Trello Triage is a custom-built, AI-powered task management tool that enhances Trello with real-time classification, smart views, and an integrated AI assistant. It was designed to reduce decision fatigue, streamline backlog triage, and give users more clarity over complex boards—without relying on external APIs or cloud-based AI services.

I designed and developed the entire platform from scratch. The backend is built with Flask and uses local LLMs (Mistral, TinyLLaMA via llama.cpp) to analyze card content and automatically assign tasks to the most relevant list. It supports dynamic classification using semantic search, keyword weighting, and priority cues, all customizable per board.

Key Features:

The frontend mirrors Trello’s layout while adding powerful new ways to interact with your data:

• Board View: A real-time, drag-and-drop Kanban-style interface.
• Calendar View: Visualize tasks by due date across weeks or months.
• Next 3 Days: A focused view for urgent upcoming tasks.
• AI Assistant: Ask questions like “What’s still in backlog?” or “What’s due tomorrow?” and receive context-aware answers in seconds.

The system uses Trello OAuth for secure authentication and includes per-user config options for classification rules, label mappings, and memory retention in chat. It runs fully self-hosted, performs well on modest hardware, and is designed for privacy-first teams who want intelligent automation without third-party dependencies.