LangChain

cascadeflow integrates with LangChain through a callback handler that wraps any BaseChatModel. It keeps the product direction intact inside LangChain and LangGraph: decisions happen inside agent execution, with budgets, traces, and runtime policy visible where the workflow actually runs.

Install

pip install "cascadeflow[langchain]"

npm install @cascadeflow/langchain @langchain/core @langchain/openai

Quick Start

import cascadeflow
from cascadeflow.integrations.langchain import get_harness_callback
from langchain_openai import ChatOpenAI

cascadeflow.init(mode="observe")

model = ChatOpenAI(model="gpt-4o")
cb = get_harness_callback()

with cascadeflow.run(budget=0.50) as session:
    result = await model.ainvoke("Explain quantum computing", config={"callbacks": [cb]})
    print(session.summary())

from langchain_openai import ChatOpenAI
from langchain_anthropic import ChatAnthropic
from cascadeflow.integrations.langchain import CascadeFlow

cascade = CascadeFlow(
    drafter=ChatOpenAI(model="gpt-4o-mini"),
    verifier=ChatAnthropic(model="claude-sonnet-4"),
    quality_threshold=0.8,
)

result = await cascade.ainvoke("Explain quantum computing")

import { ChatOpenAI } from '@langchain/openai';
import { ChatAnthropic } from '@langchain/anthropic';
import { withCascade } from '@cascadeflow/langchain';

const cascade = withCascade({
  drafter: new ChatOpenAI({ model: 'gpt-4o-mini' }),
  verifier: new ChatAnthropic({ model: 'claude-sonnet-4' }),
  qualityThreshold: 0.8,
});

const result = await cascade.invoke('Explain quantum computing');

Features

Full LCEL support (pipes, sequences, batch)
Streaming with pre-routing
Tool calling and structured output
LangSmith cost tracking metadata
Cost tracking callbacks
Domain policies with cascadeflow_domain metadata

Why This Integration Matters

Keeps LangChain apps framework-native instead of forcing a proxy hop
Makes runtime cost, latency, and trace data visible at the chain or agent level
Lets teams move from observability to governance without rewriting chain logic

Cost Tracking Callback

from cascadeflow.integrations.langchain.langchain_callbacks import get_cascade_callback

with get_cascade_callback() as cb:
    response = await cascade.ainvoke("What is Python?")
    print(f"Total cost: ${cb.total_cost:.6f}")
    print(f"Drafter cost: ${cb.drafter_cost:.6f}")
    print(f"Verifier cost: ${cb.verifier_cost:.6f}")

LangSmith Integration

When LangSmith tracing is enabled, cascadeflow adds metadata to runs:

cascade_decision: whether the drafter was accepted
modelUsed: which model produced the final response
drafterQuality: quality score from validation
savingsPercentage: cost savings achieved

export LANGSMITH_API_KEY="..."
export LANGSMITH_PROJECT="my-project"
export LANGSMITH_TRACING=true

Examples on GitHub: integrations/langchain_harness.py | packages/langchain-cascadeflow/examples/ (6 TypeScript examples)

Overview

Getting Started

Core Concepts

Harness

Integrations

Guides

Resources

Install

Quick Start

Features

Why This Integration Matters

Cost Tracking Callback

LangSmith Integration

​Install

​Quick Start

​Features

​Why This Integration Matters

​Cost Tracking Callback

​LangSmith Integration

Install

Quick Start

Features

Why This Integration Matters

Cost Tracking Callback

LangSmith Integration