Overview

Relevant source files

The SAP Cloud SDK for AI is the official Java SDK for SAP AI Core, SAP Generative AI Hub, and Orchestration Service. The SDK provides type-safe client classes for consuming foundation models, managing AI Core resources, and configuring orchestration features including templating, content filtering, data masking, and document grounding.

The codebase is a Maven multi-module project located at https://github.com/SAP/ai-sdk-java. The parent POM sdk-parent (pom.xml4-5) coordinates build configuration and dependency management across five SDK modules and one sample application module.

Key SDK Artifacts:

Group ID: com.sap.ai.sdk (core services)
Group ID: com.sap.ai.sdk.foundationmodels (OpenAI client)
Current Version: 1.15.0-SNAPSHOT (pom.xml6)
License: Apache 2.0 (pom.xml16-21)

Sources: pom.xml4-10 pom.xml16-21 README.md7-13

Purpose and Integration Patterns

The SDK enables three integration patterns for consuming AI services on SAP Business Technology Platform:

1. Orchestration Service Integration

Primary Class: OrchestrationClient (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationClient.java)
Configuration: OrchestrationModuleConfig builder pattern for pipeline modules
Capabilities: Templating, content filtering, data masking, document grounding, tool calling
Module Artifact: com.sap.ai.sdk:orchestration (orchestration/pom.xml9)

2. Direct Foundation Model Access

Primary Class: OpenAiClient (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiClient.java)
Model Catalog: OpenAiModel static constants (GPT-4o, Claude, Gemini, etc.)
Capabilities: Chat completion, streaming, embeddings, tool calling, vision
Module Artifact: com.sap.ai.sdk.foundationmodels:openai (foundation-models/openai/pom.xml11)

3. AI Core Resource Management

Primary Classes: DeploymentApi, ConfigurationApi, ScenarioApi (core/src/main/java/com/sap/ai/sdk/core/client/)
Infrastructure: AiCoreService for connectivity and deployment resolution (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java)
Capabilities: Artifact, configuration, and deployment lifecycle management
Module Artifact: com.sap.ai.sdk:core (core/pom.xml9)

4. Spring AI Framework Integration (Optional)

Implementations: OrchestrationChatModel, OpenAiChatModel, OrchestrationSpringEmbeddingModel
Spring AI Interfaces: ChatModel, EmbeddingModel, ChatOptions
Dependencies: Marked as <optional>true</optional> (orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119)

Runtime Requirements:

Java 17+ (compiler target: pom.xml52-54)
SAP Cloud SDK 5.25.0 (BOM: pom.xml58 pom.xml100-106)
Optional: Spring AI 1.1.2 (BOM: pom.xml67 pom.xml108-113)
Optional: Spring Boot 3.0+ for sample application (sample-code/spring-app/pom.xml36)

Sources: pom.xml52-67 pom.xml100-113 orchestration/pom.xml9-11 foundation-models/openai/pom.xml11-13 core/pom.xml9-11 orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119

Module Structure and Dependencies

SDK Module Architecture

Module Definitions

The parent POM (pom.xml31-37) declares the following modules:

Module Path	Artifact Coordinates	Key Classes/Packages	Purpose
`core/`	`com.sap.ai.sdk:core` (core/pom.xml9)	`AiCoreService` `DeploymentApi` `ConfigurationApi` `ScenarioApi` `DeploymentResolver`	AI Core connectivity, service binding resolution, deployment caching, generated OpenAPI clients from `aicore.yaml` (core/pom.xml187)
`orchestration/`	`com.sap.ai.sdk:orchestration` (orchestration/pom.xml9)	`OrchestrationClient` `OrchestrationModuleConfig` `OrchestrationPrompt` `OrchestrationChatModel`	Orchestration service client with templating, masking, filtering, grounding modules. Generated models from `orchestration.yaml` (orchestration/pom.xml199)
`foundation-models/openai/`	`com.sap.ai.sdk.foundationmodels:openai` (foundation-models/openai/pom.xml11)	`OpenAiClient` `OpenAiModel` `OpenAiChatModel` `OpenAiEmbeddingModel`	Direct OpenAI model access without orchestration. Generated models from `openapi-2024-10-21.yaml` (foundation-models/openai/pom.xml193)
`core-services/document-grounding/`	`com.sap.ai.sdk:document-grounding` (core-services/document-grounding/pom.xml10)	Generated API clients Generated models	Document grounding service client. Generated from `grounding.yaml` (core-services/document-grounding/pom.xml135)
`core-services/prompt-registry/`	`com.sap.ai.sdk:prompt-registry` (core-services/prompt-registry/pom.xml10)	Generated API clients Generated models	Prompt template management. Generated from `prompt-registry.yaml` (core-services/prompt-registry/pom.xml152)
`sample-code/spring-app/`	`com.sap.ai.sdk.app:spring-app` (sample-code/spring-app/pom.xml11)	`OrchestrationController` `OpenAiController` `DeploymentController` `SpringAiOrchestrationService`	Spring Boot reference application and E2E test suite. Not included in releases (pom.xml800-803)

Dependency Characteristics

Spring AI Integration: Dependencies marked <optional>true</optional> in orchestration (orchestration/pom.xml63-76) and OpenAI (foundation-models/openai/pom.xml106-119) modules
Code Generation: OpenAPI clients generated via SAP Cloud SDK plugin (pom.xml295-298) with profile -Pgenerate
Generated Code Exclusions: Model classes excluded from quality checks (pom.xml676-688) and marked @Beta (README.md56-58)

Sources: pom.xml31-37 pom.xml58 pom.xml67 pom.xml100-113 pom.xml295-298 pom.xml676-688 pom.xml800-803 core/pom.xml9 core/pom.xml187 orchestration/pom.xml9 orchestration/pom.xml63-76 orchestration/pom.xml199 foundation-models/openai/pom.xml11 foundation-models/openai/pom.xml106-119 foundation-models/openai/pom.xml193 core-services/document-grounding/pom.xml10 core-services/document-grounding/pom.xml135 core-services/prompt-registry/pom.xml10 core-services/prompt-registry/pom.xml152 sample-code/spring-app/pom.xml11

