Skip to content
Kilavuz Kilavuz
Blog

AI Scheduler Feature

This page explains the AI-assisted scheduler inside the Kilavuz demo AI app. The scheduler collects the user’s daily tasks and preferences and converts them into a structured plan with the help of an AI model.

Flow

  1. The user opens https://kilavuz.app/demo and enters a short list of things to do.
  2. Each task is given a start time and an estimated duration.
  3. The generateGptResponse action sends the task list and preferences to an OpenAI-compatible model and receives a JSON schedule.
  4. The returned plan is parsed into a GeneratedSchedule value and displayed to the user in their timezone.

Exporting the plan

The generated plan can be downloaded in a calendar-friendly format when the user opts in. This flow shares the same download shape as the question-tree Markdown export; see the API reference for details.

Limits

  • The demo AI app is not exposed to production traffic; it is intended for demonstration only.
  • Model output is not guaranteed to be valid JSON. When parsing fails, the application suggests a retry to the user; backend error messages stay in English and are not surfaced to the user.

Configuration

The OpenAI API key is injected as the OPENAI_API_KEY environment variable. The key never reaches the client; it is read only by the server-side app/src/demo-ai-app/operations.ts.