sentinelproxy-langchain

A LangChain.js callback handler that mirrors every LLM call to SentinelProxy — the self-healing layer for AI agents (hallucination, prompt-injection and PII detection, plus auto-healed system prompts).

Install

npm install sentinelproxy-langchain

@langchain/core is a peer dependency you'll already have.

Usage

import { SentinelProxyCallbackHandler } from "sentinelproxy-langchain";

const handler = new SentinelProxyCallbackHandler({
  apiKey: "sk_sentinel_…",                       // dashboard → project → Connect
  systemPrompt: "You are a support agent for Acme.",
});

await llm.invoke(messages, { callbacks: [handler] });

Pass the handler per-call as above, or attach it at construction with callbacks: [handler] to apply it to every call.

Logging is best-effort and off your critical path — failures are swallowed so it can never break the chain. PII is scrubbed at the SentinelProxy edge before anything is stored.

Build & publish

npm install
npm run build
npm publish    # unscoped public package

License

MIT