Hautech SDK

Installation

npm install @hautechai/sdk
# or
yarn add @hautechai/sdk

Usage

Creating tokens

To use the SDK, first you need to create a token via token signer.

// THIS CODE SHOULD BE RUN ON SERVER SIDE ONLY!!!

import { createTokenSigner } from '@hautechai/sdk';

const signer = createTokenSigner({
    appId: process.env.APP_ID!,
    appKeyId: process.env.APP_KEY_ID!,
    appKeySecret: process.env.APP_KEY_SECRET!,
});

Root token

For creating token that can be used on server side, you can create a root token.

const rootToken = signer.createRootToken({ expiresInSeconds: 3600 });

It's very not recommended to use root token on client side.

Account token

For creating token that can be used on client side by your users, you can create an account token.

const accountToken = signer.createAccountToken({ accountId: 'ACCOUNT_ID', expiresInSeconds: 3600 });

Initializing SDK

To initialize the SDK, you need to pass the function that returns the token to the createSDK function.

import { createSDK } from '@hautechai/sdk';

const sdk = createSDK({ authToken: () => accountToken }); // you should call the server here for getting the token

Handling request errors and retries

You can centralize HTTP error handling by providing an onRequestError hook. The hook receives the typed RequestErrorInfo and RequestContextInfo, and can instruct the SDK to retry the request, invalidate the cached token, and optionally wait before retrying.

const sdk = createSDK({
    baseUrl,
    authToken: getCurrentToken,
    onRequestError: async ({ error }) => {
        if (error.status === 401) {
            await refreshToken();
            return { retry: true, invalidateToken: true };
        }

        return { retry: false };
    },
});

// Later, when using websockets, you can disconnect explicitly
sdk.ws.disconnect();

Using SDK

Docs about how to use the SDK are available here

Seedream v5 Lite imagine operations

// Text-to-image
const t2iOperation = await sdk.operations.run.imagine.seedream['5_lite_t2i'].v1({
    input: {
        prompt: 'A cinematic skyline at sunrise above a crystal bay',
        width: 2048,
        height: 2048,
    },
});

// Image edit
const editOperation = await sdk.operations.run.imagine.seedream['5_lite_edit'].v1({
    input: {
        prompt: 'Reimagine this scene as a watercolor illustration',
        imageIds: ['existing-image-id'],
    },
});

// Wait for the generated assets
const finalizedT2i = await sdk.operations.wait(t2iOperation);
const finalizedEdit = await sdk.operations.wait(editOperation);

Both variants accept dimensions up to 3072×3072. Pricing endpoints report a cost of $0.04375 USD per run for the text-to-image and edit models—use sdk.pricing.list() or sdk.pricing.get('seedream.5_lite_t2i.v1') / sdk.pricing.get('seedream.5_lite_edit.v1') to confirm the current catalog pricing.

Google Nano Banana prompt-only vs. edit mode

// Prompt-only generation (no source images required)
const promptOnly = await sdk.operations.run.google.nano_banana_pro.edit.v1({
    input: {
        prompt: 'Generate a high-fashion catalog shot of a model in neon streetwear',
        // imageIds omitted on purpose
    },
});

// Classic edit mode with at least one source image
const editFromImage = await sdk.operations.run.google.nano_banana_pro.edit.v1({
    input: {
        prompt: 'Apply a cinematic golden-hour lighting pass to this asset',
        imageIds: ['existing-image-id'],
    },
});

The backend Swagger DTO (GoogleNanoBananaProEditV1Input) intentionally marks imageIds as optional so prompt-only runs remain valid. Provide one or more IDs when you need edit behavior; omit the field when you want the model to generate a fresh asset purely from text.

Uploading files

• In browsers, prefer passing a File object to methods like sdk.images.createFromFile and sdk.videos.createFromFile. The File.name will be included in the multipart upload automatically.
• If you pass a Blob in the browser, some environments may set the filename to a default value (e.g., "blob"). To control the filename when using a Blob, wrap it in an object with explicit metadata on server-side, or construct a File from the Blob.
• In Node.js, when you pass a string path (e.g., /path/to/image.png), the SDK sets the multipart filename to the basename of the path (e.g., image.png). You can also pass an object with { stream, filename, contentType } to control metadata explicitly.

You can also expire uploads automatically by specifying a TTL when finalizing image uploads:

await sdk.images.createFromFile('/path/to/image.png', { ttlSeconds: 3600 });

Development

Prerequisites

• Node.js 22+
• PNPM 10+

Installation

pnpm install --frozen-lockfile

Building

pnpm build

Testing

# Run all tests (unit + e2e)
pnpm test

# Run unit tests only
pnpm test:unit

# Run e2e tests only
pnpm test:e2e

# Run tests with coverage
pnpm test:cov

Linting and Formatting

# Check code formatting
npx prettier --check .

# Fix code formatting
npx prettier --write .

# Type checking
npx tsc --noEmit