Demo on how easy it is to build AI-driven applications with Spring Boot and Spring AI. It also shows how to implement advanced techniques for the adaption of foundation models with Function Calling and RAG.

Note: A Spring AI Recipe Finder implementation using the Model Context Protocol is available here.

Slides: Building AI-Driven Spring Applications with Spring AI

As Ollama doesn't yet provide a text-to-image model, recipe image generation is not available with this setup. From version 3.2 Llama is supporting Function Calling even if it's not working well with the small models.

1. Download and install Ollama on your local machine link
2. Start llama3.2 model (Ollama 0.2.8 or newer, and Llama 3.1 or newer is required for Function Calling)

   ollama run llama3.2
   

By enabling the "ollama-compose" Spring profile, the llama3.1 model will be automatically started and configured with docker compose. Depending on your system (e.g. ARM macs) this is not a recommended setup due to performance reasons.

export SPRING_PROFILES_ACTIVE=ollama-compose

Set the API key via an environment variable or in application-openai.yaml.

export SPRING_AI_OPENAI_API_KEY=<INSERT KEY HERE>

Run your application with the "openai" Spring Profile.

export SPRING_PROFILES_ACTIVE=openai

Set the API key and endpoint via environment variables or in application-azure.yaml.

export SPRING_AI_AZURE_OPENAI_API_KEY=<INSERT KEY HERE>
export SPRING_AI_AZURE_OPENAI_ENDPOINT=https://{your-resource-name}.openai.azure.com

Make sure the deployment names of the models match exactly what's in your application-azure.yaml configuration file.

Currently, only some regions support image generation with Dall-E. If you use a region that doesn't support it, you have to disable the image generation by setting ai.azure.openai.image.enabled: false in the application-azure.yaml configuration file to not run into errors.

Run your application with the "azure" Spring Profile.

export SPRING_PROFILES_ACTIVE=azure

On your local machine, a Redis database is automatically started and configured with docker compose. As a fallback if no Redis database is configured, a SimpleVectorStore instance will be used.

./gradlew bootRun

Open https://localhost:8080 in your browser. Enter the ingredients (e.g. "Cheese") you want to find a recipe for in the form and press the "find" button.

By checking the "Prefer available ingredients" checkbox, Tool Calling will be enabled. As the functionality for the API call to check the available ingredients in the fridge is mocked, they can be configured via the app.available-ingredients-in-fridge property in application.yaml.

Bacon and onions are currently configured for available ingredients in fridge. With the input "Cheese", you should get a recipe with cheese and bacon.

By checking the "Prefer own recipes" checkbox, Retrieval-Augmented Generation will be enabled.

To upload your own PDF documents for recipes to the vector database, there is a REST API endpoint implemented.

curl -XPOST -F "file=@$PWD/german_recipes.pdf" -F "pageBottomMargin=50" https://localhost:8080/api/v1/recipes/upload

Based on the sample recipes part of this repository, with the input "Cheese", you should get a recipe that goes in the direction of a cheese spaetzle muffin.

kubectl apply -f deployment/kubernetes/ollama.yaml

export SPRING_AI_AZURE_OPENAI_API_KEY=<INSERT KEY HERE>
export SPRING_AI_AZURE_OPENAI_ENDPOINT=<INSERT ENDPOINT URL HERE>
envsubst < deployment/kubernetes/openai.yaml | kubectl apply -f -

export SPRING_AI_OPENAI_API_KEY=<INSERT KEY HERE>
envsubst < deployment/kubernetes/azure-openai.yaml | kubectl apply -f -