, DEFAULT_TTS_VOICE, KOKORO_DEFAULT_VOICE, KOKORO_TTS_MODEL,
124};
aurum_core/lib.rs

aurum_core/
lib.rs
