Confidential AI API: Secure LLM Inference at Scale | Prem

Basic API Key Authentication

Authenticate with your API key:

import createRvencClient from "@premai/api-sdk";

const client = await createRvencClient({
  apiKey: process.env.PREM_API_KEY,
  clientKEK: process.env.CLIENT_KEK
});

// Verify authentication
const response = await client.chat.completions.create({
  model: "deepseek-v4-flash",
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(response.choices[0].message.content);

Use Pre-generated Keys

Use pre-generated encryption keys for better performance:

import createRvencClient, { generateEncryptionKeys } from "@premai/api-sdk";

// Generate keys once
const encryptionKeys = await generateEncryptionKeys();

// Reuse keys across multiple requests
const client = await createRvencClient({
  apiKey: process.env.PREM_API_KEY,
  encryptionKeys,
});

const response = await client.chat.completions.create({
  model: "deepseek-v4-flash",
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(response.choices[0].message.content);

Use Custom Options

Configure custom options such as timeout and buffer size:

const client = await createRvencClient({
  apiKey: process.env.PREM_API_KEY,
  clientKEK: process.env.CLIENT_KEK,
  requestTimeoutMs: 60000, // 60 seconds
  maxBufferSize: 20 * 1024 * 1024, // 20MB
});