Integration Guide

LangChain Integration

LangChain integration uses the generic wrap() method on the RANKIGI SDK. Wrap your agent or executor invocation and the input args plus the final result are SHA-256 hashed and appended to your tamper-evident chain.

Callback Handler

The Node SDK ships RankigiCallbackHandler, which hooks every LangChain lifecycle event (tool start/end, agent action, LLM end, errors) and seals each into your chain.

typescript

import { Rankigi } from "@rankigi/sdk";
import { RankigiCallbackHandler } from "@rankigi/sdk/langchain";

const rk = Rankigi.fromEnv();
const handler = new RankigiCallbackHandler({ client: rk, sessionId: "research-agent" });

const executor = new AgentExecutor({ agent, tools, callbacks: [handler] });

Prefer the generic wrap() pattern below when you want one event per invocation instead of per lifecycle hook.

Installation

Node.js / TypeScript

bash

npm install @rankigi/sdk @langchain/core langchain

Python

bash

pip install rankigi langchain-core langchain

Environment

Use npx @rankigi/cli init to issue a credential and write RANKIGI_CREDENTIAL and RANKIGI_CHAIN_ID into your .env.

bash

RANKIGI_CREDENTIAL=rnk_live_cred_v1...
RANKIGI_CHAIN_ID=chn_your_chain_id
OPENAI_API_KEY=sk-...

Quick Start. Node.js

typescript

import "dotenv/config";
import { Rankigi } from "@rankigi/sdk";
import { AgentExecutor, createOpenAIFunctionsAgent } from "langchain/agents";
import { ChatOpenAI } from "@langchain/openai";

const rk = Rankigi.fromEnv();

const llm = new ChatOpenAI({ modelName: "gpt-4" });
const agent = await createOpenAIFunctionsAgent({ llm, tools, prompt });
const executor = new AgentExecutor({ agent, tools });

// Wrap the executor invocation. The input args and the final
// output are SHA-256 hashed and appended to your chain.
const input = { input: "your task" };
const result = await rk.wrap(
  "langchain-run",
  input,
  () => executor.invoke(input),
);

await rk.close();

Quick Start. Python

python

from dotenv import load_dotenv
load_dotenv()

from rankigi import Rankigi
from langchain.agents import initialize_agent, AgentType
from langchain_openai import ChatOpenAI

rk = Rankigi.from_env()

llm = ChatOpenAI(model="gpt-4")
agent = initialize_agent(
    tools=tools,
    llm=llm,
    agent=AgentType.OPENAI_FUNCTIONS,
)

# Wrap the agent.run() call. Input and output are hashed
# and appended to your tamper-evident chain.
result = rk.wrap(
    "langchain-run",
    {"input": "your task"},
    lambda: agent.run("your task"),
)

rk.close()

Sidecar Guarantees

Blocking behaviorNever blocks agent execution

Error propagationFailures logged to stderr, never thrown

Data transmittedSHA-256 hashes only, no raw input/output

Transport (Node.js)One POST per event, 3 retries with exponential backoff

Transport (Python)Daemon thread queue, no retry