Analysis updated 2026-05-18
Play an ongoing AI narrated story offline without an internet connection or API key.
Generate scene illustrations locally as the story progresses using a Mac, Windows, or Linux machine.
Write your own narration or dialogue directly into a story alongside the AI narrator.
| newideas99/open-dungeon | u1acodefinance/polymarket-v2-arbitrage-bot | bigfrankykevin/sportsbook-bet365 | |
|---|---|---|---|
| Stars | 106 | 106 | 105 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | hard | moderate |
| Complexity | 3/5 | 4/5 | 3/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Ollama installed locally and a few GB of disk space for the language model.
Open Dungeon is an AI-powered interactive fiction app that runs entirely on your own computer. You type actions, dialogue, or narration, and an AI narrator responds and continues the story. Both the text and any scene images are generated locally, meaning nothing is sent to an external server and no account or API key is required. For the text side, the app works through Ollama, a tool that runs AI language models on your own machine. It defaults to Google's Gemma 4 models, which come in several sizes ranging from about 4 GB to 19 GB of disk space, so you can pick what fits your hardware. If you already run your own AI server, you can point the app at it instead, since it supports any backend that speaks the OpenAI API format. On the image side, the app can generate scene illustrations inline as the story progresses. On Apple Silicon Macs this uses a model called FLUX, and on Windows and Linux it can use either a graphics card or the CPU. Getting started on a Mac with Apple Silicon is a drag-and-drop installer from the Releases page. On Windows there is a batch file launcher that checks for required software, installs missing pieces where possible, and sets up the image generation backend. A manual setup path is also available for anyone who prefers to clone the repository and run commands themselves. The story interface has three input modes. Do sends a player action that the narrator responds to in second person. Say wraps your text as spoken dialogue. Story lets you write narration directly into the story yourself. Above the text box, Continue lets the narrator advance the scene without any input from you, Retry discards the last response and regenerates it, and Erase removes the most recent exchange. You can also click Edit on any message to rewrite it in place. For long stories, the app keeps as much history in the model's context window as possible, up to 256,000 tokens on the larger models. When older story text would no longer fit, the app compresses it into a rolling summary so the narrator retains the plot without forgetting earlier events. All chats, characters, and images are stored in a local database on your disk.
A local AI interactive fiction app where you type actions and an AI narrator continues the story, generating both text and scene images entirely on your own computer.
Mainly TypeScript. The stack also includes TypeScript, Ollama, Gemma.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.