Fournex Docs Preview

The backend and runtime services are in active development. This page shows the intended developer experience and API shape.

PreviewAPI surface may changeRuntime-first

Overview

Fournex is a production runtime for reinforcement learning agents. It focuses on stateful execution, deterministic control loops, and system integration. The goal is to make agent deployment feel as simple as running a service.

This is placeholder documentation. Endpoint names, SDK APIs, and deployment paths are subject to change as the backend stabilizes.

Quickstart

Start a session, step the environment, and read actions. The runtime is built around sessions and control loops.

Install (placeholder)

# SDK package name may change
pip install fournex

Start a session

from fournex import Client

client = Client(api_key="YOUR_API_KEY")

session = client.sessions.start(
    policy="agent_abc123",
    environment="CartPole-v1",
    seed=42
)

print(session.id)

Step the control loop

step = client.sessions.step(
    session_id=session.id,
    observation=[0.02, 0.01, -0.03, 0.04],
    reward=0.0,
    done=False
)

print(step.action, step.tick)

Runtime Model

Policy

A deployed RL policy artifact with a stable execution interface.

Session

Long-lived state container for a policy interacting with an environment.

Step

Single environment transition: observation in, action out.

Environment

A simulator, robotics stack, or service providing observations and rewards.

API Reference

POST/v0/sessions

Start a new session for a policy and environment.

POST/v0/sessions/{session_id}/step

Step the environment loop with observation and reward.

POST/v0/policies

GET/v0/sessions/{session_id}

Read session status, tick, and last action.

SDK Usage

The Python SDK will provide a thin wrapper over the runtime APIs. The final interface will stay close to the HTTP model.

SDK client (draft)

from fournex import Client

client = Client(api_key="YOUR_API_KEY", base_url="https://api.fournex.com")
policy = client.policies.register(
    name="my-policy",
    framework="sb3",
    artifact_path="./model.zip"
)

Deployment

Deployment guidance will cover self-hosted runtime nodes, managed clusters, and integration patterns for robotics and simulation.

Want early access or to influence the API surface? Contact the team and share your runtime requirements.