Crate flexaudio_core 

flexaudio-core — OS-independent core.

The OS-independent part of the desktop audio capture abstraction library flexaudio. Provides ring buffers, sample-rate conversion, channel mixing, 20 ms chunking, clock normalization, events, and type definitions. OS-specific capture is implemented by separate flexaudio-os-* crates that implement backend::CaptureBackend, which the facade layer wires to the core.

Fixed contract

All internal processing uses interleaved f32 / 48000 Hz / stereo 2ch / 20 ms = 960 frames/chunk. OutputFormat can change the external rate/channels (the Normalizer’s second stage resamples; e.g. 16k/1ch gives 320 frames/chunk). Output chunks remain 20 ms regardless of rate.

The public API has no callbacks. RT threads only push; consumers poll. The RT path never blocks (DROP_OLDEST / overflow drop when full). PTS comes from the device and gaps are detected.

Two-stage ring-buffer layout

[RT cb] --push--> RawRing (rtrb, RT-safe) --pop--> [ingest/process thread]
                                                      |
                                         Normalizer (mix + rubato SRC + 960-frame slicing)
                                                      |
                                                      v
                                      ChunkRing (ringbuf, DROP_OLDEST) --try_pop--> [poll]

Re-exports

pub use capabilities::Availability;

pub use capabilities::BackendKind;

pub use capabilities::Capabilities;

pub use capabilities::DeviceSelectionBehavior;

pub use capabilities::DeviceWatcherCapabilities;

pub use capabilities::OsVersion;

pub use capabilities::PidScope;

pub use capabilities::ProcessCapabilities;

pub use capabilities::SideEffectCapabilities;

pub use capabilities::SideEffectCapability;

pub use capabilities::SideEffectCondition;

pub use capabilities::SourceCapabilities;

pub use capabilities::SourceCapability;

pub use capabilities::SystemExclusionCapabilities;

pub use capabilities::UnavailableReason;

pub use capabilities::WatchScope;

pub use capabilities::WatcherMode;

pub use backend::CaptureBackend;

pub use backend::RawSink;

pub use chunk_ring::chunk_ring;

pub use chunk_ring::ChunkConsumer;

pub use chunk_ring::ChunkProducer;

pub use clock::monotonic_now_ns;

pub use clock::ClockNormalizer;

pub use normalizer::InnerProcessor;

pub use normalizer::NormalizedChunk;

pub use normalizer::Normalizer;

pub use normalizer::CHUNK_FRAMES;

pub use quant::quantize_i16;

pub use raw_ring::raw_ring;

pub use raw_ring::RawConsumer;

pub use raw_ring::RawProducer;

pub use secondary_ring::secondary_chunk_ring;

pub use secondary_ring::SecondaryChunkConsumer;

pub use secondary_ring::SecondaryChunkProducer;

pub use types::AudioChunk;

pub use types::AudioLoss;

pub use types::AudioPath;

pub use types::ChunkFlags;

pub use types::DefaultDeviceKind;

pub use types::DeviceEvent;

pub use types::DeviceInfo;

pub use types::Error;

pub use types::ErrorContext;

pub use types::ErrorGroup;

pub use types::ErrorKind;

pub use types::Event;

pub use types::LossReason;

pub use types::MixLane;

pub use types::NativeStatus;

pub use types::Operation;

pub use types::OutputFormat;

pub use types::OutputTap;

pub use types::Permission;

pub use types::ProcessInfo;

pub use types::Result;

pub use types::SecondaryChunk;

pub use types::ShutdownReport;

pub use types::SourceKind;

pub use types::StreamConfig;

pub use types::CHANNELS;

pub use types::SAMPLE_RATE;

pub use diagnostics::CaptureDiagnostics;

pub use startup::StartupDropReason;

Modules

backend

The CaptureBackend trait implemented by OS backends, and the RawSink handle used by backends to pass raw frames to the core.

capabilities

Capture support and optional side effects, independent of capture permissions.

chunk_ring

SPSC ring for completed AudioChunk values, backed by ringbuf.

clock

Helpers for translating timestamps to process-monotonic i64 nanoseconds.

diagnostics

Allocation-free realtime rejection summaries, materialized off callback.

error_context

Typed failure context and retained shutdown outcomes.

inventory

Shared discovery policy: collection failures are terminal; entry metadata is best effort.

normalizer

Normalize arbitrary device frames (any sample rate/channel count, interleaved f32) in two stages, then convert them back to the requested output format.

process_list

OS-independent post-processing for process enumeration results (ProcessInfo).

quant

Signed 16-bit PCM quantization, shared by every output path.

raw_ring

SPSC, real-time-safe ring for raw interleaved f32 device frames, backed by rtrb.

secondary_ring

SPSC ring of finished SecondaryChunks (ringbuf-backed, DROP_OLDEST).

startup

Sample-based microphone startup confirmation, shared by facade capture paths.

types

Shared types: internal canonical-form constants / AudioChunk / ChunkFlags / SourceKind / OutputFormat / StreamConfig / Event / Error.

Enums

AddonErrorKind

Addon failure identity, independent of native or model diagnostic messages.

ChunkMsOption

Public spelling of the canonical chunk-duration option.

VadErrorCode

Typed cause of an isolated VAD failure.

Functions

chunk_ms_error

One cause/remedy message with the binding’s public option name.

spawn_error

Keep the I/O classification and numeric OS cause without exposing paths or payloads.

validate_chunk_ms

Validate the canonical frame duration before acquiring capture resources.