Show HN: Fixthisbug.de – AI Code Fixes with Local LLM for Privacy

2 points by DeflectedFlux 11 hours ago

For fun (and learning) I've let an AI built FixThisBug.de, an AI-selfpowered code fixing tool that's different from typical AI coding assistants as it uses a local Ollama instance with the qwen2.5 model.

Key Features:

- AI build: build entirely with AI, over several iterations via v0 and cursor.

- Local Processing: Uses qwen2.5 model through Ollama for code analysis and fixes

- Dual Language: Full support for English and German interfaces

- "Free Tier": Limited daily fixes for casual users

- "Pro Access": Unlimited fixes for authenticated users (just for fun, its free :) )

- Privacy Compliant: GDPR-compliant & hosted in Germany

Tech Stack:

- Frontend: Next.js 15.0, React 19

- Backend: Supabase + Next.js API Routes -

- AI: Local Ollama instance running qwen2.5

- Auth: Supabase Auth with HCaptcha

- Infrastructure: German servers for EU compliance

To add a little extra to the project: Unlike GitHub Copilot or similar tools, FixThisBug.de processes everything locally on the server. No external AI service is used.

I'm particularly interested in feedback on:

1. The local LLM integration approach

2. Performance optimizations for Ollama

3. Your User experience using the service

4. Privacy considerations

Would love to hear your thoughts and suggestions!