SDKs

The cheapest "SDK" is the one you already use. Gateway-LLM speaks the OpenAI API natively, so the official OpenAI SDKs work as-is — just point them at the gateway:

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.example.com/v1",
    api_key="gw_virt_...",
)

That gets you 90% of what you need. The native SDKs below add ergonomics on top: typed routing decisions, spend records, virtual key management, streaming helpers.

Python — gateway-llm

Install:

pip install gateway-llm

Use:

from gateway_llm import GatewayClient

client = GatewayClient(
    base_url="https://gateway.example.com",
    api_key="gw_virt_...",
)

# Chat completions — same shape as openai-python
resp = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Summarise this email in two lines"}],
)
print(resp.choices[0].message.content)

# Inspect the routing decision
print(resp.gateway.deployment)         # 'openai/gpt-4o-mini'
print(resp.gateway.bucket)             # 'simple'
print(resp.gateway.cost_usd)           # 0.000099

The Python SDK also exposes a LangChain integration:

from gateway_llm.langchain import ChatGatewayLLM

llm = ChatGatewayLLM(model="auto", api_key="gw_virt_...")

TypeScript / JavaScript — @gateway-llm/sdk

Install:

npm install @gateway-llm/sdk

Use:

import { GatewayClient } from '@gateway-llm/sdk';

const client = new GatewayClient({
  baseUrl: 'https://gateway.example.com',
  apiKey: 'gw_virt_...',
});

const resp = await client.chat.completions.create({
  model: 'auto',
  messages: [{ role: 'user', content: 'Hi!' }],
});

// Typed gateway metadata
const { deployment, bucket, costUsd } = resp.gateway;

Streaming uses async iterators (works in Node 18+, Bun, Deno, and modern browsers):

const stream = await client.chat.completions.stream({
  model: 'auto',
  messages: [{ role: 'user', content: 'Tell me a story' }],
});

for await (const chunk of stream) {
  process.stdout.write(chunk.delta);
}

Go — github.com/showmikb/gateway-llm/sdk/go

Install:

go get github.com/showmikb/gateway-llm/sdk/go@latest

Use:

package main

import (
    "context"
    "fmt"

    gateway "github.com/showmikb/gateway-llm/sdk/go"
)

func main() {
    client := gateway.New(gateway.Config{
        BaseURL: "https://gateway.example.com",
        APIKey:  "gw_virt_...",
    })

    resp, err := client.Chat.Completions.Create(context.Background(), &gateway.ChatCompletionRequest{
        Model: "auto",
        Messages: []gateway.Message{{Role: "user", Content: "Hi!"}},
    })
    if err != nil { panic(err) }

    fmt.Println(resp.Choices[0].Message.Content)
    fmt.Println("decision:", resp.Gateway.Deployment)
}

Idiomatic context support, typed errors, and a streaming Recv() interface that doesn't allocate per chunk.

VS Code extension

The Gateway-LLM VS Code extension turns any IDE into a routing-aware playground. Install Gateway-LLM from the Marketplace, run Cmd+Shift+P → Gateway-LLM: Configure, drop in your gateway base URL and a virtual key, and now:

• Hover over any string literal that looks like an LLM prompt to see the cost & bucket the classifier would pick.
• Cmd+K runs the highlighted prompt against your gateway and shows the result inline.
• A status-bar widget displays your team's monthly spend in real time.

It's a thin client over the same HTTP API — works against self-hosted and hosted gateways identically.

Curl / direct HTTP

When you want zero dependencies:

curl https://gateway.example.com/v1/chat/completions \
  -H 'Authorization: Bearer gw_virt_...' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "auto",
    "messages": [{"role":"user","content":"Hi!"}]
  }'

The response includes routing headers (x-gateway-decision, x-gateway-route-bucket, x-gateway-cost-usd) so you can route programmatically without parsing the body.

What's next

• Quickstart — get the server running first if you haven't.
• Virtual API keys — issue the key the SDK is going to use.
• Smart routing — what model: 'auto' actually does.