Fantasy AI Adventure

A tiny text-based adventure RPG for the command line, built in .NET 10. Storylines, enemies, interludes, and custom character classes are generated dynamically by a local Ollama model. A 4K reward image is produced by Pollinations.ai at the end of every run.

Team

• Team name: AllAboutCoding

Members

• Tracy Rickman — @tracyrickman
• Ben Lopez — @blopez01

Category

• Primary: Creative application
• Secondary: Azure OpenAI / LLM app

Architecture

The application is a single-project .NET 10 console app (AdventureGame) organized into small, focused components:

• Program.cs — entry point and replay loop.
• Game.cs — orchestrates one full playthrough: setup → 3 encounters with narrative interludes → ending → reward image. Also owns the turn-based combat loop.
• Player.cs — player stats, class presets (Warrior / Mage / Archer / Custom), health potion, and class-specific special abilities.
• Enemy.cs — enemy model with HP, attack, and a one-shot signature move.
• LlmService.cs — talks to a locally-running Ollama instance (llama3.2) via http://localhost:11434/api/generate. Splits generation into several small, focused JSON prompts (intro, three enemies, three interludes, custom-class stats) instead of one large payload, which dramatically improves JSON reliability. Falls back to hand-authored content per piece on any failure.
• ImageService.cs — downloads and saves the end-of-game reward image from Pollinations.ai.
• Typewriter.cs — incremental text rendering for narrative sections.

Runtime flow: the user names their hero and picks a class (Warrior / Mage / Archer or a fully custom class whose stats and signature move are designed by the LLM). Ollama then produces the adventure — a themed intro, three escalating enemies, and three narrative interludes — over a series of small JSON calls. The player fights three turn-based encounters with HP carrying over, then Pollinations.ai renders a themed victory or defeat image and saves it to disk.

Setup

Prerequisites

1. .NET 10 SDK (installed with Visual Studio 2026, or via https://dotnet.microsoft.com).
2. Ollama — https://ollama.com/download
   • After installing, pull the model used by the game:

     ollama pull llama3.2
     

   • Make sure Ollama is running (it usually auto-starts as a background service on http://localhost:11434).
3. Internet connection for the reward image (Pollinations.ai — no API key required).

No environment variables or configuration files are required.

Run the game

From the repo root:

dotnet run --project AdventureGame

If Ollama isn't running or a sub-prompt fails, the game still plays — the affected pieces fall back to built-in content per theme.

License & attribution

This project is released under the MIT License.

Built with Llama. See CREDITS.md for third-party notices, including the Llama 3.2 Community License, Ollama, and Pollinations.ai.