Core Components and Request Flow

AI Core Service Integration Architecture

API Layer Characteristics

The SDK provides three API patterns for different use cases:

1. High-Level Service Clients (Recommended for AI Tasks)

OrchestrationClient (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationClient.java):
- Methods: chatCompletion(OrchestrationPrompt, OrchestrationModuleConfig), streamChatCompletion(), embed()
- Requires: OrchestrationModuleConfig for pipeline configuration
- Returns: OrchestrationChatResponse, Stream<OrchestrationChatResponse>
OpenAiClient (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiClient.java):
- Factory: OpenAiClient.forModel(OpenAiModel.GPT_4O_MINI)
- Methods: chatCompletion(String), streamChatCompletion(), embedding(List<String>)
- Returns: OpenAiChatCompletionOutput, Stream<String>, OpenAiEmbeddingOutput

2. Spring AI Framework Adapters (Optional Integration)

OrchestrationChatModel (orchestration/src/main/java/com/sap/ai/sdk/orchestration/spring/OrchestrationChatModel.java):
- Implements: org.springframework.ai.chat.model.ChatModel
- Methods: call(Prompt), stream(Prompt)
- Converts: Spring AI Prompt → OrchestrationPrompt
OpenAiChatModel (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/spring/OpenAiChatModel.java):
- Implements: org.springframework.ai.chat.model.ChatModel
- Converts: Spring AI Message → OpenAI request format

3. Generated AI Core Management APIs

DeploymentApi (core/src/main/java/com/sap/ai/sdk/core/client/DeploymentApi.java):
- Generated from: core/src/main/resources/spec/aicore.yaml (core/pom.xml187)
- Methods: list(resourceGroup), create(resourceGroup, AiDeploymentCreationRequest), get(), delete()
ConfigurationApi (core/src/main/java/com/sap/ai/sdk/core/client/ConfigurationApi.java):
- Methods: query(resourceGroup), create(), get(), update()
ScenarioApi (core/src/main/java/com/sap/ai/sdk/core/client/ScenarioApi.java):
- Methods: query(), get()

Infrastructure Components

AiCoreService (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java): Central service coordinator
- Lazy initialization of service bindings
- Deployment cache management per resource group
- Methods: getBaseDestination(), getInferenceDestination(), withResourceGroup(), reloadCachedDeployments()
DestinationResolver (core/src/main/java/com/sap/ai/sdk/core/DestinationResolver.java): Service binding resolution
- Reads AICORE_SERVICE_KEY environment variable
- Builds HttpDestination with OAuth authentication
- Adds custom headers: AI-Client-Type: AI SDK Java, AI-Resource-Group: {resourceGroup}

Sources: orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationClient.java foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiClient.java core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java core/src/main/java/com/sap/ai/sdk/core/DestinationResolver.java core/src/main/java/com/sap/ai/sdk/core/client/ core/pom.xml187

Key Capabilities

Foundation Model Access

Model Catalog Access Points

The SDK provides foundation model access through two static constant classes:

1. OrchestrationAiModel Constants

Location: orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationAiModel.java
Purpose: Models accessible through Orchestration Service pipeline
Usage: OrchestrationModuleConfig.withLlmConfig(OrchestrationAiModel.GPT_4O)
Type: String constants
Versioning: Includes model version strings (e.g., gpt-4o-2024-08-06)

2. OpenAiModel Constants

Location: foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiModel.java
Purpose: Models for direct OpenAI client access
Usage: OpenAiClient.forModel(OpenAiModel.GPT_4O_MINI)
Type: String constants
Deprecation: Deprecated models marked with @Deprecated annotation

Supported Model Providers

Provider	Example Constants	Typical Use Case
OpenAI	`GPT_4O`, `GPT_4O_MINI`, `O1`, `O3_MINI`	General chat, reasoning tasks
Anthropic	`CLAUDE_4_SONNET`, `CLAUDE_4_5_SONNET`	Complex analysis, coding
Google	`GEMINI_2_5_FLASH`, `GEMINI_2_5_PRO`	Multimodal tasks
AWS	`NOVA_PRO`, `NOVA_PREMIER`	Enterprise workloads
Meta	`LLAMA_3_3_70B`, `LLAMA_3_2_90B_VISION`	Open models, vision
Mistral AI	`MISTRAL_LARGE_LATEST`, `MISTRAL_SMALL_LATEST`	European compliance
Cohere	`COMMAND_R_PLUS`, `COMMAND_R`	RAG, search
SAP	`ABAP_1`	SAP-specific tasks

Model Selection Guidelines

Chat completion: Use OrchestrationClient with OrchestrationAiModel for pipeline features
Direct access: Use OpenAiClient with OpenAiModel for minimal latency
Embeddings: Both clients support embedding models (text-embedding-3-small, text-embedding-3-large)

Complete model catalog with versioning information available at page 1.3 (Supported Models).

Sources: orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationAiModel.java foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiModel.java

Orchestration Pipeline Configuration

Processing Flow

OrchestrationModuleConfig Methods

The OrchestrationModuleConfig class (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationModuleConfig.java) provides fluent configuration:

Method	Return Type	Purpose
`withLlmConfig(OrchestrationAiModel)`	`OrchestrationModuleConfig`	Required: Set model and inference parameters
`withTemplateConfig(TemplatingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Configure prompt templates
`withMaskingConfig(DpiMaskingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Enable PII masking
`withInputFiltering(FilterConfig)`	`OrchestrationModuleConfig`	Optional: Apply input content filters
`withOutputFiltering(FilterConfig)`	`OrchestrationModuleConfig`	Optional: Apply output content filters
`withGrounding(GroundingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Inject context from vector DB
`withTranslationConfig(TranslationConfig)`	`OrchestrationModuleConfig`	Optional: Enable language translation

All configuration objects follow immutable builder pattern. Intermediate results from each module are accessible via OrchestrationChatResponse.getIntermediateResults() (returns Map<String, Object>). Detailed configuration covered in pages 5.4-5.9.

