Introduction

SoftArchitect AI is not just another code chat. It is your shadow software architect. It is an assisted work environment that guides the most critical and often overlooked phase of development: project conception. By leveraging RAG (Retrieval-Augmented Generation) over a curated knowledge base, it ensures your application design complies with SOLID principles, Clean Architecture, and OWASP security guidelines before you write the first line of code. Its mission is clear: transform abstract ideas into development-ready technical specifications, cutting technical debt at the root.

​

The guided architectural workflow

Every session follows a structured five-stage process that acts as a preventive quality gate: Each stage builds on the last, so decisions made early — about domain boundaries, data ownership, or security posture — propagate cleanly into the final technical specification.

​

Why local-first matters

SoftArchitect AI runs entirely on your machine. Your architecture decisions, business requirements, and source context never leave your network unless you explicitly choose a cloud provider. Two modes are available:

• Privacy mode — Runs inference via Ollama on your local hardware. Zero external API calls.
• Performance mode — Connects to Groq Cloud or Google Gemini for faster inference on modest hardware, with your explicit opt-in.

The LLM_PROVIDER variable in your .env controls which mode is active.

​

Key capabilities

​

Technology stack

​

Next steps