Skip to content
System One

ekzhang/openjev-sglang

Prefill-only server that implements the TypeSafe Jev HTTP API on open models, running Qwen3.6-35B-A3B on SGLang. Deploys to Modal on a B200 with radix caching and breakable prefill CUDA graphs, scaling to zero after five idle minutes.

Image from ekzhang/openjev-sglang
Image from i.imgur.com

ekzhang/openjev-sglang on GitHub

Stars
309
Forks
37
Language
Python
Last push
21 Sep 2026

Read from GitHub on . Counts change daily.

How it uses Jev

async def systemone(payload: SystemOneRequest, request: Request):
    async def disconnect():
        while True:
            event = await request.receive()
            if event["type"] == "http.disconnect":
                return

    evaluation = asyncio.create_task(request.app.state.service.evaluate(payload))
    watcher = asyncio.create_task(disconnect())
    try:
        await asyncio.wait([evaluation, watcher], return_when=asyncio.FIRST_COMPLETED)
        if not evaluation.done():
            raise RequestError("Client disconnected", 499)
        result = await evaluation

View in src/openjev/api.py

Open on GitHub

More like this

Keep browsing