Sources: README.md129-132 orchestration/pom.xml10-11 docs/blog/Introducing.md56-76

Spring AI Integration

Optional Dependency Configuration

Spring AI dependencies are declared with <optional>true</optional> in:

orchestration/pom.xml63-76: spring-ai-model and spring-ai-commons
foundation-models/openai/pom.xml106-119: spring-ai-commons and spring-ai-model
pom.xml108-113: spring-ai-bom version 1.1.2 in dependency management

Implementation Classes

SDK Class	Spring AI Interface	Location
`OrchestrationChatModel`	`ChatModel`	`orchestration/.../spring/OrchestrationChatModel.java`
`OpenAiChatModel`	`ChatModel`	`foundation-models/openai/.../spring/OpenAiChatModel.java`
`OrchestrationSpringEmbeddingModel`	`EmbeddingModel`	`orchestration/.../spring/OrchestrationSpringEmbeddingModel.java`
`OrchestrationChatOptions`	`ChatOptions`	`orchestration/.../spring/OrchestrationChatOptions.java`
`DefaultToolCallingManager`	N/A (internal)	Handles `ToolCallback` conversion

The implementations delegate to underlying OrchestrationClient and OpenAiClient instances, converting between Spring AI data structures (Prompt, Message) and SDK request types (OrchestrationPrompt, OrchestrationModuleConfig). Page 8 documents Spring AI integration patterns.

Sources: pom.xml108-113 orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119

Getting Started

Maven Dependency Configuration

Dependency Management Import

The parent POM defines dependency versions in pom.xml98-259 Applications should import the SAP Cloud SDK BOM:

Module Dependencies

Select required modules based on use case:

Current version defined in pom.xml6 Released versions available on Maven Central at https://central.sonatype.com/namespace/com.sap.ai.sdk.

Sources: pom.xml98-113 pom.xml6 README.md74-83

Authentication Configuration

Environment Variable Setup

The SDK resolves AI Core credentials from the AICORE_SERVICE_KEY environment variable. The AiCoreService class (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java) uses SAP Cloud SDK's ServiceBindingAccessor to resolve this binding.

Credential Resolution Flow

In cloud environments, service bindings are automatically detected from VCAP_SERVICES. The destination resolution is lazy - credentials are only fetched when getBaseDestination() is first called. Alternative authentication approaches documented at page 2.2.

Sources: README.md112-126 core/pom.xml (dependencies on SAP Cloud SDK connectivity)

Basic Usage Examples

Orchestration Service Client

Minimal chat completion using OrchestrationClient:

OpenAI Direct Access

Using OpenAiClient without orchestration features:

Deployment Management

Using generated DeploymentApi:

Additional examples for streaming, tool calling, and module configuration available at pages 2.3, 5, and 6.

Sources: README.md86-109 docs/blog/Introducing.md32-103

Sample Application

Spring Boot Reference Implementation

The spring-app module (sample-code/spring-app/pom.xml11) serves dual purposes:

SDK Usage Examples: REST controllers demonstrating integration patterns (sample-code/spring-app/README.md1-6)
E2E Test Suite: JUnit tests validating functionality against live AI Core instances (sample-code/spring-app/README.md49-51)

Application Architecture

Module Configuration

Artifact: com.sap.ai.sdk.app:spring-app (sample-code/spring-app/pom.xml10-11)
Spring Boot: 3.5.9 (sample-code/spring-app/pom.xml36)
Test Execution: Skipped by default (<skipTests>true</skipTests>, sample-code/spring-app/pom.xml41), run with -DskipTests=false
Release Exclusion: Not included in release builds (pom.xml800-803)

Demonstrated Features

The OrchestrationService and SpringAiOrchestrationService classes demonstrate:

Templating with placeholder substitution
DPI masking (anonymization and pseudonymization)
Input/output content filtering
Document grounding with vector search
Tool calling with weather API and MCP
Chat memory with conversation history
Structured output with JSON schema
Streaming responses with SSE
Resource group isolation

Runtime Prerequisites

From sample-code/spring-app/README.md8-30:

Java 17+, Maven 3.8+
Environment variable: AICORE_SERVICE_KEY or .env file
Required deployments in default resource group:
- gpt-4o-mini model
- text-embedding-3-small embedding model

Running the Application

Install SDK to local Maven: mvn install -DskipTests from repository root
Start application: mvn spring-boot:run from sample-code/spring-app/
Access UI: http://localhost:8080
MCP profile: mvn spring-boot:run -Dspring-boot.run.arguments="--spring.profiles.active=mcp" (sample-code/spring-app/README.md47)

MCP configuration defined in sample-code/spring-app/src/main/resources/application.yml1-22

Detailed application documentation at page 8 (Sample Application).

Sources: sample-code/spring-app/pom.xml10-13 sample-code/spring-app/pom.xml36 sample-code/spring-app/pom.xml41 sample-code/spring-app/README.md1-52 sample-code/spring-app/src/main/resources/application.yml1-22 pom.xml800-803

Development Requirements

Compiler Configuration

Java compiler settings defined in pom.xml52-56:

Source/target: Java 17
Release: Java 17
Encoding: UTF-8 (pom.xml56)
Timestamp: 2025-04-03T13:23:00Z (pom.xml57)

Dependency Versions

Key versions from pom.xml58-74:

Dependency	Version	Property
SAP Cloud SDK	5.25.0	`cloud-sdk.version`
Spring AI	1.1.2	`spring-ai.version`
Spring Framework	6.2.15	`springframework.version`
JUnit Jupiter	6.0.1	`junit-jupiter.version`
Jackson	2.20.1	`jackson.version`

Quality Enforcement Plugins

Maven plugins configured in pom.xml606-789:

Plugin	Configuration	Enforcement Level
JaCoCo (pom.xml672-752)	Coverage ratios	100% for complexity, line, instruction, branch, method, class
Checkstyle (pom.xml607-635)	.pipeline/checkstyle.xml	Zero violations allowed
PMD (pom.xml636-670)	.pipeline/pmd.xml	Zero violations allowed
SpotBugs (pom.xml753-773)	.pipeline/spotbugs.xml	Zero violations allowed
Spotless (pom.xml459-491)	Google Java Format	Auto-fix in `process-sources` phase
OWASP Check (pom.xml776-789)	CVSS threshold 7.0	Fail build on high severity CVEs

Generated code excluded from quality checks via pom.xml676-688 Development setup and build process documented at page 10.

