This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine
What I Built
What if an abandoned computer could talk to you?
I built GhostShell AI, a spooky, interactive terminal experience featuring a fictional ghost named Wisp, who lives inside an old computer system called OLD-MAINFRAME.
Wisp responds to visitors, explains commands in simple language, and gives spooky but clear warnings about potentially dangerous commands. The goal was to turn a familiar programming interface into something creative, interactive, and educational.
I interpreted the “Ghost in the Machine” theme as bringing a personality to a forgotten computer system using open-source AI.
Demo
GhostShell AI runs locally on my computer using Streamlit and Ollama.
The application uses Gemma 3 1B to generate Wisp's responses. A screenshot of the interface is included as the cover image. I recommend adding a screenshot or short screen recording showing Wisp responding to a prompt as well.
Code
👩💻 GitHub Repository: https://github.com/titlydhara12-spec/GhostShell--AI
The repository contains the application, AI response logic, safety-classification rules, tests, and setup requirements.
Feel free to explore the code and share suggestions!
How I Built It
I used Python, Streamlit, Ollama, and Gemma 3 1B to build the experience.
Here is how the main components work:
Streamlit: Provides the interactive web interface.
Ollama: Runs the language model locally.
Gemma 3 1B: Generates Wisp's character-driven responses.
Safety.py: Checks user input against predefined command patterns and classifications.
Testsafety.py: Tests selected safety-classification cases.
The application classifies user input before deciding how to present it to the model. Commands are simulated and explained rather than executed by the application.
This is a learning prototype, not a complete security system. Its command detection uses predefined patterns and cannot identify every potentially dangerous command.
Why Does Open Innovation Matter?
Open innovation made this project possible for me as a student and beginner developer.
Using an open-weight model through Ollama gave me a way to experiment with AI locally instead of depending entirely on a hosted AI API. It also helped me understand how a language model, a user interface, and a basic safety layer can work together.
Open-source tools make experimentation more accessible, encourage developers to learn from one another, and give projects room to grow through community feedback.
I hope others can explore GhostShell AI, suggest improvements, and help make it better.
My Agent Session
The idea behind GhostShell AI is simple: a forgotten machine still has a story to tell.
The Spirit of It
The idea behind GhostShell AI is simple: a forgotten machine still has a story to tell.
Wisp is not just there to answer questions. The character turns ordinary terminal interactions into a small fictional experience, where curiosity meets a little mystery.
Building this project reminded me that AI projects don't always need to solve a huge problem to be interesting. Sometimes, combining a simple idea with the right tools is enough to create something worth exploring.
Prize Categories
Best Use of Gemma
GhostShell AI uses Gemma 3 1B through Ollama to generate Wisp's responses, making Gemma a core part of the experience.
Built with curiosity, Python, and a little bit of digital mystery. 👻
Top comments (0)