Gateway
Coming soonOne API for supported model access.
Connect your application to supported providers. Configure routing and fallbacks, then inspect requests and usage. Preview — not yet deployed.
Model Access
Keep model access in one place.
For developers and product teams building with multiple models, Gateway brings provider access and routing into a shared API surface.
Routing
Make routing deliberate.
- 01
Choose your routes
Configure supported model paths around the task and its requirements.
- 02
Plan fallbacks
Define supported alternatives when a preferred route cannot serve a request. Behavior depends on configuration and availability.
- 03
Inspect traffic
Review request logs, route decisions, and usage signals.
- 04
Bring your own key
Use customer-managed provider keys where supported. Provider terms, access limits, and key responsibilities still apply.
Served routes
Only what the branch serves.
The rows below are the served routes. Everything else is listed as not served, so nothing here reads as a promise. Model prices live with Model Intelligence; the gateway serves no billable set yet.
| Capability | Route | Note | Status |
|---|---|---|---|
| Chat completions | POST v1/chat/completions | Streaming and non-streaming, project-key auth. | Served on branch |
| Models | GET v1/models | The configured model set. | Served on branch |
| Providers | GET v1/providers | Provider inventory behind the routes. | Served on branch |
| Stream control | v1/streams/[id] | Inspect and cancel a running stream. | Served on branch |
| Images | — | Not part of the served set. | Not served |
| Embeddings | — | Not part of the served set. | Not served |
| Speech and transcription | — | Not part of the served set. | Not served |
| Video | — | Not part of the served set. | Not served |
| Realtime sessions | — | Not part of the served set. | Not served |
Workflow
Review a route before expanding it.
The request shapes below come from the branch routes. There is no deployed host yet, so they carry placeholders and this quickstart executes nothing.
- 01
Define the task
Identify the request type and its context needs.
- 02
Configure access
Choose a supported model and provider, then any permitted fallback.
- 03
Test and inspect
Review the result, request status, route, and available usage records.
- 04
See what handled the request.
Request logs and routing records help you inspect traffic. Approval gates and broader workspace evidence depend on the product surface using Gateway.
POST /api/gateway/v1/chat/completions
Authorization: Bearer <project-key>
{ "model": "<configured-model>", "messages": [ ... ] }POST /api/gateway/v1/chat/completions
Authorization: Bearer <project-key>
{ "model": "<configured-model>", "stream": true, "messages": [ ... ] }
# cancel a running stream
POST /api/gateway/v1/streams/<stream-id>/cancelGET /api/gateway/v1/models
Authorization: Bearer <project-key>
GET /api/gateway/v1/providers
Authorization: Bearer <project-key>Route shapes from app/api/gateway/v1 on the unmerged branch, recreated with placeholders. No live host — this executes nothing.
FAQ
Does Gateway guarantee fallback or delivery?
Fallback depends on configured routes and their availability. Provider limits and failures still apply.
Can Gateway route to local models?
Where both the local runtime and Gateway route are configured to support it. Local Models can also be used separately.
Is Gateway the same as Faros?
Gateway provides model access and routing. Faros is Ethen's flagship intelligence; Gateway works without it.
Connect your next build.
Gateway is in preview and not yet deployed. Read the docs while the API is prepared.