Sources: pom.xml50-96 pom.xml606-789 README.md32-52

Architecture Highlights

Lazy Initialization and Caching Strategy

AiCoreService Lifecycle

The AiCoreService class (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java) implements lazy initialization to optimize resource usage:

Method Behavior Matrix

Method Call	Resolves Service Binding	Loads Deployment List
`new AiCoreService()`	❌ No	❌ No
`withBaseDestination(HttpDestination)`	❌ No	❌ No
`withResourceGroup(String)`	❌ No	❌ No
`getBaseDestination()`	✅ Yes (lazy)	❌ No
`getInferenceDestination(model)`	✅ Yes	✅ Yes (cached)
`forModel(String)`	✅ Yes	✅ Yes (cached)
`reloadCachedDeployments()`	✅ Yes	✅ Yes (forced)

Deployment Cache Implementation

Cache Scope: Per resource group (core/src/main/java/com/sap/ai/sdk/core/DeploymentResolver.java)
Cache Expiry: 5 minutes (configurable)
Population Trigger: First call to getInferenceDestination() or forModel()
API Call: GET /v2/lm/deployments?resourceGroup={group}
Cache Key: Model name or scenario ID
Refresh: Call reloadCachedDeployments() to force cache invalidation

Immutability Pattern

AiCoreService follows immutable object design. Configuration methods return new instances:

Thread Safety Implications:

Safe to share AiCoreService instances across threads
Deployment cache is thread-safe (ConcurrentHashMap)
No mutable state after construction

Detailed architecture documentation at pages 3.2 (Connectivity & Service Integration) and 4.1 (AiCoreService).

Sources: core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java core/src/main/java/com/sap/ai/sdk/core/DeploymentResolver.java docs/adrs/004-core-class-for-service-integration.md70-112

Generated vs. Handwritten Code

OpenAPI Code Generation Strategy

The SDK uses SAP Cloud SDK's OpenAPI generator plugin (pom.xml295-298) to generate client code and model classes from OpenAPI specifications.

Generation Configuration per Module

Module	Spec File	Generated Packages	Generation Profile
core	`core/src/main/resources/spec/aicore.yaml` (core/pom.xml187)	`com.sap.ai.sdk.core.client.` `com.sap.ai.sdk.core.model.`	`-Pgenerate` (core/pom.xml160-209)
orchestration	`orchestration/src/main/resources/spec/orchestration.yaml` (orchestration/pom.xml199)	`com.sap.ai.sdk.orchestration.model.*`	`-Pgenerate` (orchestration/pom.xml172-220)
openai	`foundation-models/openai/src/main/resources/spec/openapi-2024-10-21.yaml` (foundation-models/openai/pom.xml193)	`com.sap.ai.sdk.foundationmodels.openai.generated.model.*`	`-Pgenerate` (foundation-models/openai/pom.xml166-223)
document-grounding	`core-services/document-grounding/src/main/resources/spec/grounding.yaml` (core-services/document-grounding/pom.xml135)	`com.sap.ai.sdk.grounding.client.` `com.sap.ai.sdk.grounding.model.`	`-Pgenerate` (core-services/document-grounding/pom.xml108-153)
prompt-registry	`core-services/prompt-registry/src/main/resources/spec/prompt-registry.yaml` (core-services/prompt-registry/pom.xml152)	`com.sap.ai.sdk.prompt.registry.client.` `com.sap.ai.sdk.prompt.registry.model.`	`-Pgenerate` (core-services/prompt-registry/pom.xml124-178)

Generation Command

Run from repository root: mvn clean install -Pgenerate

Quality Enforcement Exclusions

Generated code excluded from static analysis and test coverage:

PMD/Checkstyle Exclusions (pom.xml650-660 pom.xml612-614):

JaCoCo Coverage Exclusions (pom.xml676-688):

Same packages excluded from test coverage requirements
Allows 100% coverage enforcement for handwritten code only

Handwritten SDK Facade Layer

Handwritten classes provide developer-friendly APIs over generated models:

Handwritten Class	Purpose	Wraps Generated
`AiCoreService`	Service connectivity and deployment resolution	`DeploymentApi`, `ConfigurationApi`
`OrchestrationClient`	High-level orchestration client	`LLMModuleConfig`, `FilterConfig`, etc.
`OpenAiClient`	Direct model access facade	Generated OpenAI request/response models
`OrchestrationModuleConfig`	Builder for pipeline configuration	Multiple generated config classes
`OrchestrationChatModel`	Spring AI adapter	`OrchestrationClient` + generated models
`OpenAiChatModel`	Spring AI adapter	`OpenAiClient` + generated models

Generated Code Stability

Generated model classes marked @Beta (README.md56-58)
May change in minor releases without breaking semantic versioning
Handwritten facade classes provide stable API surface

OpenAPI generation process and workflow automation detailed at page 9.6 (API Specification Management).

Sources: pom.xml295-298 pom.xml612-614 pom.xml650-660 pom.xml676-688 README.md56-58 core/pom.xml160-209 core/pom.xml187 orchestration/pom.xml172-220 orchestration/pom.xml199 foundation-models/openai/pom.xml166-223 foundation-models/openai/pom.xml193 core-services/document-grounding/pom.xml108-153 core-services/document-grounding/pom.xml135 core-services/prompt-registry/pom.xml124-178 core-services/prompt-registry/pom.xml152

Release and Distribution

Distribution Channels

Channel	Purpose	URL
Maven Central	Public releases	`https://central.sonatype.com/namespace/com.sap.ai.sdk`
SAP Artifactory	Daily snapshots	`https://common.repositories.cloud.sap/artifactory/build-snapshots-cloudsdk/`

Snapshot Repository Configuration

Defined in pom.xml43-48:

Release Workflow Overview

The release process is split into two GitHub Actions workflows:

prepare-release.yaml: Version bump, branch creation, draft release
perform-release.yaml: Artifact signing, Maven Central deployment, GitHub release publication

Release Profile Configuration

Release-specific plugins configured in pom.xml804-960:

maven-source-plugin: Attaches source JARs (pom.xml815-828)
lombok-maven-plugin: Delomboks source for JavaDoc (pom.xml830-856)
maven-javadoc-plugin: Generates JavaDoc JARs (pom.xml858-875)
maven-gpg-plugin: Signs artifacts (pom.xml877-889)
central-publishing-maven-plugin: Deploys to Maven Central (pom.xml922-931)

Complete release process documentation at page 10.5.

