How We Built Metis in 7 Hours: Recovering Hidden Decisions from Open-Source Projects

Описание и применение

В течение семи часов команда Dot создала Mentis – инструмент для отслеживания проверенных коммитов в открытых репозиториях, демонстрируя быстрый прототипинг и интеграцию ИИ.

Задача

Open-source projects accumulate years of development activity, but tracking which commits have been verified becomes difficult as repositories grow.

Результат

Mentis was successfully built in under seven hours, enabling developers to easily monitor verified commits in open-source repositories, demonstrating rapid prototyping and effective AI integration.

Рабочий процесс

• 1. Selected problem statement during Hack Day.
• 2. Designed Mentis to track verified commits.
• 3. Built frontend with Next.js, TypeScript, Tailwind CSS.
• 4. Developed backend API with Python and FastAPI.
• 5. Stored data in PostgreSQL.
• 6. Integrated AI to analyze repository context.
• 7. Debugged data flow issues where AI lacked backend code context.
• 8. Completed product in ~7 hours and merged PR at 5:30 AM.
• 9. Prepared presentation for Hack Day.

Доказательства

Open-source projects accumulate years of development activity across GitHub commits, pull requests, issues, and code changes.

Источник

We wanted Mentis to help answer one important question: What verified changes have been made to this repository?

Источник

Our tech stack included: Frontend: Next.js, TypeScript, Tailwind CSS; Backend: Python, FastAPI REST API; Database: PostgreSQL

Источник

We wanted the AI to work with repository information, but it wasn't fetching the backend code as expected. That meant the AI was missing important context required for analysis.

Источник

After nearly seven hours of work, we finally merged our PR.

Источник