View as /docs.md
Getting started

Your first request

Everything is OpenAI-compatible. The only change from your existing code is the base_url. Pick a language and run one of the snippets below — if you already have the OpenAI SDK installed, you don't need to install anything else.

Quickstart

python
typescript
aig CLI
curl
# pip install aigateway-py openai
# aigateway-py for primitives (sub-accounts, evals, jobs).
# openai for chat — drop-in, that's the path our own SDK recommends.
from openai import OpenAI

client = OpenAI(
    base_url="https://api.aigateway.sh/v1",
    api_key=os.environ["AIG_KEY"],
)

stream = client.chat.completions.create(
    model="moonshot/kimi-k2.6",
    messages=[{"role": "user",
               "content": "hi"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content, end="")

What you get back

Non-streaming calls return the familiar OpenAI response envelope. Streaming calls return a series of chat.completion.chunk events; the final chunk carries a usage object with exact token counts so you can meter per-request cost without waiting for the provider to settle.

// chat.completion streaming chunks
{ "id": "cmpl-9xT...", "object": "chat.completion.chunk",
  "model": "moonshot/kimi-k2.6",
  "choices": [{ "delta": { "content": "I'll " } }] }
{ ... "choices": [{ "delta": { "content": "help you " } }] }
{ ... "choices": [{ "delta": { "content": "..." } }]  }
{ ... "usage": { "prompt_tokens": 42, "completion_tokens": 184 } }

Request shape

The request body is the standard OpenAI chat/completions schema. Any parameter the target model supports (temperature, top_p, response_format, tool_choice, max_tokens, seed) is passed through untouched. Unsupported parameters are dropped with a warning header (x-aig-warning) rather than a hard error.

POST /v1/chat/completions
Content-Type: application/json
Authorization: Bearer sk-aig-...

{ "model": "moonshot/kimi-k2.6",
  "messages": [
    { "role": "system", "content": "You are concise." },
    { "role": "user", "content": "hi" }
  ],
  "temperature": 0.2,
  "stream": true }

What to try next

← PreviousAuthentication