Sources: pom.xml43-48 pom.xml804-960 README.md155-157

Next Steps

Getting Started: Follow the quickstart guide at page 2
Module Details: Explore individual module documentation:
- AI Core Management: page 4
- Orchestration Service: page 5
- OpenAI Models: page 6
- Prompt Registry: page 7
- Spring AI Integration: page 8
Sample Code: Review the Spring Boot sample application at page 9
Contributing: See development setup and guidelines at page 10
Model Reference: Complete model listings at page 1.3

Sources: README.md134-145

Overview

Relevant source files

Key SDK Artifacts:

Group ID: com.sap.ai.sdk (core services)
Group ID: com.sap.ai.sdk.foundationmodels (OpenAI client)
Current Version: 1.15.0-SNAPSHOT (pom.xml6)
License: Apache 2.0 (pom.xml16-21)

Sources: pom.xml4-10 pom.xml16-21 README.md7-13

Purpose and Integration Patterns

The SDK enables three integration patterns for consuming AI services on SAP Business Technology Platform:

1. Orchestration Service Integration

Primary Class: OrchestrationClient (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationClient.java)
Configuration: OrchestrationModuleConfig builder pattern for pipeline modules
Capabilities: Templating, content filtering, data masking, document grounding, tool calling
Module Artifact: com.sap.ai.sdk:orchestration (orchestration/pom.xml9)

2. Direct Foundation Model Access

Primary Class: OpenAiClient (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiClient.java)
Model Catalog: OpenAiModel static constants (GPT-4o, Claude, Gemini, etc.)
Capabilities: Chat completion, streaming, embeddings, tool calling, vision
Module Artifact: com.sap.ai.sdk.foundationmodels:openai (foundation-models/openai/pom.xml11)

3. AI Core Resource Management

Primary Classes: DeploymentApi, ConfigurationApi, ScenarioApi (core/src/main/java/com/sap/ai/sdk/core/client/)
Infrastructure: AiCoreService for connectivity and deployment resolution (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java)
Capabilities: Artifact, configuration, and deployment lifecycle management
Module Artifact: com.sap.ai.sdk:core (core/pom.xml9)

4. Spring AI Framework Integration (Optional)

Implementations: OrchestrationChatModel, OpenAiChatModel, OrchestrationSpringEmbeddingModel
Spring AI Interfaces: ChatModel, EmbeddingModel, ChatOptions
Dependencies: Marked as <optional>true</optional> (orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119)

Runtime Requirements:

Java 17+ (compiler target: pom.xml52-54)
SAP Cloud SDK 5.25.0 (BOM: pom.xml58 pom.xml100-106)
Optional: Spring AI 1.1.2 (BOM: pom.xml67 pom.xml108-113)
Optional: Spring Boot 3.0+ for sample application (sample-code/spring-app/pom.xml36)

Sources: pom.xml52-67 pom.xml100-113 orchestration/pom.xml9-11 foundation-models/openai/pom.xml11-13 core/pom.xml9-11 orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119

Module Structure and Dependencies

SDK Module Architecture

Module Definitions

The parent POM (pom.xml31-37) declares the following modules:

Module Path	Artifact Coordinates	Key Classes/Packages	Purpose
`core/`	`com.sap.ai.sdk:core` (core/pom.xml9)	`AiCoreService` `DeploymentApi` `ConfigurationApi` `ScenarioApi` `DeploymentResolver`	AI Core connectivity, service binding resolution, deployment caching, generated OpenAPI clients from `aicore.yaml` (core/pom.xml187)
`orchestration/`	`com.sap.ai.sdk:orchestration` (orchestration/pom.xml9)	`OrchestrationClient` `OrchestrationModuleConfig` `OrchestrationPrompt` `OrchestrationChatModel`	Orchestration service client with templating, masking, filtering, grounding modules. Generated models from `orchestration.yaml` (orchestration/pom.xml199)
`foundation-models/openai/`	`com.sap.ai.sdk.foundationmodels:openai` (foundation-models/openai/pom.xml11)	`OpenAiClient` `OpenAiModel` `OpenAiChatModel` `OpenAiEmbeddingModel`	Direct OpenAI model access without orchestration. Generated models from `openapi-2024-10-21.yaml` (foundation-models/openai/pom.xml193)
`core-services/document-grounding/`	`com.sap.ai.sdk:document-grounding` (core-services/document-grounding/pom.xml10)	Generated API clients Generated models	Document grounding service client. Generated from `grounding.yaml` (core-services/document-grounding/pom.xml135)
`core-services/prompt-registry/`	`com.sap.ai.sdk:prompt-registry` (core-services/prompt-registry/pom.xml10)	Generated API clients Generated models	Prompt template management. Generated from `prompt-registry.yaml` (core-services/prompt-registry/pom.xml152)
`sample-code/spring-app/`	`com.sap.ai.sdk.app:spring-app` (sample-code/spring-app/pom.xml11)	`OrchestrationController` `OpenAiController` `DeploymentController` `SpringAiOrchestrationService`	Spring Boot reference application and E2E test suite. Not included in releases (pom.xml800-803)

Dependency Characteristics

Spring AI Integration: Dependencies marked <optional>true</optional> in orchestration (orchestration/pom.xml63-76) and OpenAI (foundation-models/openai/pom.xml106-119) modules
Code Generation: OpenAPI clients generated via SAP Cloud SDK plugin (pom.xml295-298) with profile -Pgenerate
Generated Code Exclusions: Model classes excluded from quality checks (pom.xml676-688) and marked @Beta (README.md56-58)

Core Components and Request Flow

AI Core Service Integration Architecture

API Layer Characteristics

The SDK provides three API patterns for different use cases:

