Inspiration
Product teams spend days translating brainstorms into formal specs. We wanted to skip the slog and let AI handle that first draft—so everyone can focus on design and delivery instead of paperwork.
What it does
Upload rough notes (text, CSV, PDF) and the app returns clean, Industry-Style requirements: actors, flows, acceptance criteria—even Jira-ready JSON. A detail slider lets you choose brief or deep dives.
How we built it
- FastAPI gathers files and validates them with Pydantic. -Text is embedded and indexed for fast reference searches.
- Gemini (via Google AI Studio) generates the requirement draft, locked to a ruleset that matches Industry's template.
- A React dashboard shows live previews and can push results to Confluence or Jira.
Challenges we ran into
- Data privacy – Stripped PII on-prem and encrypted storage.
- Right level of detail – Added the slider and token tests to avoid walls of text.
- Messy inputs – Integrated OCR and language detection for scans and mixed-language docs.
Accomplishments that we're proud of
- First draft of a full user-story pack in under a minute.
- 18 % accuracy boost after adding a banking glossary.
- