Skip to main content

​
TypeScript API Reference

cascadeflow provides three TypeScript packages for different integration patterns.

​
Quick Start

import { CascadeAgent } from '@cascadeflow/core';

const agent = new CascadeAgent({
  models: [
    { name: 'gpt-4o-mini', provider: 'openai', cost: 0.000375 },
    { name: 'gpt-4o', provider: 'openai', cost: 0.00625 },
  ],
});

const result = await agent.run('Summarize this document');
console.log(result.content);
console.log(`Model: ${result.modelUsed}, Cost: $${result.totalCost}`);

​
Packages

PackagePurposeDocs
@cascadeflow/coreCascadeAgent with speculative execution and cost trackingReference
@cascadeflow/vercel-aiVercel AI SDK middleware with streaming and tool loopsReference
@cascadeflow/langchainLangChain withCascade() wrapper and model discoveryReference

​
Install

# Core package
npm install @cascadeflow/core

# Vercel AI SDK integration
npm install @cascadeflow/vercel-ai

# LangChain integration
npm install @cascadeflow/langchain @langchain/core @langchain/openai

​
Core Concepts

Speculative execution — the cheaper model (drafter) runs first. If its response passes quality validation, cascadeflow returns it without calling the expensive model. If not, the verifier model runs as fallback. Quality validation — configurable confidence threshold, minimum token count, and optional semantic validation determine whether a draft is accepted. Cost tracking — every response includes totalCost and savingsPercentage so you can measure optimization impact.

​
Integration Patterns

​
Standalone Agent

Use @cascadeflow/core directly for full control: 
const result = await agent.run('What is TypeScript?');
// result.modelUsed → 'gpt-4o-mini' (if draft accepted)
// result.savingsPercentage → 94

​
Vercel AI SDK

Use @cascadeflow/vercel-ai for Next.js and Edge deployments: 
import { createChatHandler } from '@cascadeflow/vercel-ai';

const handler = createChatHandler(agent, {
  protocol: 'data',
  maxSteps: 5,
});

​
LangChain

Use @cascadeflow/langchain for LCEL chains and pipelines: 
import { withCascade } from '@cascadeflow/langchain';

const cascade = withCascade({
  drafter: new ChatOpenAI({ model: 'gpt-4o-mini' }),
  verifier: new ChatAnthropic({ model: 'claude-sonnet-4' }),
});

const chain = prompt.pipe(cascade).pipe(new StringOutputParser());

​
Runtime Support

All packages work in Node.js, Browser, and Edge Functions (Vercel Edge, Cloudflare Workers).