1. High-Level Service Clients (Recommended for AI Tasks)

OrchestrationClient (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationClient.java):
- Methods: chatCompletion(OrchestrationPrompt, OrchestrationModuleConfig), streamChatCompletion(), embed()
- Requires: OrchestrationModuleConfig for pipeline configuration
- Returns: OrchestrationChatResponse, Stream<OrchestrationChatResponse>
OpenAiClient (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiClient.java):
- Factory: OpenAiClient.forModel(OpenAiModel.GPT_4O_MINI)
- Methods: chatCompletion(String), streamChatCompletion(), embedding(List<String>)
- Returns: OpenAiChatCompletionOutput, Stream<String>, OpenAiEmbeddingOutput

2. Spring AI Framework Adapters (Optional Integration)

OrchestrationChatModel (orchestration/src/main/java/com/sap/ai/sdk/orchestration/spring/OrchestrationChatModel.java):
- Implements: org.springframework.ai.chat.model.ChatModel
- Methods: call(Prompt), stream(Prompt)
- Converts: Spring AI Prompt → OrchestrationPrompt
OpenAiChatModel (foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/spring/OpenAiChatModel.java):
- Implements: org.springframework.ai.chat.model.ChatModel
- Converts: Spring AI Message → OpenAI request format

3. Generated AI Core Management APIs

DeploymentApi (core/src/main/java/com/sap/ai/sdk/core/client/DeploymentApi.java):
- Generated from: core/src/main/resources/spec/aicore.yaml (core/pom.xml187)
- Methods: list(resourceGroup), create(resourceGroup, AiDeploymentCreationRequest), get(), delete()
ConfigurationApi (core/src/main/java/com/sap/ai/sdk/core/client/ConfigurationApi.java):
- Methods: query(resourceGroup), create(), get(), update()
ScenarioApi (core/src/main/java/com/sap/ai/sdk/core/client/ScenarioApi.java):
- Methods: query(), get()

Infrastructure Components

AiCoreService (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java): Central service coordinator
- Lazy initialization of service bindings
- Deployment cache management per resource group
- Methods: getBaseDestination(), getInferenceDestination(), withResourceGroup(), reloadCachedDeployments()
DestinationResolver (core/src/main/java/com/sap/ai/sdk/core/DestinationResolver.java): Service binding resolution
- Reads AICORE_SERVICE_KEY environment variable
- Builds HttpDestination with OAuth authentication
- Adds custom headers: AI-Client-Type: AI SDK Java, AI-Resource-Group: {resourceGroup}

Key Capabilities

Foundation Model Access

Model Catalog Access Points

The SDK provides foundation model access through two static constant classes:

1. OrchestrationAiModel Constants

Location: orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationAiModel.java
Purpose: Models accessible through Orchestration Service pipeline
Usage: OrchestrationModuleConfig.withLlmConfig(OrchestrationAiModel.GPT_4O)
Type: String constants
Versioning: Includes model version strings (e.g., gpt-4o-2024-08-06)

2. OpenAiModel Constants

Location: foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiModel.java
Purpose: Models for direct OpenAI client access
Usage: OpenAiClient.forModel(OpenAiModel.GPT_4O_MINI)
Type: String constants
Deprecation: Deprecated models marked with @Deprecated annotation

Supported Model Providers

Provider	Example Constants	Typical Use Case
OpenAI	`GPT_4O`, `GPT_4O_MINI`, `O1`, `O3_MINI`	General chat, reasoning tasks
Anthropic	`CLAUDE_4_SONNET`, `CLAUDE_4_5_SONNET`	Complex analysis, coding
Google	`GEMINI_2_5_FLASH`, `GEMINI_2_5_PRO`	Multimodal tasks
AWS	`NOVA_PRO`, `NOVA_PREMIER`	Enterprise workloads
Meta	`LLAMA_3_3_70B`, `LLAMA_3_2_90B_VISION`	Open models, vision
Mistral AI	`MISTRAL_LARGE_LATEST`, `MISTRAL_SMALL_LATEST`	European compliance
Cohere	`COMMAND_R_PLUS`, `COMMAND_R`	RAG, search
SAP	`ABAP_1`	SAP-specific tasks

Model Selection Guidelines

Chat completion: Use OrchestrationClient with OrchestrationAiModel for pipeline features
Direct access: Use OpenAiClient with OpenAiModel for minimal latency
Embeddings: Both clients support embedding models (text-embedding-3-small, text-embedding-3-large)

Complete model catalog with versioning information available at page 1.3 (Supported Models).

Sources: orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationAiModel.java foundation-models/openai/src/main/java/com/sap/ai/sdk/foundationmodels/openai/OpenAiModel.java

Orchestration Pipeline Configuration

Processing Flow

OrchestrationModuleConfig Methods

The OrchestrationModuleConfig class (orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationModuleConfig.java) provides fluent configuration:

Method	Return Type	Purpose
`withLlmConfig(OrchestrationAiModel)`	`OrchestrationModuleConfig`	Required: Set model and inference parameters
`withTemplateConfig(TemplatingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Configure prompt templates
`withMaskingConfig(DpiMaskingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Enable PII masking
`withInputFiltering(FilterConfig)`	`OrchestrationModuleConfig`	Optional: Apply input content filters
`withOutputFiltering(FilterConfig)`	`OrchestrationModuleConfig`	Optional: Apply output content filters
`withGrounding(GroundingModuleConfig)`	`OrchestrationModuleConfig`	Optional: Inject context from vector DB
`withTranslationConfig(TranslationConfig)`	`OrchestrationModuleConfig`	Optional: Enable language translation

Sources: README.md129-132 orchestration/pom.xml10-11 docs/blog/Introducing.md56-76

Spring AI Integration

Optional Dependency Configuration

Spring AI dependencies are declared with <optional>true</optional> in:

orchestration/pom.xml63-76: spring-ai-model and spring-ai-commons
foundation-models/openai/pom.xml106-119: spring-ai-commons and spring-ai-model
pom.xml108-113: spring-ai-bom version 1.1.2 in dependency management

Implementation Classes

SDK Class	Spring AI Interface	Location
`OrchestrationChatModel`	`ChatModel`	`orchestration/.../spring/OrchestrationChatModel.java`
`OpenAiChatModel`	`ChatModel`	`foundation-models/openai/.../spring/OpenAiChatModel.java`
`OrchestrationSpringEmbeddingModel`	`EmbeddingModel`	`orchestration/.../spring/OrchestrationSpringEmbeddingModel.java`
`OrchestrationChatOptions`	`ChatOptions`	`orchestration/.../spring/OrchestrationChatOptions.java`
`DefaultToolCallingManager`	N/A (internal)	Handles `ToolCallback` conversion

Sources: pom.xml108-113 orchestration/pom.xml63-76 foundation-models/openai/pom.xml106-119

Getting Started

Maven Dependency Configuration

Dependency Management Import

The parent POM defines dependency versions in pom.xml98-259 Applications should import the SAP Cloud SDK BOM:

Module Dependencies

Select required modules based on use case:

Current version defined in pom.xml6 Released versions available on Maven Central at https://central.sonatype.com/namespace/com.sap.ai.sdk.

