Skip to content
siema_ai

Introduction

siema_ai runs open-weight language models that are good at Polish, behind the same HTTP interface the OpenAI SDKs already speak. Point your existing OpenAI SDK at siema_ai — the base_url, the api_key, and a model name from Models and pricing — and your chat integration works as is; an embeddings integration needs the embed alias, and any vectors you’ve already stored recomputed against it. You pay up front in PLN, per token, with no subscription.

Po polsku. siema_ai to API do modeli językowych zoptymalizowanych pod język polski (Bielik, Qwen, BGE-M3), zgodne z interfejsem OpenAI. Rozliczasz się w złotych, z góry, za tokeny. Twoje zapytania nie są nigdzie zapisywane, a przetwarzanie odbywa się w Unii Europejskiej. Dokumentacja jest po angielsku; wsparcie odpowiada po polsku.

What you get

Section titled “What you get”
  • Four model aliases. polish-fast and polish-pro (Bielik) for Polish text, advanced (Qwen 27B) for harder reasoning, code and multilingual tasks, embed (BGE-M3) for vectors. See Models and pricing.
  • OpenAI compatibility. POST /v1/chat/completions, POST /v1/embeddings, GET /v1/models. Streaming with server-sent events. Unknown parameters are forwarded to the model unchanged. See OpenAI SDK compatibility.
  • Prepaid PLN credits. Top up with BLIK, card or transfer through PayU. Every request shows its cost in the console within seconds. See Billing.
  • No prompt storage. We never store your prompts or the model’s answers; we keep request metadata (tokens, timing, status, cost). See Data handling.
QuickstartStream your first Polish answer in five minutes.
API referenceEvery endpoint, schema and error, with ready-to-copy code samples.
ConsoleKeys, usage and top-ups.

Status

Section titled “Status”

siema_ai is in its first public release. The changelog lists what changed. Prices and context lengths quoted in these docs are provisional; GET /v1/models is always right.