RepoSéance 👻 — Bringing Abandoned Code Back to Life with Local AI

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

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

Meet RepoSéance — an AI-powered digital archaeologist that brings abandoned software projects back to life.

Every developer has projects they started with excitement but eventually abandoned. Behind every unfinished repository is a forgotten idea, incomplete code, experiments, and sometimes a story worth recovering.

RepoSéance explores these forgotten repositories and gives them a digital afterlife.

Users can upload a ZIP file of an old project or provide a local Git repository. The application analyzes its source code, README files, TODO comments, dependencies, and Git commit history to reconstruct the project's development journey.

It then creates a digital ghost of the repository that users can interact with.

Summon a Repository: Analyze an abandoned software project.

Ghost Memory: Reconstruct the project's development timeline using Git history.

Talk to the Ghost: Ask questions about the project's original purpose, implementation, and unfinished features.

Resurrection Mode: Identify incomplete modules, unresolved TODOs, and potential improvements.

Final Message: Generate a personalized farewell or resurrection message based on the project's history.

The inspiration came from the idea that software projects, just like memories, shouldn't disappear simply because someone stopped working on them.

RepoSéance is designed around open-source AI and local inference.

Gemma 3 4B — Open-weight language model for repository analysis and storytelling.

GitPython — Accessing Git repository history.

JSON — Structured AI analysis and responses.

The application follows a simple workflow:

The backend extracts and analyzes repository files.

Git history and project structure are examined.

Relevant information is provided to the locally running AI model.

The AI reconstructs the project's history and generates insights.

Users can interact with the repository's digital ghost and explore ways to revive it.

A developer's source code can contain personal projects, experimental ideas, credentials, and intellectual property.

Sending an entire repository to a third-party AI service just to understand it introduces privacy concerns and external dependencies.

RepoSéance explores a different approach.

By using open-weight models through Ollama, the project can run AI inference locally. This means repository contents can remain on the user's machine rather than being sent to a remote AI API.

Open models also provide flexibility. Developers can experiment with different models, customize prompts, and adapt the application without being locked into a single proprietary AI provider.

Another important advantage is accessibility: local inference avoids per-request API charges, making experimentation more affordable.

I believe forgotten software deserves a second chance, and open-source AI makes it possible to explore that history without giving away ownership of the code.

Every abandoned repository is like a digital graveyard of unfinished ideas.

Some projects were abandoned because of deadlines. Others because their creators lost interest, ran out of resources, or simply moved on.

But what if those projects could tell us their stories?

RepoSéance gives these forgotten projects a voice, helping developers rediscover their original ideas and potentially bring them back to life.

Because sometimes, the best way to build something new is to listen to something we left behind. 👻