Sources: pom.xml98-113 pom.xml6 README.md74-83

Authentication Configuration

Environment Variable Setup

Credential Resolution Flow

Sources: README.md112-126 core/pom.xml (dependencies on SAP Cloud SDK connectivity)

Basic Usage Examples

Orchestration Service Client

Minimal chat completion using OrchestrationClient:

OpenAI Direct Access

Using OpenAiClient without orchestration features:

Deployment Management

Using generated DeploymentApi:

Additional examples for streaming, tool calling, and module configuration available at pages 2.3, 5, and 6.

Sources: README.md86-109 docs/blog/Introducing.md32-103

Sample Application

Spring Boot Reference Implementation

The spring-app module (sample-code/spring-app/pom.xml11) serves dual purposes:

SDK Usage Examples: REST controllers demonstrating integration patterns (sample-code/spring-app/README.md1-6)
E2E Test Suite: JUnit tests validating functionality against live AI Core instances (sample-code/spring-app/README.md49-51)

Application Architecture

Module Configuration

Artifact: com.sap.ai.sdk.app:spring-app (sample-code/spring-app/pom.xml10-11)
Spring Boot: 3.5.9 (sample-code/spring-app/pom.xml36)
Test Execution: Skipped by default (<skipTests>true</skipTests>, sample-code/spring-app/pom.xml41), run with -DskipTests=false
Release Exclusion: Not included in release builds (pom.xml800-803)

Demonstrated Features

The OrchestrationService and SpringAiOrchestrationService classes demonstrate:

Templating with placeholder substitution
DPI masking (anonymization and pseudonymization)
Input/output content filtering
Document grounding with vector search
Tool calling with weather API and MCP
Chat memory with conversation history
Structured output with JSON schema
Streaming responses with SSE
Resource group isolation

Runtime Prerequisites

From sample-code/spring-app/README.md8-30:

Java 17+, Maven 3.8+
Environment variable: AICORE_SERVICE_KEY or .env file
Required deployments in default resource group:
- gpt-4o-mini model
- text-embedding-3-small embedding model

Running the Application

Install SDK to local Maven: mvn install -DskipTests from repository root
Start application: mvn spring-boot:run from sample-code/spring-app/
Access UI: http://localhost:8080
MCP profile: mvn spring-boot:run -Dspring-boot.run.arguments="--spring.profiles.active=mcp" (sample-code/spring-app/README.md47)

MCP configuration defined in sample-code/spring-app/src/main/resources/application.yml1-22

Detailed application documentation at page 8 (Sample Application).

Development Requirements

Compiler Configuration

Java compiler settings defined in pom.xml52-56:

Source/target: Java 17
Release: Java 17
Encoding: UTF-8 (pom.xml56)
Timestamp: 2025-04-03T13:23:00Z (pom.xml57)

Dependency Versions

Key versions from pom.xml58-74:

Dependency	Version	Property
SAP Cloud SDK	5.25.0	`cloud-sdk.version`
Spring AI	1.1.2	`spring-ai.version`
Spring Framework	6.2.15	`springframework.version`
JUnit Jupiter	6.0.1	`junit-jupiter.version`
Jackson	2.20.1	`jackson.version`

Quality Enforcement Plugins

Maven plugins configured in pom.xml606-789:

Plugin	Configuration	Enforcement Level
JaCoCo (pom.xml672-752)	Coverage ratios	100% for complexity, line, instruction, branch, method, class
Checkstyle (pom.xml607-635)	.pipeline/checkstyle.xml	Zero violations allowed
PMD (pom.xml636-670)	.pipeline/pmd.xml	Zero violations allowed
SpotBugs (pom.xml753-773)	.pipeline/spotbugs.xml	Zero violations allowed
Spotless (pom.xml459-491)	Google Java Format	Auto-fix in `process-sources` phase
OWASP Check (pom.xml776-789)	CVSS threshold 7.0	Fail build on high severity CVEs

Generated code excluded from quality checks via pom.xml676-688 Development setup and build process documented at page 10.

Sources: pom.xml50-96 pom.xml606-789 README.md32-52

Architecture Highlights

Lazy Initialization and Caching Strategy

AiCoreService Lifecycle

The AiCoreService class (core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java) implements lazy initialization to optimize resource usage:

Method Behavior Matrix

Method Call	Resolves Service Binding	Loads Deployment List
`new AiCoreService()`	❌ No	❌ No
`withBaseDestination(HttpDestination)`	❌ No	❌ No
`withResourceGroup(String)`	❌ No	❌ No
`getBaseDestination()`	✅ Yes (lazy)	❌ No
`getInferenceDestination(model)`	✅ Yes	✅ Yes (cached)
`forModel(String)`	✅ Yes	✅ Yes (cached)
`reloadCachedDeployments()`	✅ Yes	✅ Yes (forced)

Deployment Cache Implementation

Cache Scope: Per resource group (core/src/main/java/com/sap/ai/sdk/core/DeploymentResolver.java)
Cache Expiry: 5 minutes (configurable)
Population Trigger: First call to getInferenceDestination() or forModel()
API Call: GET /v2/lm/deployments?resourceGroup={group}
Cache Key: Model name or scenario ID
Refresh: Call reloadCachedDeployments() to force cache invalidation

Immutability Pattern

AiCoreService follows immutable object design. Configuration methods return new instances:

Thread Safety Implications:

Safe to share AiCoreService instances across threads
Deployment cache is thread-safe (ConcurrentHashMap)
No mutable state after construction

Detailed architecture documentation at pages 3.2 (Connectivity & Service Integration) and 4.1 (AiCoreService).

