Skip to main content

Quickstart

With nilAI, it is possible to run AI models within a trusted execution environment (TEE) via the secretLLM SDK. This makes it possible to build new private AI applications or migrate existing ones to run on a secure nilAI node so that your data remains private.

Getting Started

In this quickstart, we will interact with a private AI chat/response application via Next.js. Let's get started by cloning our examples repo.

gh repo clone NillionNetwork/blind-module-examples
cd blind-module-examples/nilai/secretllm_nextjs_nucs

Authentication

cp .env.example .env

Now we need to set NILLION_API_KEY using a key from the Nillion Subscription Portal below.

Usage

secretLLM quickstart

We have two approaches for using the nilAI API via secretLLM:

  • Direct API Access: Recommended for a solo developer/organization.
  • Delegation Token: Provides permissions to another user or organization.

Direct API Access

  1. Get the API key from your .env file.
  2. Check if the message and API key exist.
  3. Initialize the nilAI OpenAI client with baseURL, apiKey, and nilauthInstance.
  4. Make a request to the chat client with the model and message.
  5. Receive the response from message.content.
nilai/secretllm_nextjs_nucs/app/api/chat/route.ts
loading...

Delegation Token Access

  1. Similar to Direct API access, except using DELEGATION_TOKEN authentication.
  2. Server initializes a delegation token server.
  3. Client produces a delegation request.
  4. Server creates the delegation token.
  5. Client uses the delegation token with the model and message for the request.
  6. Response is delivered.
nilai/secretllm_nextjs_nucs/app/api/chat-delegation/route.ts
loading...

Customization

You can also customize the types of models you want to use. Currently available models are listed here.

What you've done

🎉 Congratulations! You just built and interacted with a privacy‑preserving LLM application:

  1. You (Builder) get access to the secretLLM SDK.
  2. You (User) can provide a prompt to the LLM.
  3. The LLM understands your prompt and returns an answer via direct or delegated access.

This demonstrates a core principle of private AI: you can create endless private AI applications via Nillion.

?

Were you able to complete the quickstart?