Model Gateway

# Primary providers
export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
export GOOGLE_API_KEY=...

# Additional providers
export MISTRAL_API_KEY=...
export GROQ_API_KEY=...
export TOGETHER_API_KEY=...
export OPENROUTER_API_KEY=...

# Local providers
export OLLAMA_BASE_URL=http://localhost:11434
export LITELLM_BASE_URL=http://localhost:4000

import { chat } from './providers';

const response = await chat(
  'anthropic',  // Provider
  messages,     // Conversation
  tools,        // Available tools
  'claude-sonnet-4'  // Specific model (optional)
);

const response = await chat(
  'auto',    // Let gateway choose
  messages,
  tools
);

interface LLMResponse {
  content: string;           // Text response
  toolCalls?: ToolCall[];    // Tool invocations
  finishReason: 'stop' | 'tool_use' | 'length' | 'error';
  usage?: {
    inputTokens: number;
    outputTokens: number;
  };
}

const tool = {
  name: 'shell',
  description: 'Execute a shell command',
  parameters: {
    type: 'object',
    properties: {
      command: {
        type: 'string',
        description: 'The command to run',
      },
    },
    required: ['command'],
  },
};

{
  "name": "shell",
  "description": "Execute a shell command",
  "input_schema": { ... }
}

{
  "type": "function",
  "function": {
    "name": "shell",
    "description": "Execute a shell command",
    "parameters": { ... }
  }
}

// If Anthropic fails, try OpenAI
const response = await chatWithFallback(
  ['anthropic', 'openai'],
  messages,
  tools
);

// Uses native Anthropic SDK
const client = new Anthropic({ apiKey });
const response = await client.messages.create({
  model: 'claude-sonnet-4',
  max_tokens: 8192,
  messages: anthropicMessages,
  tools: anthropicTools,
});

// Uses native OpenAI SDK
const client = new OpenAI({ apiKey });
const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: openaiMessages,
  tools: openaiTools,
});

// Uses Google Generative AI SDK
const genAI = new GoogleGenerativeAI(apiKey);
const model = genAI.getGenerativeModel({ model: 'gemini-2.0-flash' });

// Uses OpenAI SDK with custom base URL
const client = new OpenAI({
  apiKey,
  baseURL: 'https://api.together.xyz/v1',
});

const models = await gateway.listModels('openrouter');
// Returns model list with context lengths, pricing, etc.

interface ModelInfo {
  id: string;
  name: string;
  context_length?: number;
  pricing?: {
    prompt: string;     // per 1K tokens
    completion: string; // per 1K tokens
  };
}

export ANTHROPIC_API_KEY=...  # Primary
export OPENAI_API_KEY=...     # Backup
export OLLAMA_BASE_URL=...    # Offline

const response = await chat(...);
console.log('Tokens:', response.usage);

try {
  return await chat(provider, messages, tools);
} catch (error) {
  if (error.status === 429) {
    await delay(1000);
    return await chat(provider, messages, tools);
  }
  throw error;
}