Orchestrates SSE line-by-line translation between OpenAI and Anthropic streaming formats, routing events through parser, middleware chain, and serializer. Reuse for building protocol-bridging LLM proxies.
// Source: https://github.com/dean0x/autobeat | License: MIT | Copyright 2024-2026 dean0x
// Provided via justclaw.it.com; keep this notice with the code.
/**
* Stream translator — orchestrates OpenAI parser + middleware + Anthropic serializer.
*
* ARCHITECTURE: The stream translator sits between the raw SSE lines from the
* target (OpenAI) and the SSE output to the source (Claude Code). It processes
* one SSE line at a time, yielding zero or more Anthropic SSE strings.
*
* Line routing:
* - `data: [DONE]` → end sentinel, no-op
* - `data: <json>` → parse JSON, route through target parser + middleware + source serializer
* - `event: <name>` → ignored (OpenAI doesn't use named events)
* - `: <comment>` → ignored
* - Empty line → ignored (SSE event separator)
*/
import type { Str
... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key