How do you call Plugsky from JavaScript and TypeScript?

Key facts

Install	npm install openai
Base URL	https://plugsky.com/v1
Runtimes	Node.js 18+, browser, Bun, Deno, Cloudflare Workers
Frameworks	Vercel AI SDK, LangChain.js supported via OpenAI-compatible clients
Streaming	stream: true with the OpenAI delta format
API shape	OpenAI-compatible chat completions, tools and embeddings
Free plan	plugsky-micro and plugsky-lite, no card required
Product status	Live

TL;DR

• The OpenAI Node library works as-is with a baseURL override.
• Vercel AI SDK needs only a baseURL in the provider options.
• Browser and edge code can call the REST endpoint with fetch.
• Streaming, tools and embeddings all use OpenAI shapes.
• Start on the free plan with plugsky-micro and plugsky-lite.

How it works, step by step

1. Run npm install openai and keep your existing TypeScript types.
2. Create a client with apiKey from an environment variable and baseURL set to Plugsky.
3. Call chat.completions.create with a Plugsky model and your messages.
4. Add stream: true if you want incremental tokens in your UI.
5. Use tools for function calling and response_format for JSON output.
6. Store keys server-side and proxy browser calls through your backend.

Try it yourself

Open the OpenAI-compatible API tester →

Install and configure

The fastest integration is the official OpenAI Node library: run npm install openai, then construct a client with your Plugsky key and baseURL: "https://plugsky.com/v1". Everything else stays the same — types, retries and streaming helpers included. Guided installs for Go, Java, .NET and Rust follow the same pattern if your stack is mixed.

Read the key from an environment variable such as PLUGSKY_KEY and never ship it to a browser bundle.

Node.js with the OpenAI library

A minimal call passes a Plugsky model and a messages array and reads resp.choices[0].message.content. For chat UIs, set stream: true and iterate the async stream, appending chunk.choices[0].delta.content as it arrives. Function calling uses the same tools array as OpenAI, so schemas you already maintain keep working.

If you need automatic routing, set the model to plugsky-fusion and the platform picks a model per request based on your workspace strategy.

Vercel AI SDK and other frameworks

The Vercel AI SDK works through its OpenAI provider: pass openai("plugsky-pro", { baseURL, apiKey }) to generateText or streamText. LangChain.js uses ChatOpenAI with the same base URL override. Haystack, Semantic Kernel and AutoGen are supported in the same way on their respective ecosystems; the common thread is the OpenAI-compatible wire format, so no vendor-specific adapter is required.

Browser, edge and production notes

For browsers use fetch to POST to /v1/chat/completions with your bearer token. On edge runtimes the same endpoint works, but be careful with request size and timeouts: long completions should stream, and very large conversation histories should be trimmed before sending.

• Proxy browser traffic through your backend so keys are never exposed.
• Handle 429 responses using the Retry-After header; SDK retries do this automatically.
• Request bodies are capped at 16 MB — split or trim when you approach it.

Honest comparison

Approach	OpenAI Node library	Vercel AI SDK	Raw fetch
Setup	npm install openai, set baseURL	Provider with baseURL	None — HTTP only
TypeScript types	Bundled	Bundled	You define them
Streaming	Async iterator helper	streamText helper	Manual SSE parsing
Tool calling	OpenAI tools array	Framework helpers	Manual JSON handling
Best for	Most Node services	Next.js and React apps	Edge and minimal bundles