Sources: core/src/main/java/com/sap/ai/sdk/core/AiCoreService.java core/src/main/java/com/sap/ai/sdk/core/DeploymentResolver.java docs/adrs/004-core-class-for-service-integration.md70-112

Generated vs. Handwritten Code

OpenAPI Code Generation Strategy

The SDK uses SAP Cloud SDK's OpenAPI generator plugin (pom.xml295-298) to generate client code and model classes from OpenAPI specifications.

Generation Configuration per Module

Module	Spec File	Generated Packages	Generation Profile
core	`core/src/main/resources/spec/aicore.yaml` (core/pom.xml187)	`com.sap.ai.sdk.core.client.` `com.sap.ai.sdk.core.model.`	`-Pgenerate` (core/pom.xml160-209)
orchestration	`orchestration/src/main/resources/spec/orchestration.yaml` (orchestration/pom.xml199)	`com.sap.ai.sdk.orchestration.model.*`	`-Pgenerate` (orchestration/pom.xml172-220)
openai	`foundation-models/openai/src/main/resources/spec/openapi-2024-10-21.yaml` (foundation-models/openai/pom.xml193)	`com.sap.ai.sdk.foundationmodels.openai.generated.model.*`	`-Pgenerate` (foundation-models/openai/pom.xml166-223)
document-grounding	`core-services/document-grounding/src/main/resources/spec/grounding.yaml` (core-services/document-grounding/pom.xml135)	`com.sap.ai.sdk.grounding.client.` `com.sap.ai.sdk.grounding.model.`	`-Pgenerate` (core-services/document-grounding/pom.xml108-153)
prompt-registry	`core-services/prompt-registry/src/main/resources/spec/prompt-registry.yaml` (core-services/prompt-registry/pom.xml152)	`com.sap.ai.sdk.prompt.registry.client.` `com.sap.ai.sdk.prompt.registry.model.`	`-Pgenerate` (core-services/prompt-registry/pom.xml124-178)

Generation Command

Run from repository root: mvn clean install -Pgenerate

Quality Enforcement Exclusions

Generated code excluded from static analysis and test coverage:

PMD/Checkstyle Exclusions (pom.xml650-660 pom.xml612-614):

JaCoCo Coverage Exclusions (pom.xml676-688):

Same packages excluded from test coverage requirements
Allows 100% coverage enforcement for handwritten code only

Handwritten SDK Facade Layer

Handwritten classes provide developer-friendly APIs over generated models:

Handwritten Class	Purpose	Wraps Generated
`AiCoreService`	Service connectivity and deployment resolution	`DeploymentApi`, `ConfigurationApi`
`OrchestrationClient`	High-level orchestration client	`LLMModuleConfig`, `FilterConfig`, etc.
`OpenAiClient`	Direct model access facade	Generated OpenAI request/response models
`OrchestrationModuleConfig`	Builder for pipeline configuration	Multiple generated config classes
`OrchestrationChatModel`	Spring AI adapter	`OrchestrationClient` + generated models
`OpenAiChatModel`	Spring AI adapter	`OpenAiClient` + generated models

Generated Code Stability

Generated model classes marked @Beta (README.md56-58)
May change in minor releases without breaking semantic versioning
Handwritten facade classes provide stable API surface

OpenAPI generation process and workflow automation detailed at page 9.6 (API Specification Management).

Release and Distribution

Distribution Channels

Channel	Purpose	URL
Maven Central	Public releases	`https://central.sonatype.com/namespace/com.sap.ai.sdk`
SAP Artifactory	Daily snapshots	`https://common.repositories.cloud.sap/artifactory/build-snapshots-cloudsdk/`

Snapshot Repository Configuration

Defined in pom.xml43-48:

Release Workflow Overview

The release process is split into two GitHub Actions workflows:

prepare-release.yaml: Version bump, branch creation, draft release
perform-release.yaml: Artifact signing, Maven Central deployment, GitHub release publication

Release Profile Configuration

Release-specific plugins configured in pom.xml804-960:

maven-source-plugin: Attaches source JARs (pom.xml815-828)
lombok-maven-plugin: Delomboks source for JavaDoc (pom.xml830-856)
maven-javadoc-plugin: Generates JavaDoc JARs (pom.xml858-875)
maven-gpg-plugin: Signs artifacts (pom.xml877-889)
central-publishing-maven-plugin: Deploys to Maven Central (pom.xml922-931)

Complete release process documentation at page 10.5.

Sources: pom.xml43-48 pom.xml804-960 README.md155-157

Next Steps

Getting Started: Follow the quickstart guide at page 2
Module Details: Explore individual module documentation:
- AI Core Management: page 4
- Orchestration Service: page 5
- OpenAI Models: page 6
- Prompt Registry: page 7
- Spring AI Integration: page 8
Sample Code: Review the Spring Boot sample application at page 9
Contributing: See development setup and guidelines at page 10
Model Reference: Complete model listings at page 1.3

Sources: README.md134-145

Overview

Purpose and Integration Patterns

Module Structure and Dependencies

Core Components and Request Flow

Key Capabilities

Foundation Model Access

Orchestration Pipeline Configuration

Spring AI Integration

Getting Started

Maven Dependency Configuration

Authentication Configuration

Basic Usage Examples

Sample Application

Development Requirements

Architecture Highlights

Lazy Initialization and Caching Strategy

Generated vs. Handwritten Code

Release and Distribution

Next Steps

On this page

Overview

Purpose and Integration Patterns

Module Structure and Dependencies

Core Components and Request Flow

Key Capabilities

Foundation Model Access

Orchestration Pipeline Configuration

Spring AI Integration

Getting Started

Maven Dependency Configuration

Authentication Configuration

Basic Usage Examples

Sample Application

Development Requirements

Architecture Highlights

Lazy Initialization and Caching Strategy

Generated vs. Handwritten Code

Release and Distribution

Next Steps

On this page