This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine.

👻 GhostShell: A Haunted AI Living Inside Your Terminal

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine.

What I Built

What if opening an AI chatbot felt less like using an application and more like waking up a forgotten machine?

That idea inspired GhostShell, a spooky, terminal-inspired AI chatbot that brings a fictional ghost to life using locally running AI models.

GhostShell lives inside a dark terminal-style interface. Instead of behaving like a regular chatbot, it has its own fictional personality: mysterious, witty, slightly sarcastic, and occasionally unsettling. You can talk to it casually, ask questions, or get help with programming What I Built

What if opening an AI chatbot felt less like using an application and more like waking up a forgotten machine?

That idea inspired GhostShell, a spooky, terminal-inspired AI chatbot that brings a fictional ghost to life using locally running AI models.

GhostShell lives inside a dark terminal-style interface. Instead of behaving like a regular chatbot, it has its own fictional personality: mysterious, witty, slightly sarcastic, and occasionally unsettling. You can talk to it casually, ask questions, or get help with programming while experiencing a different kind of AI interaction.

I wanted to explore the idea of a ghost living inside an abandoned computer—not an actual supernatural entity, but an AI character that makes an ordinary conversation feel a little more mysterious.

The project combines a haunted visual atmosphere with local AI inference. Its personality is guided by a system prompt, while the AI model generates responses based on the conversation.

GhostShell is my experiment in making AI interaction feel more like an experience than a simple question-and-answer exchange.

Demo

👻 Try GhostShell: [Add your deployed demo URL here]

If a public deployment is not available, I'll share a short screen recording demonstrating the interface and a conversation with GhostShell.

Code

💻 GitHub Repository: [https://github.com/sne-ha091/GhostShell]

The repository contains the Flask backend, frontend files, dependency information, and setup instructions for running GhostShell locally.

How I Built It

I built GhostShell using a combination of open-source tools and web technologies.

The frontend: HTML, CSS, and JavaScript power the terminal-inspired interface and conversation experience. The dark visual style helps establish the atmosphere of a forgotten machine.

The backend: Python and Flask handle the application's routes and chat API. When a user sends a message, the frontend sends it to the Flask backend, which prepares the conversation and forwards it to the local AI model.

The AI: I use Ollama to run open-weight language models locally. My application is configured to support Qwen 2.5 (qwen2.5:latest) and Gemma 3 (gemma3:4b), provided the selected model is installed on the machine.

The connection: The Flask backend communicates with Ollama through its local API. The model generates a response, Flask returns it to the frontend, and GhostShell displays it in the conversation.

The personality: One of the most interesting parts was designing GhostShell's system prompt. It instructs the model to behave like a fictional entity trapped inside an abandoned computer, using eerie humor and natural dialogue while remaining helpful with ordinary questions.

The architecture looks like this:

User → Web Interface → Flask Backend → Local Ollama Model → Flask Backend → GhostShell's Reply

Because inference runs locally, GhostShell does not require an external hosted AI API for its core chat functionality. The project is also an opportunity for me to learn more about backend development, API integration, and building applications around open-weight AI models.

Why Does Open Innovation Matter?

Open innovation makes projects like GhostShell possible for developers who are still learning and experimenting.

Using Ollama and open-weight models gives me the opportunity to explore AI integration without depending on a paid, hosted AI API for every conversation. I can experiment with different models, adjust the system prompt, observe how their responses change, and learn how AI-powered applications work behind the scenes.

It also makes AI development feel more accessible. I can build the frontend, connect a backend, run a model locally, and experiment with a distinctive idea using tools available to the developer community.

For me, open innovation is not just about using existing tools. It is about taking them, understanding how they work, and combining them to create something personal.

GhostShell would be much harder for me to experiment with if every interaction required a paid external AI service. Local inference gives me more control over experimentation, although performance, hardware requirements, and privacy still depend on the setup.

Open-source tools helped turn a spooky idea into a working project—and gave me an opportunity to learn while building something of my own.

The Spirit of It

The idea behind GhostShell is simple: a forgotten computer suddenly has someone to talk to again.

I wanted the experience to feel as though a machine that had been sitting silently for years had finally awakened. The user types a message, the terminal responds, and for a moment, the interface feels like a small scene from a science-fiction horror story.

Of course, there is no real ghost inside the machine. There is a local AI model, a backend, a frontend, and a system prompt shaping the character. But that is part of what makes the experiment interesting: a few familiar technologies can create an interaction that feels different from an ordinary chatbot.

Building GhostShell has also reminded me that AI projects don't always need to be about solving a huge problem. Sometimes, they can be about.exploring an idea, learning new technologies, and making something that sparks curiosity.

This is still an evolving project. I'd like to improve its dialogue, refine its personality, and add more atmospheric terminal effects so that interacting with GhostShell feels increasingly immersive.

For now, the machine is awake.

The terminal is waiting. Do you dare to say hello?

Prize Categories

Selected category or categories: [Insert the exact eligible prize categories from the official challenge rules.]

I'll confirm the category requirements before submitting to make sure GhostShell fits the selected category or categories.