Struct ChunkStreamer 

pub struct ChunkStreamer<'a, C: AudioCodec + ?Sized> { /* private fields */ }

Drives a whole-clip AudioCodec in a chunked / streaming loop: push PCM chunks to get codes, push code chunks to get PCM. Each chunk is processed independently (no carried causal state) — this is exactly how moshi/kyutai drive Mimi today (whole-clip calls on one-frame inputs).

For truly artifact-free, minimal-latency incremental decoding a codec needs internal state carry (causal-conv ring buffers + transformer KV); see rlx-qwen3-tts’s StreamingDecoder for the reference design. This wrapper gives consumers a uniform streaming API over the codecs that only expose whole-clip encode/decode.

Implementations

impl<'a, C: AudioCodec + ?Sized> ChunkStreamer<'a, C>

pub fn new(codec: &'a C, num_quantizers: Option<usize>) -> Self

pub fn with_bitrate(codec: &'a C, target_bps: f32) -> Self

Pick the quantizer count from a target bitrate (bits/sec).

pub fn encode_chunk(&mut self, pcm: &[f32]) -> Result<RvqCodes>

Encode the next PCM chunk to codes.

pub fn decode_chunk(&mut self, codes: &RvqCodes) -> Result<Vec<f32>>

Decode the next code chunk to PCM.

pub fn frames_emitted(&self) -> usize

pub fn samples_emitted(&self) -> usize

pub fn reset(&mut self)