Gateway
The single path between every engine and the AI providers — provider-neutral routing across models and tiers.
In the constellation
Gateway highlighted in the live map — hover or tap any node to explore.
Why it exists
Build anything on AI and you end up wiring provider SDKs straight into your code — model names, API keys, and retry logic hard-coded all over the place. When a provider raises prices, deprecates a model, or goes down, you’re editing source in a dozen files to react.
You also can’t see what any of it costs until the bill arrives, and switching providers means a rewrite. The plumbing between your product and the AI models becomes the most brittle thing you own.
What Gateway is
Gateway is the single path between your software and every AI provider. Your code asks for a quality tier — fast, standard, or top-of-the-line — and Gateway picks the right model behind the scenes, falls back automatically if one is down, and tracks the cost of every call.
Because it speaks the OpenAI-compatible API, anything you’ve already built works without changes. Model names and provider choices live in one config file, never in your application — so an upgrade is a setting, not a code change.
What it does
Tiers, not model names
Ask for the quality you need and Gateway maps it to the best model — swap providers without touching a line of code.
Automatic failover
When a provider degrades or goes dark, requests reroute through a fallback chain with a circuit breaker — your app stays up.
Cost tracking built in
Every call is metered per key and per tier in your own database, so spend is visible in real time, not at invoice time.
One provider-neutral endpoint
Thirteen providers behind a single OpenAI-compatible URL — no SDK sprawl, no vendor lock-in, no rewrite to switch.
The line between judgment and machinery
AccelMars draws one hard line through every product: what an AI decides, and what runs deterministically. Gateway sits on the deterministic side.
Deterministic by design
- ▪ Routing a request to the right provider and tier
- ▪ Fallback, retries, and the circuit breaker
- ▪ Cost metering and request accounting
Gateway routes to the AI — but Gateway itself is pure plumbing. Every routing, failover, and billing decision is deterministic, so the layer your whole stack depends on never has a mind of its own.
How it connects
Live in production — the single AI bridge under every AccelMars engine, running today at scale.
Fully open source (Apache 2.0) — run it yourself, inspect every line, no closed core.