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 deliberate major version.
12//!
13//! ## STT example (provider path)
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//!
47//! ## Engine path (preferred for library hosts)
48//!
49//! ```rust,no_run
50//! use aurum_core::AurumEngine;
51//!
52//! # fn demo() -> aurum_core::error::Result<()> {
53//! let engine = AurumEngine::load()?;
54//! let _ = engine.doctor();
55//! let _ = engine.support_bundle(None);
56//! // engine.transcribe_pcm(&samples, &opts).await?; // STT on engine pools
57//! engine.shutdown(); // closes + clears idle models in *this* engine only
58//! # Ok(())
59//! # }
60//! ```
61
62pub mod audio;
63pub mod batch;
64pub mod bench;
65pub mod cache;
66pub mod cancel;
67pub mod capabilities;
68pub mod cleanup;
69pub mod config;
70pub mod doctor;
71pub mod domain;
72pub mod download;
73pub mod dto;
74pub mod engine;
75pub mod error;
76pub mod eval;
77pub mod model;
78pub mod observability;
79pub mod output;
80pub mod partial;
81pub mod pcm;
82pub mod postprocess;
83pub mod profile;
84pub mod providers;
85pub mod remote;
86pub mod runtime;
87pub mod secret;
88pub mod support;
89#[cfg(feature = "tts")]
90pub mod tts;
91pub mod window;
92
93pub use audio::{load_audio, AudioInput, WHISPER_SAMPLE_RATE};
94pub use batch::{
95    build_items, discover_inputs, fingerprint_file, manifest_path, merge_for_resume, work_indices,
96    BatchItem, BatchItemStatus, BatchManifest, BatchSummary, AUDIO_EXTENSIONS, BATCH_MANIFEST_NAME,
97    BATCH_MANIFEST_VERSION,
98};
99pub use cancel::CancelFlag;
100pub use capabilities::{
101    preflight_cleanup, preflight_stt, preflight_tts, ProviderCapabilities,
102    CAPABILITY_SCHEMA_VERSION,
103};
104pub use cleanup::{
105    apply_cleanup, apply_cleanup_with_segments, cleanup_text, CleanupProviderKind, CleanupReport,
106    CleanupResult, CleanupStyle, OpenRouterCleanup, RulesCleanup, SegmentCleanupPolicy,
107    TextCleanup,
108};
109pub use config::{Config, EffectiveConfigDiagnostic, RawConfig, ValidatedConfig};
110pub use doctor::{run_doctor, DoctorCheck, DoctorReport, DoctorSeverity, DOCTOR_SCHEMA_VERSION};
111pub use domain::{FiniteDurationSecs, ModelId, SampleRateHz};
112pub use dto::{ErrorDto, SttResultDto, ERROR_SCHEMA_VERSION, STT_RESULT_SCHEMA_VERSION};
113pub use engine::AurumEngine;
114pub use error::{AurumError, ErrorCategory, Result, TranscriptionError};
115pub use eval::{
116    build_report, char_error_rate, repetition_ratio, score_stt, silence_false_positive,
117    smoke_corpus, word_error_rate, EvalCorpus, EvalReport, SttFixture, SttScore,
118};
119pub use model::{list_models, DownloadProgress, EnsureModelOptions, ModelInfo, ModelStatus};
120pub use observability::{
121    process_metrics, DiagnosticBundle, Metrics, MetricsSnapshot, SpanTimer, METRICS_SCHEMA_VERSION,
122};
123pub use output::{
124    commit_text, format_result, write_result, write_result_to_path, CommitMode, OutputFormat,
125    OutputTransaction, SymlinkPolicy, DEFAULT_MAX_OUTPUT_BYTES,
126};
127pub use partial::{PartialSession, PartialSessionConfig, PartialUpdate};
128pub use pcm::PcmBuffer;
129pub use postprocess::{normalize_result_with_report, NormalizationReport};
130pub use profile::{
131    format_recommendation, resolve_profile, ProfileResolution, QualityProfile,
132    PROFILE_EVIDENCE_VERSION,
133};
134pub use providers::local::{clear_context_cache, process_global_stt_pool, SttContextPool};
135pub use providers::{
136    LocalWhisperProvider, OpenRouterProvider, OpenRouterSttMode, Segment, TranscriptionOptions,
137    TranscriptionProvider, TranscriptionResult,
138};
139pub use remote::{HardenedHttpClient, RemotePolicy};
140pub use runtime::{
141    GovernorConfig, Lifecycle, LifecycleState, OpContext, PermitKind, ResourceGovernor,
142};
143pub use secret::SecretString;
144pub use support::{
145    build_support_bundle, default_bundle_path, SupportBundle, SUPPORT_BUNDLE_VERSION,
146};
147pub use window::{PartialClock, PartialWindowPolicy};
148
149#[cfg(feature = "tts")]
150pub use tts::{
151    format_adapters as format_tts_adapters, format_custom_list as format_tts_custom_list,
152    format_inspect as format_tts_inspect, format_model_list as format_tts_model_list,
153    format_voice_list as format_tts_voice_list, inspect_pack as inspect_tts_pack,
154    list_adapters as list_tts_adapters, list_models as list_tts_models,
155    list_voices as list_tts_voices, process_global_tts_pool,
156    propose_add_local as propose_tts_add_local, resolve_voice_for_model,
157    run_kitten_catalogue_conformance, run_pack_conformance, verify_pack as verify_tts_pack,
158    write_add_manifest as write_tts_add_manifest, write_wav_i16_mono_atomic,
159    write_wav_i16_mono_transaction, BackendKind as TtsBackendKind, LocalTtsProvider,
160    SynthesisOptions, SynthesisProvider, SynthesisResult, TrustMode, TtsSessionPool,
161    DEFAULT_TTS_MODEL, DEFAULT_TTS_VOICE, KOKORO_DEFAULT_VOICE, KOKORO_TTS_MODEL,
162};
aurum_core/lib.rs

aurum_core/
lib.rs
