How We Built ReleaseFlow: AI-Powered Release Notes

ReleaseFlow started from a real pain point: writing release notes is tedious, inconsistent, and usually the last thing developers want to do.

The core idea

Paste your commit messages, PR descriptions, or bullet points — and get polished, App Store-ready release notes in seconds. The AI handles the formatting, tone, and structure. You just review and publish.

Tech stack decisions

We built ReleaseFlow as a Flutter Web app backed by Supabase. Flutter gave us a fast, consistent UI across browsers. Supabase handled auth, the database (release history), and real-time features without us managing infrastructure.

The AI layer uses a large language model with custom system prompts tailored per output format: App Store, Play Store, professional documentation, etc.

GitHub & GitLab integration

One of the most-used features is the GitHub import. Users connect their repo, pick a branch, choose how many commits to fetch, and the app builds the input automatically. This removes the copy-paste step entirely.

We used the GitHub REST API with OAuth token auth — supporting both public repos and private ones via personal access tokens.

Multi-language output

ReleaseFlow supports 8 languages. This was surprisingly straightforward to implement — the AI model handles translation natively, and we pass the target language as part of the system prompt.

What we learned

Prompt engineering is 80% of the work in an AI product. The difference between a generic output and a genuinely useful one comes down to how precisely you define tone, structure, and format in the system prompt.

We iterated on prompts for weeks before the output felt consistently professional.