@milkeyskills/sdk

Milkey Skills SDK helps teams connect Milkey skills infrastructure into existing AI products and agent workflows.

Use it when you want to:

• plug Milkey skills into existing OpenAI-compatible clients
• add reusable skill execution to Anthropic, Gemini, or other AI stacks
• keep your model provider client while adding Milkey-managed skills as tools

Install

npm install @milkeyskills/sdk

If your app already uses the OpenAI SDK:

npm install @milkeyskills/sdk openai

Quick Start

import OpenAI from "openai";
import { milkey } from "@milkeyskills/sdk";

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

const milkeyClient = milkey.createClient({
  baseUrl: process.env.MILKEY_BASE_URL!,
  apiKey: process.env.MILKEY_API_KEY!,
});

const tools = milkey.openai.chat.tools({
  client: milkeyClient,
});

const messages = [
  {
    role: "user",
    content: "Find the best Milkey skill for PostgreSQL query optimization.",
  },
];

const first = await openai.chat.completions.create({
  model: "gpt-5",
  messages,
  tools,
});

const assistant = first.choices[0]?.message;
if (!assistant) {
  throw new Error("No assistant message returned.");
}

messages.push({
  role: "assistant",
  content: assistant.content ?? "",
  tool_calls: assistant.tool_calls?.map((toolCall) => ({
    id: toolCall.id,
    type: "function",
    function: {
      name: toolCall.function.name,
      arguments: toolCall.function.arguments,
    },
  })),
});

if (assistant.tool_calls?.length) {
  const toolMessages = await milkey.openai.chat.messages(assistant, milkeyClient);
  for (const toolMessage of toolMessages) {
    messages.push(toolMessage);
  }

  const second = await openai.chat.completions.create({
    model: "gpt-5",
    messages,
    tools,
  });

  console.log(second.choices[0]?.message?.content ?? "");
} else {
  console.log(assistant.content ?? "");
}

For production use, keep iterating until the model stops returning tool_calls and enforce a max turn count or request timeout. See the OpenAI chat completion examples for a bounded loop.

Supported Integrations

• OpenAI-compatible chat completions
• OpenAI responses and realtime-style hosted tool delivery
• Anthropic inline and hosted MCP integrations
• Gemini function-calling integrations, Gemini Interactions hosted MCP helpers, and helpers for the official Google Gen AI SDK
• AI SDK native inline tool helpers

Hosted remote MCP delivery is currently best supported through OpenAI Responses, Anthropic, and Gemini Interactions. OpenAI Chat, Gemini generateContent, and AI SDK default to inline tools for portability and predictability.

Provider Matrix

| Provider | API Variant | mode: "auto" | Supported Modes | Recommended Helper | | --- | --- | --- | --- | --- | | OpenAI | Chat Completions | inline | inline, hosted | milkey.openai.chat.tools(...) | | OpenAI | Responses API | hosted | inline, hosted | milkey.openai.responses.tools(...) | | OpenAI | Realtime | hosted | hosted | milkey.openai.realtime.tools(...) | | Anthropic | Messages API | hosted | inline, hosted | milkey.anthropic.config(...) | | Gemini | generateContent | inline | inline | milkey.gemini.config(...) | | Gemini | Interactions API | hosted | hosted | milkey.gemini.interactions.config(...) | | AI SDK | Inline tools | inline | inline | milkey.aiSdk.tools(...) |

Migration Notes

• Existing delivery options continue to work.
• New mode options are additive and support inline, hosted, and auto where meaningful.
• mode: "auto" now resolves to the best-path adapter mode for that exact provider/API variant.
• OpenAI inline tool schemas now normalize optional properties for strict: true compatibility.
• Anthropic hosted users can prefer milkey.anthropic.config(...) to get the required MCP beta automatically.
• Gemini users now have two documented paths:
  • milkey.gemini.config(...) for generateContent inline function calling
  • milkey.gemini.interactions.config(...) for hosted MCP on the Interactions API
• AI SDK users now have two paths:
  • milkey.aiSdk.tools(...) for inline tools

Versioning

This package follows Semantic Versioning.

• 0.1.0 = first public release
• 0.1.1 = bug fixes
• 0.2.0 = backward-compatible feature updates
• 1.0.0 = stable major release

Update to the latest version with:

npm update @milkeyskills/sdk

Pin a specific version with:

npm install @milkeyskills/[email protected]

Verify Availability

Check that the package is live on npm:

npm view @milkeyskills/sdk --registry=https://registry.npmjs.org

Install it into a project:

npm install @milkeyskills/sdk

Verify the package resolves locally:

import { milkey } from "@milkeyskills/sdk";

console.log(typeof milkey.createClient);

If that prints "function", the SDK is installed correctly.