Stop Building Chatbots: Why We Switched to Prompt-Oriented Programming (POP)

Описание и применение

Перевод Markdown‑спецификаций в Kanban‑таблицу с помощью строгих JSON‑шаблонов LLM позволяет полностью автоматизировать создание задач за считанные секунды.

Задача

Using open‑ended LLM prompts to generate database mutations caused hallucinated JSON and inconsistent data, breaking application state.

Результат

A 10‑page spec is converted into a fully mapped Kanban board in under 15 seconds, eliminating manual data entry and ensuring deterministic output.

Рабочий процесс

• 1. Upload raw Markdown spec to backend
• 2. Pass spec to Gemini API with a system instruction and JSON schema definition
• 3. Gemini returns a JSON array of task objects
• 4. Backend validates array against database schema
• 5. Bulk create Kanban tickets
• 6. Board is fully populated without manual entry

Доказательства

By wrapping the LLM in these strict POP constraints, the output became completely deterministic.

Источник

In under 15 seconds, a raw Markdown document is translated into a fully mapped, assignable Kanban board without a single manual data entry step.

Источник