开源AI工具：连接Google ADK TypeScript和Vercel AI Gateway和100+LLM模型

Requirements

• Node.js >= 18.0.0
• @google/adk >= 0.2.2

Installation

bun add adk-llm-bridge @google/adk

pnpm add adk-llm-bridge @google/adk

npm install adk-llm-bridge @google/adk

Quick Start

import { LlmAgent } from '@google/adk';
import { AIGateway } from 'adk-llm-bridge';

const agent = new LlmAgent({
  name: 'assistant',
  model: AIGateway('anthropic/claude-sonnet-4'),
  instruction: 'You are a helpful assistant.',
});

That's it. All ADK features work: tools, streaming, multi-agent, etc.

Tool Calling Example

import { FunctionTool, LlmAgent } from '@google/adk';
import { Anthropic } from 'adk-llm-bridge';
import { z } from 'zod';

const getWeather = new FunctionTool({
  name: 'get_weather',
  description: 'Get current weather for a city',
  parameters: z.object({
    city: z.string().describe('City name'),
  }),
  execute: ({ city }) => {
    return { status: 'success', weather: 'sunny', city };
  },
});

const agent = new LlmAgent({
  name: 'weather-assistant',
  model: Anthropic('claude-sonnet-4-5'),
  instruction: 'You help users check the weather.',
  tools: [getWeather],
});

Examples

See the examples directory:

• basic-agent-ai-gateway - Multi-agent HelpDesk with AI Gateway
• basic-agent-openrouter - Multi-agent HelpDesk with OpenRouter
• basic-agent-openai - Multi-agent HelpDesk with OpenAI
• basic-agent-anthropic - Multi-agent HelpDesk with Anthropic
• basic-agent-xai - Multi-agent HelpDesk with xAI
• basic-agent-lmstudio - Multi-agent HelpDesk with LM Studio
• express-server - Production HTTP API with sessions, streaming, tools

Configuration

Environment Variables

AI Gateway:

AI_GATEWAY_API_KEY=your-api-key
AI_GATEWAY_URL=https://ai-gateway.vercel.sh/v1  # optional

OpenRouter:

OPENROUTER_API_KEY=your-api-key
OPENROUTER_SITE_URL=https://your-site.com  # optional, for ranking
OPENROUTER_APP_NAME=Your App Name          # optional, for ranking

Direct Providers:

OPENAI_API_KEY=your-openai-key
ANTHROPIC_API_KEY=your-anthropic-key
XAI_API_KEY=your-xai-key

Programmatic Configuration

Pass options directly to the factory functions:

import { AIGateway, OpenRouter, Anthropic } from 'adk-llm-bridge';

// AI Gateway with custom URL
model: AIGateway('anthropic/claude-sonnet-4', {
  apiKey: process.env.MY_API_KEY,
  baseURL: 'https://my-gateway.example.com/v1',
})

// OpenRouter with site info
model: OpenRouter('anthropic/claude-sonnet-4', {
  apiKey: process.env.OPENROUTER_API_KEY,
  siteUrl: 'https://your-site.com',
  appName: 'Your App',
})

// Anthropic with custom max tokens
model: Anthropic('claude-sonnet-4-5', {
  apiKey: process.env.ANTHROPIC_API_KEY,
  maxTokens: 8192,
})

Configuration Options

AIGateway:

**OpenRouter:

**OpenAI:

**Anthropic:

**XAI:

**Custom:

API Reference

Using LLMRegistry (Alternative)

You can also register providers with ADK's LLMRegistry to use string-based model names:

import { LlmAgent, LLMRegistry } from '@google/adk';
import { AnthropicLlm } from 'adk-llm-bridge';

LLMRegistry.register(AnthropicLlm);

const agent = new LlmAgent({
  name: 'assistant',
  model: 'claude-sonnet-4-5',  // String-based model name
  instruction: 'You are a helpful assistant.',
});