SAP AI Core

Important Note

Third-Party Provider: This SAP AI Core provider (@mymediset/sap-ai-provider) is developed and maintained by Mymediset, not by SAP SE. While it uses the official SAP AI SDK and integrates with SAP AI Core services, it is not an official SAP product. For official SAP AI solutions, please refer to the SAP AI Core Documentation.

SAP AI Core is SAP's enterprise-grade AI platform that provides access to leading AI models from OpenAI, Anthropic, Google, Amazon, and more through a unified, secure, and scalable infrastructure. The SAP AI Core provider for the AI SDK enables seamless integration with enterprise AI capabilities while offering unique advantages:

  • Multi-Model Access: Support for 40+ models including GPT-4, Claude, Gemini, and Amazon Nova
  • Automatic Authentication: Built-in credential handling via SAP AI SDK
  • Data Masking: Built-in SAP Data Privacy Integration (DPI) for privacy
  • Content Filtering: Azure Content Safety and Llama Guard support
  • Cost Management: Enterprise billing and usage tracking through SAP BTP
  • High Availability: Enterprise-grade infrastructure with SLA guarantees
  • Hybrid Deployment: Support for both cloud and on-premise deployments
  • Tool Calling: Full function calling capabilities for compatible models
  • Multi-modal Support: Text and image inputs for compatible models

Learn more about SAP AI Core's capabilities in the SAP AI Core Documentation.

Setup

The SAP AI Core provider is available in the @mymediset/sap-ai-provider module. You can install it with:

pnpm add @mymediset/sap-ai-provider ai

Provider Instance

To create an SAP AI Core provider instance, use the createSAPAIProvider function. The provider is now synchronous and authentication is handled automatically via the SAP AI SDK:

import { createSAPAIProvider } from '@mymediset/sap-ai-provider';
// Authentication via AICORE_SERVICE_KEY environment variable
const sapai = createSAPAIProvider();

You can obtain your SAP AI Core service key from your SAP BTP Cockpit by creating a service key for your AI Core instance, then set it as the AICORE_SERVICE_KEY environment variable.

Language Models

You can create SAP AI Core models using the provider instance and model name:

// Azure OpenAI models
const gpt4Model = sapai('gpt-4o');
// Anthropic models (via AWS Bedrock)
const claudeModel = sapai('anthropic--claude-3.5-sonnet');
// Google Vertex AI models
const geminiModel = sapai('gemini-2.5-flash');
// Amazon Nova models (via AWS Bedrock)
const novaModel = sapai('amazon--nova-pro');

Supported Models

The provider supports a wide range of models available in your SAP AI Core deployment:

Azure OpenAI Models

  • gpt-4o, gpt-4o-mini
  • gpt-4.1, gpt-4.1-mini, gpt-4.1-nano
  • o1, o3, o3-mini, o4-mini

Anthropic Models (AWS Bedrock)

  • anthropic--claude-3-haiku, anthropic--claude-3-sonnet, anthropic--claude-3-opus
  • anthropic--claude-3.5-sonnet, anthropic--claude-3.7-sonnet
  • anthropic--claude-4-sonnet, anthropic--claude-4-opus

Google Vertex AI Models

  • gemini-2.0-flash, gemini-2.0-flash-lite
  • gemini-2.5-flash, gemini-2.5-pro

Amazon Nova Models (AWS Bedrock)

  • amazon--nova-premier, amazon--nova-pro, amazon--nova-lite, amazon--nova-micro

AI Core Open Source Models

  • mistralai--mistral-large-instruct, mistralai--mistral-medium-instruct, mistralai--mistral-small-instruct
  • cohere--command-a-reasoning

Note: Model availability may vary based on your SAP AI Core subscription and region. Some models may require additional configuration or permissions.

Examples

Here are examples of using SAP AI Core with the AI SDK:

generateText

import { createSAPAIProvider } from '@mymediset/sap-ai-provider';
import { generateText } from 'ai';
// Authentication via AICORE_SERVICE_KEY environment variable
const sapai = createSAPAIProvider();
const { text } = await generateText({
model: sapai('gpt-4o'),
prompt: 'What are the benefits of enterprise AI platforms?',
});
console.log(text);

streamText

import { createSAPAIProvider } from '@mymediset/sap-ai-provider';
import { streamText } from 'ai';
const sapai = createSAPAIProvider();
const result = await streamText({
model: sapai('anthropic--claude-3.5-sonnet'),
prompt: 'Write a short story about AI.',
});
for await (const textPart of result.textStream) {
process.stdout.write(textPart);
}

Tool Calling

import { createSAPAIProvider } from '@mymediset/sap-ai-provider';
import { generateText, tool } from 'ai';
import { z } from 'zod';
const sapai = createSAPAIProvider();
const result = await generateText({
model: sapai('gpt-4o'),
prompt: 'What is the current status of our inventory system?',
tools: {
checkInventory: tool({
description: 'Check current inventory levels',
parameters: z.object({
item: z.string().describe('Item to check'),
location: z.string().describe('Warehouse location'),
}),
execute: async ({ item, location }) => {
// Your inventory system integration
return { item, location, quantity: 150, status: 'In Stock' };
},
}),
},
maxSteps: 3,
});
console.log(result.text);

