Input Audio HTTP

How it fits

Typical backends

Caveats

• Hard constraintClients must send raw audio bytes, not multipart form data. transport fixes whether HTTP, WebSocket, or both channels are open and cannot change at runtime, and max_audio_bytes rejects oversized uploads before decoding.
• I/O contractEach upload gets a fresh session-id emitted on the second output; consumers correlate that id with the AudioFrame chunks on the first output to group frames by upload. The final chunk of an upload may be shorter than frame_size_ms.
• FallbackContainers are identified by sniffing the payload bytes; an unrecognised payload aborts that upload with a decode error and emits zero frames rather than failing the component.
• Parameter interactionWith raw_format empty the container is autodetected, but setting it to a PCM layout makes every upload read as that raw format, which only works together with raw_sample_rate and raw_channels describing the byte stream.
• Parameter interactiondrop_frames discards emitted chunks when the consumer lags instead of blocking the producer, and sample_rate left at the sentinel keeps each upload's native rate while a positive value resamples on the fly.
• State lifetimedrop_frames, frame_size_ms, and sample_rate are hot-swappable; transport, max_audio_bytes, raw_format, raw_sample_rate, and raw_channels bind once at startup.