SDK Wrappers

Get started in 30 seconds. One import change. Zero config.

Overview

Stockyard SDK wrappers are thin drop-in replacements for official LLM client libraries. Change one import line and all traffic automatically routes through your Stockyard proxy — no base URL configuration, no HTTP client swapping, no middleware setup.

Python

pip install stockyard-openai

Node

npm install @stockyard/openai

go get github.com/stockyard-dev/stockyard-openai-go

Python

Install

pip install stockyard-openai

Before (vanilla OpenAI)

from openai import OpenAI
client = OpenAI(api_key="sk-...")

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)

After (with Stockyard)

from stockyard_openai import OpenAI
client = OpenAI(api_key="sk-...")

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)

One line changed. Everything else stays the same. The wrapper automatically routes through http://localhost:4200/v1 (or your STOCKYARD_URL).

Node

Install

npm install @stockyard/openai

Before (vanilla OpenAI)

import OpenAI from 'openai';
const client = new OpenAI();

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello' }],
});

After (with Stockyard)

import OpenAI from '@stockyard/openai';
const client = new OpenAI();

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello' }],
});

Go

Install

go get github.com/stockyard-dev/stockyard-openai-go

Before (vanilla OpenAI)

import "github.com/openai/openai-go"

client := openai.NewClient()

After (with Stockyard)

import "github.com/openai/openai-go"
import "github.com/stockyard-dev/stockyard-openai-go"

client := openai.NewClient(
    option.WithBaseURL(stockyard.BaseURL()),
    option.WithHTTPClient(&http.Client{Transport: stockyard.Transport()}),
)

The Go wrapper provides helper functions for base URL and transport configuration. All requests flow through Stockyard with full middleware support.

Environment Variables

All SDK wrappers respect these environment variables:

Variable	Default	Description
`STOCKYARD_URL`	`http://localhost:4200`	Stockyard proxy base URL
`STOCKYARD_API_KEY`	—	API key for authenticated Stockyard instances
`OPENAI_API_KEY`	—	Your OpenAI API key (passed through to provider)
`STOCKYARD_TIMEOUT`	`30s`	Request timeout for proxy calls
`STOCKYARD_RETRY`	`true`	Enable client-side retries on transient errors
`STOCKYARD_VERIFY_SSL`	`true`	Verify SSL certificates (disable for local dev)

Cloud users: Set STOCKYARD_URL to your cloud endpoint (e.g., https://your-team.stockyard.dev) and the SDK handles everything else.

FAQ

Do the wrappers support streaming?: Yes. Streaming works identically to the official SDKs. The wrapper passes through SSE streams from the proxy without buffering.
What if Stockyard is down?: By default, the wrappers fail open — if the proxy is unreachable and STOCKYARD_RETRY is enabled, the SDK retries up to 3 times. You can also configure a fallback to call the provider directly.
Can I use multiple providers through the same SDK?: Yes. Stockyard routes based on the model name. Use model: "claude-sonnet-4-20250514" through the OpenAI SDK wrapper and Stockyard routes it to Anthropic automatically.
Do I need to change my tests?: No. The wrapper has the same interface as the official SDK. Your existing mocks and test fixtures work unchanged.
What about Anthropic and other provider SDKs?: Wrappers for the Anthropic SDK, Google Vertex SDK, and others are in development. In the meantime, set base_url manually to point at your Stockyard instance.

← Exchange Ops Guide →