A minimal Spring Boot web app with a browser chat window backed by an ADK agent. The agent calls its model through Spring AI, using the ADK Spring AI integration (google-adk-spring-ai).

• spring-ai-starter-model-google-genai creates a Spring AI ChatModel.
• The ADK Spring AI auto-configuration wraps it in a SpringAI bean, which is an ADK BaseLlm.
• AgentConfig builds an LlmAgent on that model and an InMemoryRunner.
• ChatController exposes POST /api/chat, and static/index.html is the chat page. Each page load gets its own ADK session, so the agent remembers the conversation until you reload. Sessions are kept in memory until the app stops.

To use another provider, replace the Google GenAI starter in pom.xml with any other Spring AI chat model starter (OpenAI, Anthropic, Ollama, ...) and set its spring.ai.* properties.

• Java 17+
• A Google Cloud project with Vertex AI enabled, and Application Default Credentials (gcloud auth application-default login).

The app uses Vertex AI by default. Set GOOGLE_CLOUD_PROJECT, and optionally GOOGLE_CLOUD_LOCATION (default global) and GEMINI_MODEL. To use the Gemini Developer API instead, set GOOGLE_GENAI_USE_VERTEXAI=false and GOOGLE_API_KEY.

From the repository root:

export GOOGLE_CLOUD_PROJECT=your-project-id
./mvnw install -DskipTests -pl examples/spring_chat -am
./mvnw spring-boot:run -pl examples/spring_chat

Then open http://localhost:8080. The chat endpoint has no authentication, so the app listens on localhost only, which keeps other machines out. Set SERVER_ADDRESS=0.0.0.0 to accept connections from other hosts. Don't expose it beyond a trusted environment without adding authentication.