How to Train Your AI to Understand Your Specific Workflow

Your workflow is unique

A lawyer’s “client meeting” isn’t the same as a designer’s “feedback round.” A doctor’s “charting” differs from a teacher’s “grading.” Kronux ships with a powerful default—but the real magic happens when you make it yours.

Start with the taxonomy

Kronux organizes time into categories. The AI suggests categories based on your raw logs. To train it:

Use consistent language — When you log “call with client,” stick to that phrasing. The AI learns your vocabulary.
Correct early — If the AI mislabels something, drag it to the right category. Each correction improves future suggestions.
Build your tree — Add subcategories that match how you think. “Billing” under “Legal” or “Critique” under “Design”—the structure reflects your work.

Customize the system prompt

Advanced users can tweak the AI’s instructions. Open Settings → AI → System Prompt. You can add context like:

“I’m a freelance developer. Categorize based on: client work, personal projects, learning, admin. Use my existing category names when possible.”

The AI respects your prompt and aligns suggestions with your taxonomy.

Connect other models

Kronux defaults to a local Ollama model. If you use Claude, Llama, or another model locally, you can point Kronux at it. Same privacy—your data never leaves your machine—with the reasoning style you prefer.

Iterate over time

The more you use Kronux, the better it gets. Your corrections, your categories, and your patterns become the training signal. No cloud, no uploads—just your Mac getting smarter about your work.