_A simple, open-source AI coding assistant that runs entirely on your machine using Ollama. No cloud, no data sharing, just helpful AI for developers._
Hey dev community! π
I wanted to share a project I've been working on called Code Genie - an AI coding assistant that runs completely locally on your machine. No subscriptions, no cloud APIs, and most importantly, your code never leaves your computer.
Why I Built This
Like many developers, I've been experimenting with AI coding tools. They're genuinely helpful, but I kept running into the same concerns:
- Privacy: Sending proprietary code to cloud services felt risky
- Cost: Monthly subscriptions add up, especially for indie developers
- Control: I wanted something I could customize for my workflow
- Offline capability: Not always having reliable internet access
So I built Code Genie using Ollama to run everything locally. It's not perfect, but it's been useful in my own workflow, and I thought others might benefit too.
What It Actually Does
Code Genie is essentially a collection of specialized AI agents that help with different development tasks:
- Code generation from natural language descriptions
- Security scanning for common vulnerabilities
- Test generation and quality checks
- Performance analysis and optimization suggestions
- Automated documentation creation
The key feature is that you can describe what you want in plain English, and it helps you build it:
codegenie
> create a REST API endpoint for user registration with email validation
It'll analyze your project context, generate the code, create tests, and even update documentation.
The Multi-Agent Approach
Instead of one monolithic AI, Code Genie uses specialized agents that collaborate:
- ποΈ Architect Agent - Helps with design decisions
- π¨βπ» Developer Agent - Handles code implementation
- π Security Agent - Scans for vulnerabilities
- β‘ Performance Agent - Suggests optimizations
- π§ͺ Testing Agent - Generates and runs tests
- π Documentation Agent - Keeps docs updated
This separation makes it easier to improve individual components without breaking everything else.
Getting Started (It's Pretty Simple)
If you want to try it out:
- Install Ollama
- Clone the repo:
git clone https://github.com/Sherin-SEF-AI/code-genie.git
cd code-genie
./scripts/install.sh
- Start coding:
codegenie
That's it. No API keys, no sign-ups, no configuration headaches.
Honest Limitations
I want to be upfront about what this tool is and isn't:
- It's not a replacement for understanding code - you still need to review what it generates
- It requires decent hardware (8GB+ RAM recommended) since it runs locally
- The AI models aren't as powerful as cloud-based alternatives like GPT-4
- It's still early days - bugs exist and I'm actively improving it
Why Open Source?
I made Code Genie open source (MIT license) because:
- Transparency - You can see exactly what it does with your code
- Community - Better tools come from collaboration
- Learning - I learned a lot from open-source projects, wanted to give back
- Trust - No hidden telemetry or data collection
The code isn't perfect, and I'm sure there are better ways to implement some features. That's why I'd love feedback, contributions, or even just hearing about your experience using it.
What's Next?
I'm working on:
- VS Code extension for better IDE integration
- More specialized agents (DevOps, frontend, backend)
- Improved context awareness
- Better error handling and recovery
- Team collaboration features
Try It Out?
If you're interested in a privacy-first AI coding assistant, give Code Genie a shot:
π GitHub: github.com/Sherin-SEF-AI/code-genie
π Docs: Full documentation in the repo
π¬ Feedback: Open an issue or start a discussion - genuinely interested in hearing what works and what doesn't
If you find it useful, a star on GitHub would mean a lot! β
Final Thoughts
Code Genie is my attempt at building a practical AI tool that respects developer privacy while still being genuinely helpful. It's not revolutionary, but it's honest about what it does and doesn't do.
The AI coding space is moving fast, and I'm excited to see where it goes. Tools that run locally and respect privacy should be part of that future.
Thanks for reading, and happy coding! π
Built with β€οΈ using Python and Ollama by Sherin Joseph Roy
Questions? Comments? Let me know below! π
Top comments (0)