👻 GhostTerminal Pro — A Haunted AI Companion with Google Gemma

Hacktoberfest Open-Source AI Challenge: Ghost in the Machine Submission 👻🌼

What if your code snippets could receive a reply from a ghost living inside a forgotten computer?

For the Hacktoberfest 2026 Week 4 challenge — “Ghost in the Machine,” I built GhostTerminal Pro, a spooky AI-powered application that transforms text and code snippets into eerie commentary, humorous digital curses, and suggestions inspired by computing history.

🔗 GitHub Repository: Avinash-sdbegin/ghost-terminal-pro

Developers spend plenty of time reading code, logs, and text. I wanted to turn this ordinary interaction into something creative and entertaining.

GhostTerminal Pro gives submitted text a fictional cyber-ghost personality. Enter a thought, a sentence, or a code snippet, summon the ghost, and receive a themed response through a simple web interface.

The goal is to combine a little Halloween fun with a practical experiment in generative AI.

GhostTerminal Pro generates three sections:

For example, I tested the application with:

print ( " Hello World " ) Enter fullscreen mode Exit fullscreen mode The model responded with a spooky interpretation of a simple greeting, followed by a humorous warning about phantom characters appearing in future code.

The generated response is creative commentary, not a verified code review. The application does not execute commands or access the user's terminal.

The interface allows users to enter text or code, submit it, and view the generated ghostly response.

The notebook initializes the API client, prepares the prompt, sends the user's text to Gemma, and displays the response in the Gradio interface.

You can explore the implementation here:

📓 Open the Google Colab notebook on GitHub

Gemma makes it possible for developers to experiment with open-weight models and build applications around them, rather than relying exclusively on closed model families.

For GhostTerminal Pro, I used Gemma through a hosted Google API. This made it easier to integrate the model into a lightweight Python application and experiment with prompt design.

There is an important trade-off: the current version does not run inference locally or work offline. Submitted text is sent to the hosted API, and model availability or rate limits can interrupt generation.

In a future version, I would like to explore local inference, which could offer more control over how submitted text is processed, depending on the model, hardware, and deployment setup.

GhostTerminal Pro is an educational prototype built for experimentation and entertainment.

Avoid submitting passwords, private messages, or sensitive information. Always review generated content rather than treating it as verified technical advice.

Some improvements I would like to work on are:

GhostTerminal Pro is my interpretation of “Ghost in the Machine”: a spooky little companion that adds personality to everyday text and code.

Building this project gave me an opportunity to experiment with Gemma, connect a model to a Gradio interface, and learn more about API integration and reliability.