SDKs

Point the OpenAI, Anthropic and Google SDKs at the AI Gateway by changing the base URL and the key.

Keep the SDK you use. Change two things: its base URL, to the AI Gateway, and its API key, to your virtual key. Each SDK sends the key in its usual header, and Altrum reads it there. Every example reads the key from the ALTRUM_VIRTUAL_KEY environment variable.

OpenAI

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.altrum.ai/openai/v1", api_key=os.environ["ALTRUM_VIRTUAL_KEY"])

reply = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Summarise this ticket in one line."}],
)
print(reply.choices[0].message.content)

import OpenAI from 'openai'

const client = new OpenAI({ baseURL: 'https://api.altrum.ai/openai/v1', apiKey: process.env.ALTRUM_VIRTUAL_KEY })

const reply = await client.chat.completions.create({
  model: 'openai/gpt-4o-mini',
  messages: [{ role: 'user', content: 'Summarise this ticket in one line.' }],
})
console.log(reply.choices[0].message.content)

Anthropic

import os
from anthropic import Anthropic

client = Anthropic(base_url="https://api.altrum.ai/anthropic", api_key=os.environ["ALTRUM_VIRTUAL_KEY"])

reply = client.messages.create(
    model="anthropic/claude-sonnet-4-5",
    max_tokens=256,
    messages=[{"role": "user", "content": "Summarise this ticket in one line."}],
)
print(reply.content[0].text)

Google

import os
from google import genai
from google.genai import types

client = genai.Client(
    api_key=os.environ["ALTRUM_VIRTUAL_KEY"],
    http_options=types.HttpOptions(base_url="https://api.altrum.ai/genai"),
)

reply = client.models.generate_content(
    model="gemini/gemini-3.5-flash",
    contents="Summarise this ticket in one line.",
)
print(reply.text)

Any of these clients can call any model your key allows: the provider prefix in the model name decides where the request goes. The Google SDK cannot list models through Altrum. See the API reference.

Send Altrum headers

To use a saved prompt, steer the cache or set your own request reference, send the request headers with the client or with one request:

Python

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.altrum.ai/v1",
    api_key=os.environ["ALTRUM_VIRTUAL_KEY"],
    default_headers={"x-alt-cache-key": "customer-881"},
)

reply = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Where is my order?"}],
    extra_headers={"x-alt-prompt-id": "support-reply"},
)
print(reply.choices[0].message.content)

Handle a refusal

When a guardrail, a budget or a rate limit stops a request, the SDK raises its usual API error, with the status and a message from Altrum saying what stopped it. Catch it as you catch any provider error. Handle errors shows which errors to retry and has code to copy, and Errors lists every code.