Skip to main content
Jope Inference Server · Developer Portal

Quick start, REST reference, ZMQ reference.

Pick an entry point. Quick Start walks through a working implementation in 30 minutes; the two reference pages are the full spec for day-to-day lookup.

Quick Start
Python · Walkthrough

Run the server in 30 minutes

Install the stack, paste 50 lines, hit a mock from a client. Swap in the real model afterwards. Code samples are copy-paste-ready.

  • ⏱ ~30 minutes end-to-end
  • 🐍 FastAPI · pyzmq · msgpack
  • 🧪 Hosted mock for wire-level testing
Open walkthrough →
REST · Reference

Management API

Hot-swap active model, start training jobs, probe health. Every endpoint with request / response schemas and a live Try-It panel.

  • :5556/ · HTTP/1.1 · JSON
  • OpenAPI 3.1 · rendered by Scalar
  • Click Try-It to hit the mock live
Open REST reference →
ZMQ · Reference

Inference Channel

One REQ-REP channel carrying predict, ping, and model_info. MessagePack binary · long-lived socket · ≤ 20 ms p95 round-trip.

  • :5555/ · TCP · MessagePack
  • AsyncAPI 2.6 · custom view
  • Per-message payload examples
Open ZMQ reference →
Source specifications

Feed these into openapi-generator / asyncapi generate to scaffold Python and C# clients. The specs, this portal, and the narrative doc at ZMQ Protocol are the single source of truth.