Getting started

Quickstart

An account, a key, and one changed line. The panel on the right is the request you will end up sending.

Create an account

Sign in with an email address at /login. There is no anonymous generation on this platform — every completion is attributable to an account, including the free ones.

A new account gets a monthly free allowance that renews on the 1st. When it is spent, calls fail with 429 free_allowance_exhausted until it renews or the account is upgraded — see Access & pricing. Build your retry logic accordingly: that one is not retryable.

Create an API key

Go to API keys and create one. The full value is shown once, at creation — only a SHA-256 hash is stored, so a lost key is rotated, never recovered.

Use a ad_test_ key while you are wiring things up and a ad_live_ key when you mean it. Details on Authentication.

Put it in the environment

shell

export AD_API_KEY="ad_live_…"

Never in browser JavaScript, a mobile binary, or a committed file: a key carries the full spending authority of the account that owns it.

Change one base_url

The inference surface is OpenAI-compatible, so the official SDKs work unmodified. Point them at https://adversariallm.ai/v1.

Python · openai

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AD_API_KEY"],
    base_url="https://adversariallm.ai/v1",
)

completion = client.chat.completions.create(
    model="example-model",
    messages=[
        {
            "role": "user",
            "content": "Summarise this advisory in two lines.",
        },
    ],
)
print(completion.choices[0].message.content)

Node · openai

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AD_API_KEY,
  baseURL: "https://adversariallm.ai/v1",
});

const completion = await client.chat.completions.create({
  "model": "example-model",
  "messages": [
    {
      "role": "user",
      "content": "Summarise this advisory in two lines."
    }
  ]
});
console.log(completion.choices[0].message.content);

No SDK? The same call with curl:

curl

curl -sS https://adversariallm.ai/v1/chat/completions \
  -X POST \
  -H "Authorization: Bearer $AD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"example-model","messages":[{"role":"user","content":"Summarise this advisory in two lines."}]}'

Stream it

Set stream: true to get text/event-stream frames terminated by data: [DONE]. Only completion frames are emitted — nothing else is mixed into the stream, so a naive parser cannot trip over one.

Python · streaming

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AD_API_KEY"],
    base_url="https://adversariallm.ai/v1",
)

stream = client.chat.completions.create(
    model="example-model",
    messages=[
        {
            "role": "system",
            "content": "You are terse.",
        },
        {
            "role": "user",
            "content": "Explain CVE triage.",
        },
    ],
    stream_options={
        "include_usage": True,
    },
    max_tokens=512,
    stream=True,
)

for chunk in stream:
    if not chunk.choices:
        continue  # final usage frame
    piece = chunk.choices[0].delta.content
    if piece:
        print(piece, end="", flush=True)