Models gateway

models.thundera.eu is the one place every Thundera product gets AI models from. Apps call a single API; the gateway routes to the right provider, enforces each key's budget and guardrails, and logs every request.

Why one gateway

• One API, many providers. OpenAI, Google, DeepSeek, Amazon Bedrock and more behind one interface. Swapping a model is a configuration change, not a code change.
• Budgets per key and per user. Each app has its own key; plan limits (for example a plan's monthly AI allowance) become per-user budgets the gateway enforces.
• Guardrails. Allowed models, rate limits and spend caps are set per key.
• Observability. Every request is logged with model, tokens, latency and cost, so usage and spend are visible per app and per user.
• Highly available. At least three replicas, rolled out one at a time.

Using it from an app

Each app authenticates with its own gateway key, stored in SSM at /thundera/<app>/models-api-key and injected as an environment variable. Never ship it to a browser; call the gateway from your backend.

The full API reference — endpoints, model discovery, streaming, errors and budgets — lives in the gateway's own documentation:

Open the models gateway docs