Skip to main content

Module streaming

Module streaming 

Source
Expand description

Enhanced SSE streaming with delta tokens, finish reasons, and usage info.

This module provides OpenAI-compatible Server-Sent Events (SSE) streaming primitives:

  • StreamChunk / StreamChoice / StreamDelta — wire-format structs that match the OpenAI chat.completion.chunk schema.
  • SseFormatter — stateless helpers that format SSE event strings.
  • TokenStream — a byte-level buffer that accumulates raw token bytes and yields decoded Strings as soon as a valid UTF-8 sequence is complete.
  • StreamStats — throughput accounting for a single generation request.

Structs§

SseFormatter
Stateless SSE event formatter.
StreamChoice
A single choice within a StreamChunk.
StreamChunk
A single SSE streaming chunk (OpenAI-compatible delta format).
StreamDelta
The incremental content delta for one chunk.
StreamStats
Per-request generation throughput statistics.
TokenStream
Byte-level detokenizer buffer with partial-token accumulation.