spring-ai-zero-to-hero

Example applications showing how to use Spring AI to build Generative AI projects.

Software Prerequisites

You need the following software installed: Java 21, docker, ollama, httpie, ard your favourite Java IDE. This is a lot of GBs to download so please make sure to have all this stuff installed before the conference workshop, as the conference wifi may be slow, so you might not be able to run the samples.

Java development tooling

• Java 21 you can use sdkman.io
• Maven
• Favourite Java IDE one of
  • IntelliJ
  • VSCode
  • Eclipse Spring Tool Suite

Http Client

• Command line http client httpie is recommended, the instructions use it, if you don't have it please install it. If you are handy with curl you can use that too.

Containerization tools

• Docker so we can use test containers & for local dependencies

Local AI Models

ollama makes running models on your laptop easy and very educational. You can run the models locally and learn how they work.

• Install ollama by following the instructions on the ollama website this YouTube video shows the ollama install process.

Save the conference Wi-Fi

Please make sure that the software list above is installed on your laptop before the workshop starts. After install:

1. clone this repo to your laptop
2. Run the ./download-deps.sh script pull local AI models, and container images.
3. Run the check-deps.sh script to check that the all the required software is installed, the output of the script on my machine looks like.

./check-deps.sh
============================
Checking Java installation:
============================
Java is installed. Version details:
openjdk version "21.0.4" 2024-07-16 LTS
OpenJDK Runtime Environment (build 21.0.4+9-LTS)
OpenJDK 64-Bit Server VM (build 21.0.4+9-LTS, mixed mode, sharing)

===============================
Checking Ollama installation:
===============================
Ollama is installed. Version details:
ollama version is 0.3.10

========================================
Checking if llama3.1 model is pulled:
========================================
llama3.1 model is pulled and available.

==============================
Checking Docker installation:
==============================
Docker is installed. Version details:
Docker version 27.1.1, build 6312585

Checking Docker image: pgvector/pgvector:pg16
Docker image pgvector/pgvector:pg16 is pulled.

Checking Docker image: dpage/pgadmin4:8.6
Docker image dpage/pgadmin4:8.6 is pulled.

===============================
Checking HTTPie installation:
===============================
HTTPie is installed. Version details:
3.2.3

if you run into issues try running the commands in the check-deps.sh script one at a time.

API Keys

You will be provided with API keys for online AI services during the workshop, these keys will only be valid during the workshop. Highly recommend you get your own keys to continue experimenting after the workshop.

OpenAI

• You need OpenAI API key to run the examples with OpenAI.
• Refer to this page to get an API key.

Azure OpenAI

• You need Azure OpenAI service instance in the Azure portal. This requires to fill out a form at the moment, which usually takes at most 24h to process.
• Create the service at https://portal.azure.com/#create/Microsoft.CognitiveServicesOpenAI
• Create an Azure OpenAI deployment at https://oai.azure.com/portal

Outline

Generative AI is a transformational technology impacting our world in profound ways and creating unprecedented opportunities. This workshop is designed for Spring developers looking to add generative AI to existing applications or to implement brand new AI apps using the Spring AI project.

We assume no previous AI experience. The workshop will teach you key AI concepts and how to apply them in your applications, using the Spring AI project.

The workshop is hands-on. Bring your laptop and a willingness to learn. We will provide Spring AI based sample code and the API keys for the AI services. By the end of the day you will know how to add generative AI features to your Spring apps.

Key Concepts covered:

• A Concise History of AI/ML
• Introduction to Generative AI Models
• Prompt Engineering Techniques & Best Practices
• Vector Databases
• RAG: Retrieval Augmented Generation
• Extending LLMs with Function Calling
• Evaluation: How to tell if the AI is doing what you think it should be doing
• Architecture of AI powered applications
• How to integrate AI into existing applications
• The landscape of tools and libraries of building AI powered applications

Hands on Code Exercises with Spring AI:

• Quickstart: Creating a “Hello World” application for Generative AI in just minutes
• Prompt Engineering Techniques using Prompt Templating and Roles
• Mapping AI output to POJOs
• Implementing RAG (Retrieval Augmented Generation)
• Exploring Function calling: Enable the AI to access APIs on demand
• Evaluation Driven Development
• Using Spring AI with GraalVM

Repo Organization

Spring AI provides a consistent API to work with many different types of AI providers. For example, the same code wil work with OpenAI, Google Vertex AI, Azure OpenAI, and local AI models. The major directories in this repo are:

• /components/data/ this directory contains various types of example data sets used by the examples in the repo.

• /components/api/ this directory contains the code that interacts with the AI providers. The code in this directory is the same for all the AI providers. Each project in this directory focuses on a different aspect of the Spring AI API, within a project you will see that the package names end with numbers indicating the order in which the code in each project should be studied.

• /components/patterns/ this directory contains the code that demonstrates how to use the Spring AI API to implement common AI application patterns such as retrieval augmented generation. The code in this directory is the same for all the AI providers.

• /applications/ this directory contains the spring boot applications
  that interact with the specific AI providers. The configuration of each project in this directory is different, for example, setting API keys and configuring the AI service with the correct endpoint. To try out the samples in this repo you will be launching the apps in this directory. Each subdirectory contains a readme.md file with instructions on how to run the application. -/pgvector/ this directory contains a docker compose file to launch postgres with the pgvector extension. This is used to demonstrate how to use vector databases with Spring AI.

• docs/ this directory contains the documentation for the repo.

Recommendations to get the most out of the repo

1. Run the samples with the different AI providers to see how the same code works with different providers.
2. Run the gateway application and inspect the API requests/responses to see what interaction with the AI projects looks like on the wire.
3. Make sure to run ollama and download the llama3 model to see how easy it is to run local AI models.
4. The code in this repo is designed to be read in order, so start with the code in the api directory and work your way through the projects. Once you have looked at the code in the api directory move on to the code in the patterns' directory.
5. Spring AI project is evolving quickly, it is possible that the code in this repo will be using a snapshot release of the Spring AI project, or that it falls behind the latest version. If you run into problem with this repo, send a pull request or open an issue.