0.x), so there is no dated public release history yet — this page opens
with a single Preview entry describing the current capability set. Dated entries will be added here
as changes ship.
How to read this page
- Entries are newest-first. Each future entry will note the date, the affected surface (gateway API, a specific SDK, or docs), and whether a change is additive or breaking.
/v1changes are additive. Within the/v1data plane, changes are additive — new optional fields and new endpoints — so existing integrations keep working. Anything that would break an integration will be called out explicitly.- SDKs are versioned independently on their own registries:
@roadie/sdkon npm,roadie-sdkon PyPI (beta), and RoadieKit via Swift Package Manager.
Preview — current capabilities
Preview. The platform is in active development ahead of general availability. Interfaces may still
change; where an SDK is beta or a component is not yet publicly distributed, its own reference page
says so.
- Multi-provider gateway. One unified API for chat and embeddings that routes to OpenAI and Anthropic, with model aliases, ordered targets, and cross-provider fallback. See Models, aliases & fallback.
- OpenAI-compatible surface. A drop-in
/openai/v1surface (chat completions, embeddings, images, models) so the stock OpenAI SDKs work with only a base-URL and key change. See Using OpenAI SDKs. - Bring your own keys (BYOK). Your OpenAI and Anthropic credentials are held server-side and never shipped to clients. See Providers & BYOK.
- Client tokens & device attestation. Short-lived, per-end-user tokens let browser and mobile apps call Roadie directly — minted from your backend, from a federated identity with a publishable key, or from an on-device Apple App Attest proof. See Client tokens and Mobile device auth.
- Images. Generation and editing over the OpenAI-compatible image routes, with the native iOS
client RoadieKit. Image output is base64 and
nis capped at 1; see Images. - Usage controls. Per-org, per-project, per-key, and per-end-user rate limits, token/request quotas, and spend budgets. See Rate limits, quotas & budgets and Set usage controls.
- Metering, cost & logs. Every response carries token
usage, an estimatedcost, and anX-Request-Id; usage is metered and requests are logged. See Observability. - Billing & plans. Plan tiers and metered billing underpin the usage controls above. See Billing & plans.
SDKs in this preview
Next steps
Introduction
What Roadie is and how the pieces fit together.
Node quickstart
Send your first chat completion in ten minutes.
TypeScript SDK
The isomorphic first-party SDK reference.
Error reference
Every error type and code, with how to handle each.