Multi-modal Input

import { createSAPAIProvider } from '@mymediset/sap-ai-provider';
import { generateText } from 'ai';
const sapai = createSAPAIProvider();
const result = await generateText({
model: sapai('gpt-4o'),
messages: [
{
role: 'user',
content: [
{ type: 'text', text: 'Analyze this business process diagram.' },
{
type: 'image',
image: new URL('https://example.com/diagram.jpg'),
},
],
},
],
});
console.log(result.text);

Data Masking (SAP DPI)

Use SAP's Data Privacy Integration to automatically mask sensitive data:

import {
createSAPAIProvider,
buildDpiMaskingProvider,
} from '@mymediset/sap-ai-provider';
import { generateText } from 'ai';
const dpiConfig = buildDpiMaskingProvider({
method: 'anonymization',
entities: [
'profile-email',
'profile-person',
{
type: 'profile-phone',
replacement_strategy: { method: 'constant', value: 'REDACTED' },
},
],
});
const sapai = createSAPAIProvider({
defaultSettings: {
masking: {
masking_providers: [dpiConfig],
},
},
});
const result = await generateText({
model: sapai('gpt-4o'),
prompt: 'Email john@example.com about the meeting.',
});
console.log(result.text);

Content Filtering

import {
createSAPAIProvider,
buildAzureContentSafetyFilter,
} from '@mymediset/sap-ai-provider';
const sapai = createSAPAIProvider({
defaultSettings: {
filtering: {
input: {
filters: [
buildAzureContentSafetyFilter('input', {
hate: 'ALLOW_SAFE',
violence: 'ALLOW_SAFE_LOW_MEDIUM',
}),
],
},
},
},
});

Configuration

Provider Settings

interface SAPAIProviderSettings {
resourceGroup?: string; // SAP AI Core resource group (default: 'default')
deploymentId?: string; // Specific deployment ID (auto-resolved if not set)
destination?: HttpDestinationOrFetchOptions; // Custom destination
defaultSettings?: SAPAISettings; // Default settings for all models
}

Model Settings

interface SAPAISettings {
modelVersion?: string; // Model version (default: 'latest')
modelParams?: {
maxTokens?: number; // Maximum tokens to generate
temperature?: number; // Sampling temperature (0-2)
topP?: number; // Nucleus sampling parameter
frequencyPenalty?: number; // Frequency penalty (-2 to 2)
presencePenalty?: number; // Presence penalty (-2 to 2)
n?: number; // Number of completions
parallel_tool_calls?: boolean; // Enable parallel tool calls
};
masking?: MaskingModule; // Data masking configuration
filtering?: FilteringModule; // Content filtering configuration
}

Environment Variables

When running on SAP BTP, bind an AI Core service instance to your application. The SDK will automatically detect the service binding from VCAP_SERVICES.

Local Development

Set the AICORE_SERVICE_KEY environment variable with your service key JSON:

AICORE_SERVICE_KEY='{"serviceurls":{"AI_API_URL":"https://..."},"clientid":"...","clientsecret":"...","url":"..."}'

Get your service key from SAP BTP:

  1. Go to your SAP BTP Cockpit
  2. Navigate to your AI Core instance
  3. Create a service key
  4. Copy the JSON and set it as the environment variable

Enterprise Features

SAP AI Core offers several enterprise-grade features:

  • Multi-Tenant Architecture: Isolated environments for different business units
  • Cost Allocation: Detailed usage tracking and cost center allocation
  • Custom Models: Deploy and manage your own fine-tuned models
  • Hybrid Deployment: Support for both cloud and on-premise installations
  • Integration Ready: Native integration with SAP S/4HANA, SuccessFactors, and other SAP solutions

For more information about these features and advanced configuration options, visit the SAP AI Core Documentation.

Migration from v1

Version 2.0 is a complete rewrite using the official SAP AI SDK. Here are the key changes:

Authentication

Before (v1):

const provider = await createSAPAIProvider({
serviceKey: process.env.SAP_AI_SERVICE_KEY,
});

After (v2):

// Set AICORE_SERVICE_KEY env var instead
const provider = createSAPAIProvider();

Provider is now synchronous

Before (v1):

const provider = await createSAPAIProvider({ serviceKey });

After (v2):

const provider = createSAPAIProvider();

Data Masking Configuration

Before (v1):

const dpiMasking = {
type: 'sap_data_privacy_integration',
method: 'anonymization',
entities: [{ type: 'profile-email' }],
};

After (v2):

import { buildDpiMaskingProvider } from '@mymediset/sap-ai-provider';
const dpiMasking = buildDpiMaskingProvider({
method: 'anonymization',
entities: ['profile-email'],
});

Additional Resources