LangChain integration

JavaScript

import { CloudmindCallbackHandler } from "@cloudmind-ai/js/langchain";

const handler = new CloudmindCallbackHandler({
  name: "support-chain",
  metadata: { tenant: orgId },
});

await chain.invoke({ question }, { callbacks: [handler] });

Python

from cloudmind.langchain import CloudmindCallbackHandler

handler = CloudmindCallbackHandler(name="support-chain")
chain.invoke({"question": question}, config={"callbacks": [handler]})

What gets traced

LangChain object	Cloudmind span kind
LLM / ChatModel	llm
Retriever	retrieval
Tool	tool
Chain / Runnable	custom
AgentExecutor step	custom

Nesting follows LangChain's own callback tree, so the Cloudmind trace mirrors your chain structure.

Global handler

To trace everything without passing a handler at each call site:

import langchain
from cloudmind.langchain import CloudmindCallbackHandler

langchain.callbacks = [CloudmindCallbackHandler()]

LangGraph

LangGraph nodes are traced as nested spans automatically when the handler is attached to the compiled graph. Each node becomes a child span of the graph invocation, and conditional edges are recorded as span metadata so you can see which path a run took.