Project Setup

This topic walks through the Maven pom.xml BOM import, the provider starter dependencies, and the complete application.yml configuration needed to run Power RAG with all three LLM providers.

Spring AI BOM Import

Declare the Spring AI BOM inside <dependencyManagement>. This pins all Spring AI artifact versions so you never have to manage them individually.

<dependencyManagement>
  <dependencies>
    <dependency>
      <groupId>org.springframework.ai</groupId>
      <artifactId>spring-ai-bom</artifactId>
      <version>1.1.2</version>
      <type>pom</type>
      <scope>import</scope>
    </dependency>
  </dependencies>
</dependencyManagement>

Provider Starters

Add one starter per LLM provider you want to support. No version is needed — the BOM manages them.

<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-model-anthropic</artifactId>
</dependency>
<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-model-google-genai</artifactId>
</dependency>
<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-model-ollama</artifactId>
</dependency>

Each starter auto-configures a ChatModel bean for that provider. You then wrap those models in ChatClient beans — see Topic 05.

application.yml Configuration

The full AI-related configuration lives in application.yml. All sensitive values use the ${ENV_VAR:default} pattern — the real value comes from environment variables, with a safe default for local development.

spring:
  autoconfigure:
    exclude: org.springframework.ai.model.ollama.autoconfigure.OllamaEmbeddingAutoConfiguration
  ai:
    anthropic:
      api-key: ${ANTHROPIC_API_KEY:}
      chat:
        options:
          model: claude-sonnet-4-6
    google:
      genai:
        api-key: ${GOOGLE_API_KEY:}
        chat:
          options:
            model: gemini-2.5-flash
        embedding:
          text:
            options:
              model: gemini-embedding-001
              dimensions: ${powerrag.embedding.dimensions:768}
    ollama:
      base-url: ${OLLAMA_BASE_URL:http://localhost:11434}
      chat:
        options:
          model: qwen2.5-coder:7b
    vectorstore:
      qdrant:
        host: ${QDRANT_HOST:localhost}
        port: ${QDRANT_PORT:6334}
        collection-name: power_rag_docs
        dimensions: ${powerrag.embedding.dimensions:768}

powerrag:
  embedding:
    dimensions: ${POWERRAG_EMBEDDING_DIMENSIONS:768}
  guardrails:
    input-model-id: ${POWERRAG_GUARDRAIL_MODEL:gemini-2.5-flash}

Key Configuration Notes

Anthropic (Claude)

• Default model: claude-sonnet-4-6 — fast, capable, cost-effective
• max-tokens: 8096 — maximum output token budget per response
• temperature: 0.7 — balanced between creativity and determinism

Google GenAI (Gemini)

• Default chat model: gemini-2.5-flash
• Embeddings: gemini-embedding-001 at 768 dimensions for Qdrant and the semantic cache
• Input guardrails: default gemini-2.5-flash via powerrag.guardrails.input-model-id (Gemini 2.5 Flash — Google’s fast tier)

Ollama (Local Models)

• base-url defaults to http://localhost:11434
• Used for chat when the user selects Ollama (e.g. qwen2.5-coder:7b); embedding autoconfig is excluded

Qdrant Vector Store

• Port 6334 is the gRPC port used by Spring AI (not the REST port 6333)
• powerrag.embedding.dimensions (default 768, env POWERRAG_EMBEDDING_DIMENSIONS) must match Gemini embedding output and Qdrant
• collection-name: power_rag_docs — this collection is auto-created by Spring AI on startup