Problem

Most modern AI applications are incredibly heavy, requiring gigabytes of external software packages, libraries, and complex setups just to run. This makes them slow, difficult to deploy on lightweight or mobile devices, and highly vulnerable to security tracking. Furthermore, sending raw data directly to cloud AI models risks leaking sensitive personal information and exposes the application to prompt-injection attacks that can hijack the AI's behavior.

Inspiration

Building AI tools usually requires downloading huge software packages and heavy libraries. I wanted to prove that you can build a fast, powerful AI assistant with zero extra downloads. Black Dragon AI was made to be a clean, single-file system that runs directly on your device with maximum speed and total privacy.

What it does

Black Dragon AI connects a "Cyber-Dark" developer chat interface directly to Google's powerful Gemini AI.

Secure Chat: It gives smart engineering answers using the gemini-2.5-pro model. Built-in Firewall: It instantly blocks hackers who try to trick the AI or steal its secret instructions. Privacy Guard: It automatically hides sensitive details like phone numbers and emails before sending data online. Self-Contained Server: It hosts its own website and local database right from your device. How we built it

The entire project was built from scratch without using any heavy frameworks:

The Backend: Written in pure, native Python using only built-in tools to talk directly to Google’s Gemini REST API. The Frontend: A responsive web interface designed with a dark hacker aesthetic that links seamlessly to the Python server. Local Database: Powered by an embedded SQLite database to securely save chat memories and compliance logs on your own device. Challenges we ran into No Extra Libraries: Because we used zero external packages, handling complex JSON data packages and web security using raw Python was difficult and required precise coding. Mobile Environment: Developing, testing, and uploading the entire system directly from a mobile terminal setup forced the code to be incredibly clean, compact, and efficient. Accomplishments that we're proud of We bypassed heavy software wrappers to speak directly to Google's AI backend using raw code. We created an incredibly secure, full-stack AI ecosystem that takes up virtually zero megabytes of space. The entire platform runs beautifully as a light, high-speed single-file engine. What we learned

I learned the deep inner workings of how web APIs and internet protocols handle data. By skipping heavy standard packages, I mastered how to build secure web routing, write lightweight firewalls, and build clean software that runs fast on any device.

#What's next for Black Dragon

Next, we want to add a localized document scanner so the engine can read and analyze entire project code folders instantly. We also plan to add an offline backup mode so the firewall features keep working even if the internet goes down.