Providers - cascadeflow

Built-in Providers

Provider	`provider` value	Environment variable
OpenAI	`openai`	`OPENAI_API_KEY`
Anthropic	`anthropic`	`ANTHROPIC_API_KEY`
Groq	`groq`	`GROQ_API_KEY`
Together AI	`together`	`TOGETHER_API_KEY`
Hugging Face	`huggingface`	`HUGGINGFACE_API_KEY`
OpenRouter	`openrouter`	`OPENROUTER_API_KEY`
Ollama	`ollama`	Not required
vLLM	`vllm`	Depends on deployment

Vercel AI adapters add Azure, Bedrock, Cohere, DeepSeek, Fireworks, Google, Mistral, Perplexity, Vertex, xAI, and other provider identifiers.

Model Configuration

const model = {
  name: 'gpt-4o-mini',
  provider: 'openai' as const,
  cost: 0.00015,
  apiKey: process.env.OPENAI_API_KEY,
  baseUrl: 'https://api.openai.com/v1',
  maxTokens: 1000,
  temperature: 0.2,
};

cost is the configured cost per 1,000 tokens used for cascade ordering. Provider-reported usage and built-in pricing are used when the provider supports detailed cost calculation.

Local Endpoints

const local = new CascadeAgent({
  models: [
    {
      name: 'qwen2.5:7b',
      provider: 'ollama',
      cost: 0,
      baseUrl: 'http://localhost:11434',
    },
  ],
});

Registry

import { getAvailableProviders, providerRegistry } from '@cascadeflow/core';

console.log(getAvailableProviders());
console.log(providerRegistry.has('openai'));

The Provider interface contains generate(), optional stream(), calculateCost(), and isAvailable() methods. Register a custom provider class when the built-in providers do not cover your endpoint. See Providers and Presets for configuration patterns.