GhostMemo: Keeping Memories Alive with a Local AI Ghost Storyteller.

This submission addresses the Hacktoberfest Open-Source AI Challenge Week 4, titled "Ghost in the Machine." The project, called GhostMemo, is a compact memory journal that incorporates a ghost-story feature. The creator developed it to provide saved memories with a new, eerie existence: users input a memory, save it within their browser, and then utilize an AI model to convert it into a fictitious tale.

The memories are preserved in the browser's local storage, and each memory can be removed from the journal. The generated story is presented as fiction, as the AI model might introduce details that were not originally included in the memory. The website showcases the project, but its AI storyteller operates in a local version using Ollama, which is installed on the creator's computer.

Since GitHub Pages cannot connect to the Ollama model on the creator's laptop, the story generation happens within the user's own machine, rather than on a remote server. The source code for GhostMemo is available on GitHub.

The development of GhostMemo relies on basic HTML, CSS, and JavaScript, all contained within a single index.html file. The JavaScript code manages the saving and display of memories using the browser's local storage. For the ghost-story feature, the page connects to a locally running Ollama model using the Gemma 3 1B model. When the user clicks the ghost-story button, the app forwards the memory to the local Ollama model, which then generates a response.

This response is displayed alongside a label indicating that certain details in the story may be fictional. The GhostMemo project does not require an API key, as the interaction with the Ollama model occurs locally, on the user's computer. This open-source approach allows the creator to experiment with AI storytelling on their own computer without the need for a paid API or copious amounts of code dedicated to handling API keys.

By utilizing an open-weight model, the user maintains control over the memory data and the AI-generated stories, which is crucial for a project centered around personal memories. The initial test of the application began with a brief narrative describing knocks at a locked door. Upon inputting this memory into the system, Gemma transformed it into an extensive ghost story, incorporating additional details that were not part of the original memory.

This unexpected outcome highlights the potential of AI models to produce creative, sometimes surprising narratives based on user-provided content.