Skip to content

README.adoc

Latest commit

 

History

History
174 lines (115 loc) · 9.76 KB

README.adoc

File metadata and controls

174 lines (115 loc) · 9.76 KB
Raw

Spring Boot Example with Spring AI Agent

Introduction

This example demonstrates how to use Apache Camel with Spring AI to build AI-powered integration applications. The example showcases three key capabilities:

  1. Chat - Basic conversational AI using the Spring AI Chat component

  2. Tools - Function calling where the AI can invoke Camel routes as tools

  3. Vector Store - Semantic search using document embeddings with Qdrant

The example uses Ollama as the AI model provider with the granite4:3b model for chat and embeddinggemma:300m for embeddings, and Qdrant as the vector database.

Design Principles

This example follows Apache Camel Spring AI’s design philosophy of separating concerns:

  • Business Logic in Camel Routes - The route definitions focus purely on integration logic and data flow

  • Infrastructure Configuration in application.yaml - All non-functional requirements such as AI model selection, connection parameters, temperature settings, and external service endpoints are externalized to configuration files

This separation enables:

  • Easy switching between different AI providers (e.g., Ollama, OpenAI, Azure) without changing route code

  • Environment-specific configurations (dev, staging, production) with different models or endpoints

  • Clear distinction between what the integration does (routes) and how it connects (configuration)

Prerequisites

Camel CLI

Install Camel CLI to easily run infrastructure services:

$ curl -Ls https://sh.jbang.dev | bash -s - trust add https://github.com/apache/camel/
$ curl -Ls https://sh.jbang.dev | bash -s - app install --fresh --force camel@apache/camel

Or if you already have JBang installed:

$ jbang app install camel@apache/camel

Ollama

You need to have Ollama installed and running locally. Install Ollama from https://ollama.ai and then pull the required models:

$ ollama pull granite4:3b
$ ollama pull embeddinggemma:300m

The granite4:3b model is used for chat and the embeddinggemma:300m model is used for embeddings.

Make sure Ollama is running on http://localhost:11434 (default port) or update the application.yaml accordingly.

Note
 While you can use camel infra run ollama to run Ollama via Docker, be aware of performance limitations. On macOS, the Docker image cannot utilize GPU acceleration, making it extremely slow. Additionally, when running the Camel Ollama test-infra, both the Ollama Docker image and the default model (granite4, ~3GB) need to be downloaded on first run, which can be very time-consuming. For better performance, it’s recommended to install Ollama natively.

Qdrant

Use Camel CLI to run Qdrant:

$ camel infra run qdrant

Qdrant will be available on http://localhost:6333.

Components Used

  • camel-spring-ai-chat-starter - Camel Spring AI chat integration

  • camel-spring-ai-vector-store-starter - Camel Spring AI vector store integration

  • camel-spring-ai-tools-starter - Camel Spring AI tools/function calling

  • spring-ai-starter-model-ollama - Spring AI Ollama model support

  • spring-ai-starter-vector-store-qdrant - Qdrant vector store integration

Routes Overview

Note
 The routes use timer delays to ensure proper startup sequence. The chat route executes immediately on startup, the vector store query executes after a 10 second delay, and the tool chat route executes after a 20 second delay.

ChatRoute

Demonstrates basic chat functionality. Sends questions to the AI model and logs the responses. Executes immediately on startup.

ToolRoute

Shows how to define tools (functions) that the AI can call. The example includes a weather tool that the AI can invoke when asked about weather information. The tool is defined using the spring-ai-tools component with parameters. The tool chat route executes after a 20 second delay to ensure all services are ready.

VectorStoreRoute

Demonstrates semantic search using vector embeddings with Qdrant:

  • Watches the input directory for .txt files and adds them to the Qdrant vector store

  • Performs a sample similarity search query after a 10 second delay

  • Documents are automatically embedded and stored in Qdrant

  • Returns the top 3 most similar documents with a similarity threshold of 0.5

Build

You can build this example using:

$ mvn package

Run

You can run this example using:

$ mvn spring-boot:run

Usage

  1. The chat route will automatically execute and ask questions about Apache Camel

  2. The tool route will demonstrate AI function calling with the weather tool (executes after 20 seconds)

  3. The vector store route will:

    • Add sample documents from the input folder to the Qdrant vector store on startup

    • Execute a sample similarity search query after 10 seconds

    • You can add more .txt files to the input folder to expand the knowledge base

Expected Output

You should see in the console:

  • AI responses to the chat questions

  • Weather tool being called when the AI needs weather information

  • Vector store adding documents to Qdrant and performing a similarity search

  • The most relevant documents with similarity scores for the query

The example will continue running and watching the input directory for new documents. Press Ctrl+C to stop.

