UNST33R API

Connect and list models

Create a key in the console. The SDK base URL is https://api.unst33r.si/v1. Model discovery requires a valid key and does not run inference or charge your balance. Its list reflects your key permissions and available routes.

curl https://api.unst33r.si/v1/models \
  -H "Authorization: Bearer $UNST33R_API_KEY"

Send a chat request

Choose a model returned by /v1/models; replace the example model if your key cannot access it. Inference requires sufficient balance. See pricing and the console for usage and orders.

curl https://api.unst33r.si/v1/chat/completions \
  -H "Authorization: Bearer $UNST33R_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"unst33r-pro","messages":[{"role":"user","content":"Hello"}],"max_completion_tokens":128}'

Python SDK

Install the OpenAI Python package and set UNST33R_API_KEY in your environment.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.unst33r.si/v1",
    api_key=os.environ["UNST33R_API_KEY"],
)
print(client.models.list())
response = client.chat.completions.create(
    model="unst33r-pro",
    messages=[{"role": "user", "content": "Hello"}],
    max_completion_tokens=128,
)
print(response.choices[0].message.content)

Supported request range

• Text Chat Completions. Use one output: omit n or set it to 1.
• Use either max_completion_tokens or max_tokens, never both. The default output limit is 512 tokens; larger limits require model-specific verification and configuration.
• Pro and Flash accept omitted reasoning or low, high, max. DeepSeek also accepts off. Other reasoning values are rejected rather than silently changed.
• DeepSeek currently accepts non-streaming text only. SSE and tool calls are not enabled for it.
• Pro/Flash streaming requires usage reporting; the server adds stream_options.include_usage: true if omitted. Their tool profile is limited to one explicitly selected function with flat string arguments and reasoning_effort: low.
• Sampling controls, structured outputs, images, audio, embeddings and Responses API are outside the supported range. Do not assume every OpenAI SDK option is supported.

Errors and retries

401 means missing or invalid credentials; 403 means access was denied. Unknown API paths return 404 and unsupported methods return 405. Invalid request options return 400. Temporary service or upstream errors may return 429 or 5xx.

An interrupted response does not prove that inference was not billed. Avoid blindly retrying an uncertain request. Check your usage and contact support through the console. A payment return page alone does not confirm payment; check the order status and balance.