GX10 Queued LLM Gateway
An authenticated API that queues long-running inference jobs against the GX10 box (vLLM), handles model switching, and returns results by polling or webhook callback.
- One bearer API key — send
Authorization: Bearer <key> - Submit a job → poll
/v1/jobs/:idor receive a webhook - Models:
gpt-oss-120b·llama4-scout(32K) ·llama4-scout-long(256K) - Built for big jobs that can run overnight
Full API reference ships with this
service as API.md — drop it into an AI agent and it can drive
the API end-to-end.