@trustgateai/sdk

trustgate-node

TrustGate Node.js SDK with Shadow AI detection: workflow context, middleware for Vercel AI SDK and LangChain, n8n header helper, and SSE streaming.

Install

npm install trustgate-node

Optional peer dependencies for middleware:

npm install ai @ai-sdk/openai          # Vercel AI SDK
npm install @langchain/openai         # LangChain

Core client

import { TrustGate, getWorkflowContextFromEnv } from 'trustgate-node';

const client = new TrustGate({
  baseUrl: process.env.TRUSTGATE_BASE_URL!,
  apiKey: process.env.TRUSTGATE_API_KEY,
  // workflowContext optional; otherwise read from process.env (n8n, Vercel, GitHub)
});

// Non-streaming
const res = await client.fetch('/v1/chat/completions', {
  method: 'POST',
  body: JSON.stringify({ model: 'gpt-4', messages: [...] }),
});

// SSE streaming (workflow_id sent in request for background attribution)
const streamRes = await client.fetchStream('/v1/chat/completions', {
  method: 'POST',
  body: JSON.stringify({ model: 'gpt-4', messages: [...], stream: true }),
});
// Consume streamRes.body (ReadableStream) as usual for SSE

Context provider

Workflow metadata is read from process.env so TrustGate can attribute traffic (Shadow AI detection):

Source	Example env vars
n8n	N8N_WORKFLOW_ID, N8N_EXECUTION_ID, N8N_WORKFLOW_NAME
Vercel	VERCEL, VERCEL_ENV, VERCEL_URL, VERCEL_GIT_*
GitHub	GITHUB_ACTION, GITHUB_WORKFLOW, GITHUB_RUN_ID, GITHUB_REPOSITORY

import { getWorkflowContextFromEnv } from 'trustgate-node';

const ctx = getWorkflowContextFromEnv();
// { source: 'n8n', workflow_id: '...', execution_id: '...', metadata: {...} }

Middleware: Vercel AI SDK

Point the OpenAI provider at TrustGate so all calls go through the gateway with workflow context:

import { createOpenAI } from '@ai-sdk/openai';
import { createTrustGateOpenAIOptions } from 'trustgate-node/middleware/vercel-ai';

const openai = createOpenAI(
  createTrustGateOpenAIOptions({
    baseUrl: process.env.TRUSTGATE_BASE_URL!,
    apiKey: process.env.TRUSTGATE_API_KEY,
  })
);

// Use with streamText / generateText as usual
import { streamText } from 'ai';
const result = streamText({
  model: openai('gpt-4'),
  messages: [{ role: 'user', content: 'Hello' }],
});

Middleware: LangChain

Use TrustGate as the OpenAI endpoint for LangChain with one config object:

import { ChatOpenAI } from '@langchain/openai';
import { createTrustGateLangChainConfig } from 'trustgate-node/middleware/langchain';

const config = createTrustGateLangChainConfig({
  baseUrl: process.env.TRUSTGATE_BASE_URL!,
  apiKey: process.env.TRUSTGATE_API_KEY,
});

const llm = new ChatOpenAI({
  ...config,
  modelName: 'gpt-4',
  temperature: 0,
});

n8n HTTP Request node: header snippet

Get a pre-formatted JSON object of headers for n8n “HTTP Request” nodes so each request includes TrustGate auth and workflow context:

import { getN8nHeaderSnippet } from 'trustgate-node';

const { headers, json } = getN8nHeaderSnippet();
// headers: [ { name: 'X-Trustgate-Source', value: 'n8n' }, ... ]
// json: '{"X-Trustgate-Source":"n8n","X-Trustgate-Workflow-Id":"...", ...}'

In n8n, set Send Headers (or equivalent) to the headers array or paste the json into your node config. You can pass apiKey and baseUrl explicitly or rely on TRUSTGATE_API_KEY and TRUSTGATE_BASE_URL in the environment.

Streaming (SSE)

• Use client.fetchStream(path, init, options) for SSE endpoints. The client sets Accept: text/event-stream and sends workflow context (e.g. workflow_id) in the initial request, so the gateway can attribute the background stream.
• The returned Response body is the raw stream; consume it with response.body.getReader() or pass through to your framework’s SSE handling.

License

MIT