Quick Start

Installation

Install the Nvisy SDK from npm:

npm install @nvisy/sdk

yarn add @nvisy/sdk

pnpm add @nvisy/sdk

NPM Package

View the package on npm

Requirements

Node.js 20 or later (or any ES2022+ runtime with fetch and a global File)
An API token — create one under API Tokens in your account settings

Uploads construct a File, which Node.js exposes globally from version 20 onwards. On Node.js 18, import it explicitly with import { File } from "node:buffer".

Your First Request

Construct a client with your API token and read your account:

import { Nvisy } from "@nvisy/sdk";

const client = new Nvisy({ apiToken: process.env.NVISY_API_TOKEN! });

const account = await client.account.getAccount();
const workspaces = await client.workspaces.listWorkspaces();

console.log(account.username, workspaces.items.length);

Everything except authentication and account management is scoped to a workspace, identified by its slug. List your workspaces first to find it.

Configuration

The constructor accepts additional options:

const client = new Nvisy({
  apiToken: "your-api-token",       // Required
  baseUrl: "https://api.nvisy.com", // Optional
  userAgent: "MyApp/1.0.0",         // Optional
  withLogging: true,                // Optional — debug logging
  headers: {                        // Optional
    "X-Custom-Header": "value",
  },
});

Option	Type	Required	Default
`apiToken`	`string`	Yes	—
`baseUrl`	`string`	No	`https://api.nvisy.com`
`userAgent`	`string`	No	`@nvisy/sdk v.<version>`
`withLogging`	`boolean`	No	`false`
`headers`	`Record<string, string>`	No	—

The Redaction Workflow

Redaction is a two-phase process: a run first detects sensitive data, then you apply redactions after reviewing the findings.

Upload a file

const [file] = await client.files.uploadFiles(workspaceSlug, blob);

Start a run against a pipeline

Creation returns immediately; analysis continues in the background.

let run = await client.runs.createRun(workspaceSlug, pipelineSlug, {
  fileId: file.id,
});

Wait for analysis to finish

Subscribe to the run’s event stream. It emits the current status immediately, then each transition, and ends once the run settles.

const response = await client.runs.events(workspaceSlug, run.id);
const reader = response.body!.pipeThrough(new TextDecoderStream()).getReader();

let status = run.status;

while (true) {
  const { done, value } = await reader.read();
  if (done) break;

  for (const line of value.split("\n")) {
    if (!line.startsWith("data:")) continue;
    ({ status } = JSON.parse(line.slice(5)));
  }
}

if (status !== "analyzed") {
  throw new Error(`Run ${run.id} ended as ${status}`);
}

Review detections

const audit = await client.runs.getDetections(workspaceSlug, run.id);

Apply redactions

const redacted = await client.runs.redact(workspaceSlug, run.id);

A pipeline defines how a document is analyzed, and a policy defines what is considered sensitive. Create them once, then reuse across runs.

The run row in Postgres is the source of truth, so a missed broadcast is recoverable — re-read the run with getRun if a connection drops. Note that the native EventSource cannot send an Authorization header, which is why the SDK streams over fetch.

TypeScript

Python

Rust

Installation

NPM Package

Requirements

Your First Request

Configuration

The Redaction Workflow

Next Steps

API Reference

Examples

​Installation

NPM Package

​Requirements

​Your First Request

​Configuration

​The Redaction Workflow

​Next Steps

API Reference

Examples

Installation

Requirements

Your First Request

Configuration

The Redaction Workflow

Next Steps