Ghost in the Terminal：讓被遺忘的電腦開口説話的 AI 專案

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

👻 Ghost in the Terminal: What If a Forgotten Computer Could Talk?

Have you ever wondered what would happen if an old computer, left alone for years, suddenly started talking?

That idea inspired me to build Ghost in the Terminal, a small AI-powered project that brings a fictional forgotten machine to life.

I wanted to create something that was more than just another chatbot. I wanted the experience to feel a little mysterious, like opening an old terminal and discovering that something is still running inside it.

So I built a dark, terminal-inspired web app where you can talk to a ghost, generate forgotten memories, and summon mysterious messages.

And yes, you can try it yourself! 👻

What I Built

Ghost in the Terminal is an interactive AI experience built with Python and Streamlit, powered by the open-weight openai/gpt-oss-20b language model through Groq-hosted inference.

The app has three main features:

• 👻 Haunted Chat: Talk to a fictional computer ghost. Ask questions, start a conversation, or explore the mystery of an abandoned machine.
• 📜 Forgotten Memories: Generate spooky stories about forgotten computers, lost programs, mysterious memories, and machines left behind.
• 🔮 Random Séance: Summon a mysterious message with a single click.

I also created a dark interface to match the atmosphere of the project.

My interpretation of the Ghost in the Machine theme was simple: what if a machine could tell stories about the things it had supposedly forgotten?

The result is a small experiment that combines AI, imagination, and a little digital horror.

Demo

You can try the live application here:

👻 Launch Ghost in the Terminal

The app is deployed online, and I've tested the three main features.

Open it, start a conversation with the ghost, generate a forgotten memory, or try your luck with the Random Séance.

Just remember: the machine might have something to say. 🕯️

Code

The complete project source code is available on GitHub:

💻 View Ghost in the Terminal on GitHub

The repository includes the application code, dependencies, and setup instructions so other developers can explore the project and run their own version.

If you try it or have ideas for making it better, I'd love to hear your thoughts!

How I Built It

I built this project using tools that make experimenting with AI accessible without needing a powerful computer.

Here's the technology behind it:

• Python: The main programming language used for the application logic.
• Streamlit: Used to build the interactive web interface.
• Groq Python SDK: Connects the application to the hosted inference service.
• openai/gpt-oss-20b: The open-weight language model that generates the ghost's responses, stories, and mysterious messages.
• Streamlit Community Cloud: Used to deploy the application so people can try it online.

One important detail is that the model does not run locally on my computer. The app sends prompts to Groq's hosted inference service, which generates the responses.

This made it possible to build and deploy an AI-powered experience without requiring a dedicated GPU.

The most interesting part was bringing the different features together into one experience. I wanted the app to feel like interacting with a character rather than simply entering prompts into a generic chatbot.

Why Does Open Innovation Matter?

For me, open innovation is about making it easier for more people to experiment, learn, build, and share.

You shouldn't need an expensive computer or a large development team to start exploring AI ideas.

Using an open-weight model gave me the opportunity to build a creative application around a model whose weights are publicly available under its applicable license. Python and Streamlit helped me turn that idea into something interactive, while hosted inference made it possible to run without powerful local hardware.

It also means other developers can inspect the application code, learn from the implementation, and experiment with their own ideas.

There are still limitations. The application needs an internet connection, depends on the inference provider's availability and usage limits, and doesn't run entirely offline. Open-weight models also have their own licenses and terms that developers need to understand.

But I think that's part of the point: accessible tools give people a starting point. What they create with those tools is up to them.

A small project can still be a meaningful way to learn, experiment, and share something with the community.

The Spirit of It 👻

The idea behind this project is that an abandoned machine might not be as silent as it appears.

Imagine sitting in a dark room, opening an old terminal, and typing a simple question.

Instead of an error message, a reply appears.

Maybe it's a forgotten memory. Maybe it's a story about a program that no longer exists. Or maybe it's just a language model doing what it does best.

I like the space between those possibilities. It gives a familiar piece of technology a little mystery and makes an ordinary interaction feel like the beginning of a story.

That was the feeling I wanted to capture with Ghost in the Terminal.

The project is still a small experiment, but it was exciting to take an idea from a concept to a working application that people can try online.

And who knows? Maybe the ghost has more stories to tell.

Prize Categories

I'm submitting Ghost in the Terminal to the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine.

My project explores creative use of an open-weight language model to build an interactive AI experience.

Thanks to the organizers and everyone in the community making opportunities like this possible. Building this project has been a great way to experiment with AI and turn a spooky idea into something real.

If you've tried the app, let me know what you would ask a forgotten computer! 👻