Your models. One dependable API.

Run private AI like a shared service.

AI Hub connects your applications to models running on machines you control. Durable runs, isolated tenants, and outbound-only workers without exposing your model host.

Create your hub Read the docs

Self-hosted Rails control plane · OpenAI-compatible API

request.py

import os
from openai import OpenAI

# Point an existing client at your hub
client = OpenAI(
    api_key=os.environ["AI_HUB_TOKEN"],
    base_url="https://your-hub.example/v1",
)

response = client.responses.create(
    model="assistant.general@1",
    input="Summarize this report",
)

response.output_text

Completed by local-worker2.4s

Outbound-only workersDurable SQLite queueTenant isolationSchema validation

Small infrastructure

A control plane, not another platform dependency.

Applications submit work over HTTPS. Workers beside your models make short outbound claims, run compatible jobs, and return validated results.

01ApplicationsSubmit familiar API requests

→

02AI HubAuthenticate, queue, lease

←

03Local workersCall private models

OpenAI-compatible

Use Responses or Chat Completions with your application token and a custom base URL.

Durable by default

Leases, retries, idempotency, and a local result outbox protect work across network interruptions.

Definitions travel

Applications publish immutable instructions and schemas. Generic workers fetch exactly what each job needs.

Operations included

Inspect runs, hosting health, applications, tokens, and task versions from one tenant-scoped console.

Start locally

Connect the model you already run.

One Rails app, one durable volume, and a lightweight worker beside your model.

View integration guide