An app to generate D&D party shared connections
This app is a web app that uses the llama3 8B model to generate shared connections between a party of Dungeons and Dragons player characters.
It uses the Ollama library to run the model.
The application uses server side rendering to generate the HTML pages.
The application uses Tailwind CSS for styling.
The application structure is a foray into Vertical Slice Architecture, where each capability (feature) is a separate package with one Kotlin source file for each capability (create a party of characters, generate shared connections, etc.).
Within each capability source file, there will be the route for the capability, along with the "service" and repository functions needed to facilitate the capability. Models used by the service functions are meant to represent DDD entities and are stored in a models package under the capability package. Entity helpers (with the MongoDB collection name and helper functions to convert Kotlin data classes to and from Mongo Documents) are stored outside the capabilities package, in a top-level db package.
The plugins package configures various aspects of the application, including the HTTP Server, MongoDB connection, and the HTTP Client that interacts with Ollama.
The web package contains the index route and HTML elements along with a Styles CSS file that offers sets of Tailwind CSS classes for components.
Looking at the route and service functions for each capability, the route is structured using a try/catch with any errors being returned as an error message and either a 404 when data is not found in Mongo or an HTTP status 400 error for exceptions.
Service functions are wrapped in a runCatching and log exceptions in an onFailure block. Result objects are returned from service functions holding either data or the exception that occurred.
TODO currently the connection to Ollama is done within an Application function in the plugins package and this needs to be moved elsewhere.
Add 127.0.0.1 host.docker.internal to your host machine's /etc/hosts file.
Run docker compose up to create a single node MongoDB replica set and start the Ollama server.
Set the environment variable LLM_URL=http://localhost:11434/api/
The first time the app runs, Ollama will download the llama3 8B parameter model, so it'll take a while to start.