@codeworksh/aikit

AiKit is a TypeScript SDK that provides a unified API for working with multiple LLM providers, automatic model discovery, provider configurations, token and cost tracking, and mid-session hand-off to other models. It gives you the basic primitives for streaming LLM responses without the extra bloat, letting you handle the orchestration yourself.

Table of Contents

• Installation
• Supported Providers
• Quick Start
• Tools
• Providers and Models
• CLI Tools
  • Model Generation
  • OAuth Providers
• Contribute

Installation

npm install @codeworksh/aikit

Node >=22.0.0 is required.

Supported Providers

Built on top of the Vercel AI SDK, @codeworksh/aikit supports a wide array of providers natively:

• Anthropic (@ai-sdk/anthropic)
• OpenAI (@ai-sdk/openai)
• Google (@ai-sdk/google)
• Google Vertex AI (@ai-sdk/google-vertex)
• xAI (@ai-sdk/xai)
• OpenRouter (@openrouter/ai-sdk-provider)
• OpenAI Compatible APIs (@ai-sdk/openai-compatible)

Quick Start

import { llm, stream, Message } from "@codeworksh/aikit";

// Resolve a model from the built-in registry
const model = await llm("anthropic", "claude-haiku-4-5-20251001");
if (!model) throw new Error("Model not found");

// Setup conversation context
const context: Message.Context = {
    systemPrompt: "You are a helpful coding assistant.",
    messages: [
        Message.createUserMessage({
            role: "user",
            time: { created: Date.now() },
            parts: [{ type: "text", text: "Write a simple loop in TypeScript." }],
        }),
    ],
};

// Option 1: Stream events
const s = stream(model, context, { apiKey: process.env.ANTHROPIC_API_KEY });

for await (const event of s) {
    if (event.type === "text.delta") {
        process.stdout.write(event.delta);
    }
}

const finalMessage = await s.result();
console.log(`\nMessage ID: ${finalMessage.messageId}`);

// Option 2: Get complete response without streaming
const message = await stream.complete(model, context, { apiKey: process.env.ANTHROPIC_API_KEY });
console.log(message.parts);

Tools

Tools enable LLMs to interact with external systems. @codeworksh/aikit provides native support for function calling and schema validation using TypeBox.

import { Type, llm, stream, Message, validateToolArguments } from "@codeworksh/aikit";

const calculatorTool = {
    name: "calculator",
    description: "Evaluate arithmetic expressions",
    parameters: Type.Object({
        expression: Type.String(),
    }),
};

const context: Message.Context = {
    messages: [
        Message.createUserMessage({
            role: "user",
            time: { created: Date.now() },
            parts: [{ type: "text", text: "What is 25 * 18?" }],
        }),
    ],
    tools: [calculatorTool],
};

const response = await stream.complete(model, context, { apiKey: process.env.ANTHROPIC_API_KEY });

// Check for tool calls in the response
for (const part of response.parts) {
    if (part.type === "toolCall") {
        console.log(`Executing tool: ${part.name}`);

        // Validates arguments against TypeBox schema automatically
        const args = validateToolArguments(calculatorTool.parameters, part.arguments);
        // Execute your tool logic here...
    }
}

Providers and Models

@codeworksh/aikit uses a registry to fetch model specifications and metadata.

import { llm } from "@codeworksh/aikit";

// Get all available providers
const providers = await llm.providers();
console.log(providers);

// Get all models for a specific provider
const anthropicModels = await llm.models("anthropic");

// Get a specific model directly
const gpt4 = await llm.model("openai", "gpt-4o");

CLI Tools

@codeworksh/aikit ships with a CLI tool for managing local metadata and authentication.

You can invoke it using npx aikit or pnpm aikit.

Model Generation

Generate a models.gen.json file fetching the latest metadata about providers and their available models from models.dev.

pnpm aikit modelgen [path]

OAuth Providers

Manage OAuth credentials for providers that require it, such as OpenAI Codex and GitHub Copilot.

# Start an OAuth login flow in your browser
pnpm aikit auth --openai-codex

# Codex with a device code instead: no localhost callback port, so it works
# over SSH, in containers, and when something else holds port 1455
pnpm aikit auth --openai-codex --device

# GitHub Copilot uses the device flow; add --enterprise <domain> for GitHub Enterprise
pnpm aikit auth --github-copilot

# Check the status of your stored credentials
pnpm aikit auth --openai-codex --status
pnpm aikit auth --github-copilot --status

# Refresh your current credentials
pnpm aikit auth --openai-codex --refresh

# Clear stored credentials
pnpm aikit auth --openai-codex --logout
pnpm aikit auth --github-copilot --logout

Contribute

@codeworksh/aikit is open to community contribution. Please ensure you submit an issue before submitting a pull request. The project prefers open community discussion before accepting new features.