very day, we generate gigabytes of ideas, code, and drafts in our conversations with AI like Gemini, ChatGPT, and Claude. But here’s the harsh truth: we don't truly own this data.
Tomorrow, your account could be suspended (I just lost access to an AI account recently!), the service might go down, or the API might change. And the official "Export Data" features? They usually give you an unreadable JSON file that is impossible to navigate or search through.
I got tired of depending on the cloud for my digital memory. I wanted my AI chats to be stored safely on my local drive, available completely offline, with instant search.
So, I built Gemini Vault.
What is Gemini Vault?
It’s a fully local backup tool and offline viewer for your AI conversations.
GitHub Repo: https://github.com/vuhuutam459-max/gemini-vault
How it works under the hood
I wanted this to be as lightweight and independent as possible:
Zero-dependencies core: The processor runs on the pure Python 3.10+ standard library.
Millisecond Search: I used SQLite with FTS5 (Full-Text Search). You can search through thousands of chats instantly.
Source-Agnostic Core: The database and search engine are designed to be easily expandable without breaking the core logic.
Local UI: A sleek Single Page Application dashboard (index.html) that works completely offline. It supports LaTeX (KaTeX) for math, syntax highlighting for code, and handles broken images gracefully.
The Journey
Building this wasn't just smooth sailing. Initially, I tried parsing data via internal APIs, but realized that violates Terms of Service and isn't stable. Shifting to processing official GDPR/Takeout JSON exports was the turning point for stability. I also had to battle through environment limitations and resource constraints to make the parsing process robust.
What’s next?
Currently, the vault perfectly saves text, code, and formulas from Google Gemini. My next goals for the near future are:
Multi-Source Support (ChatGPT & Claude): The architecture is already prepared for this. I am currently finishing the adapter patterns to normalize exports from OpenAI and Anthropic into the same unified database. This is coming very soon!
Local Media Caching: Automatically downloading inline images from chats so the backup is 100% autonomous even if the original web links die.
Neon Cyberpunk Theme: Adding a custom UI toggle for a dark tech-noir aesthetic to match the project's vibe.
Try it out!
The project is fully open-source (MIT License). If you value your data and want to keep your AI brainstorming sessions safe, check it out.
I’m looking for feedback! I would love to hear what you think, and I warmly welcome any stars, forks, or pull requests on GitHub.
Repo: https://github.com/vuhuutam459-max/gemini-vault
Let's take our data back! 🚀
Top comments (0)