Welcome

siati.ai is a sovereign LLM platform: OpenAI-compatible API, models served on our own hardware in Lugano, governed by Swiss law (nLPD). No CLOUD Act middleman, no data export, invoices in CHF.

These docs cover everything: integration with the API, RAG over your own documents, infrastructure choices, compliance.

Where to start#

• New here? Jump to the Quickstart — five minutes from sign-up to your first chat completion.
• Migrating from OpenAI? Three lines of code: see OpenAI SDK migration.
• Need RAG over your documents? Walk through the RAG quickstart.
• Building a mobile / chat app? See Chat sessions API — persistent conversations with SSE streaming, auto-title, stop-on-demand.
• Procuring for a regulated industry? Read Sovereignty and nLPD compliance.

What is live right now#

These pages query the production database directly:

• Models catalog — every model currently served, its hardware home, its price tier.
• Infrastructure (fleet) — every GPU/CPU backend, its health, its routing weight.
• Pricing — pay-as-you-go per tier and subscription plans, with the cost formula.

The mental model#

Three concepts you'll use everywhere:

1. Model — what generates the text. Apertus 70B (Swiss), Mistral Large 2 (EU), Qwen 2.5 72B, DeepSeek-R1 Distill, etc. Listed in /v1/models.
2. Tier — how fast / how prioritised you want it. slow / medium / fast / ludicrous. Pay more, get a faster GPU and queue priority.
3. Backend — the physical server running a model. Transparent to you: the BackendRouter picks the best one for (your_tier, your_model).

Conventions in these docs#

• All examples use curl first, then SDK snippets (Python, JavaScript) where relevant.
• Replace $SIATI_API_KEY with your actual key from the dashboard.
• Base URL: https://api.siati.ai for the developer API, https://my.siati.ai/api/v1/ for the mobile/app contract.

Press ⌘K to search anything in these docs.