Module runtime 

The framework-owned per-processor task loop (PROCESSOR-DESIGN §2.2/§2.3/§2.5).

Two channels per processor so System frames jump ahead of Data/Control (pipecat does this with a PriorityQueue, frame_processor.py:119; a second channel is the cheaper, branch-free Rust equivalent):

• bounded normal channel (Data/Control) → natural backpressure for media;
• unbounded system channel (Start/Cancel/Interruption/Error/…) → an interruption can never block on a full queue.

The loop biases the system channel, runs the processor’s lifecycle hooks on lifecycle frames, drains interruptible frames on Frame::Interruption, and converts a process_frame Err into an upstream Frame::Error.

Structs

ProcessorRx

The receive half held by a processor task: a bounded normal channel and an unbounded system channel.

ProcessorTx

The send half: routes a frame by Frame::class (System → the unbounded system channel, Data/Control → the bounded normal channel). Cheaply cloned into each neighbour’s Link.

Constants

NORMAL_CHAN_CAP

Default capacity of the bounded normal (Data/Control) channel — the bench-rs value (~1.3 s of audio headroom @ 50 fps). PROCESSOR-DESIGN §2.2.

Functions

channel

Build a processor channel pair (the bounded normal + unbounded system halves) tagged with the owning processor’s name.

run_processor

The framework-owned per-processor loop. Runs until the input channels close or a terminal frame (End/Cancel) is forwarded.

Type Aliases

EnvelopeSender

Link’s sender alias for a neighbour (PROCESSOR-DESIGN §2.2).