2025-11-25T17:24:21.770+01:00  INFO 51706 --- [ - timer://chat] chatRoute                                : Sending question to AI: What is Apache Camel?
2025-11-25T17:24:29.858+01:00  INFO 51706 --- [ - timer://chat] chatRoute                                : AI Response: Apache Camel is an open-source integration framework based on known Java connectivity standards...


2025-11-25T17:24:21.771+01:00  INFO 51706 --- [ - file://input] addDocuments                             : Adding document camel-intro.txt to vector store
2025-11-25T17:24:22.048+01:00  INFO 51706 --- [ - file://input] addDocuments                             : Successfully added camel-intro.txt to vector store
2025-11-25T17:24:22.048+01:00  INFO 51706 --- [ - file://input] addDocuments                             : Adding document spring-ai-intro.txt to vector store
2025-11-25T17:24:22.183+01:00  INFO 51706 --- [ - file://input] addDocuments                             : Successfully added spring-ai-intro.txt to vector store
2025-11-25T17:34:08.998+01:00  INFO 52622 --- [mer://queryDocs] queryDocuments                           : Querying vector store with: What is Apache Camel and what does it do?
2025-11-25T17:34:09.206+01:00  INFO 52622 --- [mer://queryDocs] queryDocuments                           : Retrieved 1 documents from vector store
2025-11-25T17:34:09.217+01:00  INFO 52622 --- [mer://queryDocs] queryDocuments                           : Document text: Apache Camel is an open-source integration framework based on Enterprise Integration Patterns. It provides a rule-based routing and mediation engine which enables you to define routing and mediation rules in various domain-specific languages.


2025-11-25T17:51:41.445+01:00  INFO 54308 --- [imer://toolChat] toolChatRoute                            : Sending question with tool capability to AI: What is the weather like in Rome, Italy? Answer with English slang
2025-11-25T17:51:43.520+01:00  INFO 54308 --- [imer://toolChat] weatherTool                              : Weather tool called for location: Rome
2025-11-25T17:51:43.739+01:00  INFO 54308 --- [imer://toolChat] weatherTool                              : Geocoding result - Latitude: 41.89193, Longitude: 12.51133
2025-11-25T17:51:43.858+01:00  INFO 54308 --- [imer://toolChat] weatherTool                              : Weather tool response: The weather in  is 12.4 degrees Celsius with 87% humidity, 100% cloud cover, wind speed 2.9 km/h from 7 degrees, and 0.2 mm precipitation.
2025-11-25T17:51:47.786+01:00  INFO 54308 --- [imer://toolChat] toolChatRoute                            : AI Response with tool usage: So, Rome's got it pretty mild out there right now – about **12.4°C**, feels like a cool breeze with that high humidity hanging around at **87%**. No rain in sight (just **0.2mm**), and the sky’s practically overcast (**100% cloud cover**) thanks to those gentle winds from the north‑west at **7°** – perfect for grabbing an espresso outside while you ponder life, love, or just how many more layers you need before it gets chilly again!

Configuration

The AI configuration can be modified in src/main/resources/application.yaml:

  • Change the Ollama base URL if running on a different host/port

  • Switch to a different model (ensure it’s pulled in Ollama first)

  • Adjust the temperature for more/less creative responses

  • Modify Qdrant host/port if running on a different location

  • Change the collection name for different use cases

  • Modify topK in VectorStoreRoute for more/fewer search results

Production Profile with OpenAI

The example includes a production profile that uses OpenAI instead of Ollama. To use it, activate the prod Maven profile and set the required environment variables:

export OPENAI_CHAT_MODEL=gpt-4o
export OPENAI_CHAT_BASE_URL=https://api.openai.com
export OPENAI_CHAT_API_KEY=your-openai-api-key
export OPENAI_EMBEDDING_MODEL=text-embedding-3-small
export OPENAI_EMBEDDING_BASE_URL=https://api.openai.com
export OPENAI_EMBEDDING_API_KEY=your-openai-api-key

mvn spring-boot:run -Pprod -Dspring-boot.run.profiles=prod

The production profile configuration is in src/main/resources/application-prod.yaml and uses environment variables for flexibility. You can customize the models and base URLs to use different OpenAI-compatible endpoints (e.g., Azure OpenAI, local OpenAI-compatible servers).

Help and contributions

If you hit any problem using Camel or have some feedback, then please let us know.

We also love contributors, so get involved :-)

The Camel riders!