@mantey/saia-ai-sdk

Vercel AI SDK provider for GWDG's SAIA (Scalable Artificial Intelligence Accelerator) API. This package provides an OpenAI-compatible interface for the SAIA API, allowing seamless integration with the Vercel AI SDK.

Installation

npm install @mantey/saia-ai-sdk

pnpm add @mantey/saia-ai-sdk

yarn add @mantey/saia-ai-sdk

Setup

1. Get your SAIA API Key

To use SAIA, you need access to the GWDG SAIA service:

1. Visit the GWDG SAIA Documentation
2. Request access to the SAIA service if you're affiliated with a GWDG member institution
3. Obtain your API key

2. Install Dependencies

This package requires the Vercel AI SDK:

npm install ai @mantey/saia-ai-sdk

Note:

• ai - The Vercel AI SDK (required for using any of vercel's core AI SDK)
• @ai-sdk/openai-compatible, @ai-sdk/provider, and @ai-sdk/provider-utils are regular dependencies and will be installed automatically

Usage

Basic Setup

import { createSaia } from '@mantey/saia-ai-sdk';
import { generateText } from 'ai';

const saia = createSaia({
  apiKey: process.env.SAIA_API_KEY,
});

const { text } = await generateText({
  model: saia('meta-llama-3.1-8b-instruct'),
  prompt: 'Explain quantum computing in simple terms',
});

console.log(text);

Using the Default Instance

import { saia } from '@mantey/saia-ai-sdk';
import { generateText } from 'ai';

// Uses SAIA_API_KEY environment variable
const { text } = await generateText({
  model: saia('meta-llama-3.1-8b-instruct'),
  prompt: 'Write a haiku about TypeScript',
});

Streaming Responses

import { createSaia } from '@mantey/saia-ai-sdk';
import { streamText } from 'ai';

const saia = createSaia({
  apiKey: process.env.SAIA_API_KEY,
});

const { textStream } = await streamText({
  model: saia('meta-llama-3.1-8b-instruct'),
  prompt: 'Count from 1 to 10',
});

for await (const chunk of textStream) {
  process.stdout.write(chunk);
}

Chat Completions

import { createSaia } from '@mantey/saia-ai-sdk';
import { generateText } from 'ai';

const saia = createSaia({
  apiKey: process.env.SAIA_API_KEY,
});

const { text } = await generateText({
  model: saia('meta-llama-3.1-8b-instruct'),
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'What is the capital of France?' },
  ],
});

Vision Models (Image Understanding)

import { createSaia } from '@mantey/saia-ai-sdk';
import { generateText } from 'ai';
import fs from 'fs';

const saia = createSaia({
  apiKey: process.env.SAIA_API_KEY,
});

// Read image and convert to base64
const imageBuffer = fs.readFileSync('./path/to/image.jpg');
const base64Image = imageBuffer.toString('base64');

const { text } = await generateText({
  model: saia('gemma-3-27b-it'),
  messages: [
    {
      role: 'user',
      content: [
        { type: 'text', text: 'What is in this image?' },
        {
          type: 'image',
          image: `data:image/jpeg;base64,${base64Image}`
        },
      ],
    },
  ],
});

console.log(text);

Note: SAIA requires images to be uploaded as base64-encoded data or file uploads. Fetching images from URLs is not supported.

Embeddings

import { createSaia } from '@mantey/saia-ai-sdk';
import { embed } from 'ai';

const saia = createSaia({
  apiKey: process.env.SAIA_API_KEY,
});

const { embedding } = await embed({
  model: saia.textEmbeddingModel('multilingual-e5-large-instruct'),
  value: 'Hello, world!',
});

console.log(embedding);

Configuration

Provider Options

import { createSaia } from '@mantey/saia-ai-sdk';

const saia = createSaia({
  // Your SAIA API key (can also use SAIA_API_KEY environment variable)
  apiKey: 'your-api-key',

  // Custom base URL (optional)
  baseURL: 'https://chat-ai.academiccloud.de/v1',

  // Custom headers (optional)
  headers: {
    'X-Custom-Header': 'value',
  },

  // Custom fetch implementation (optional)
  fetch: customFetch,
});

Available Models (From SAIA Documentation)

Chat Models

General purpose text generation models:

• meta-llama-3.1-8b-instruct
• llama-3.3-70b-instruct
• openai-gpt-oss-120b
• mistral-large-instruct
• teuken-7b-instruct-research

Code Generation:

• qwen2.5-coder-32b-instruct
• codestral-22b

Reasoning Models:

• deepseek-r1
• qwen3-235b-a22b
• qwq-32b
• qwen3-32b

RAG-Enabled Models:

• meta-llama-3.1-8b-rag
• llama-3.1-sauerkrautlm-70b-instruct

Vision Models (Multimodal)

Models that can analyze images and videos:

• gemma-3-27b-it
• qwen2.5-vl-72b-instruct
• medgemma-27b-it - Medical image analysis (Text + Image)
• internvl2.5-8b
• qwen2.5-omni-7b

Note: Images must be uploaded as part of requests (base64 or file upload). Fetching images from URLs is not supported.

Embedding Models

Text embedding models for semantic search and similarity:

• e5-mistral-7b-instruct
• multilingual-e5-large-instruct - Multilingual support
• qwen3-embedding-4b

Environment Variables

You can set your SAIA API key via environment variable:

export SAIA_API_KEY=your-api-key

Or use a .env file:

SAIA_API_KEY=your-api-key

TypeScript Support

This package is written in TypeScript and provides full type definitions. All model IDs, settings, and options are properly typed.

import type {
  SAIAChatModelId,
  SAIAEmbeddingModelId,
  SAIAProviderOptions
} from '@mantey/saia-ai-sdk';

const modelId: SAIAChatModelId = 'meta-llama-3.1-8b-instruct';

API Reference

createSaia(options?)

Creates a new SAIA provider instance.

Parameters:

• options (optional): Provider configuration options
  • apiKey: Your SAIA API key
  • baseURL: Custom base URL (default: https://chat-ai.academiccloud.de/v1)
  • headers: Custom headers to include in requests
  • fetch: Custom fetch implementation

Returns: SAIA provider instance compatible with Vercel AI SDK

saia

Default SAIA provider instance that uses the SAIA_API_KEY environment variable.

Examples

Check out the examples directory for more usage examples.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT

Links

• Vercel AI SDK Documentation
• GWDG SAIA Documentation
• GitHub Repository

Support

For issues and questions:

• Open an issue on GitHub
• Check the Vercel AI SDK documentation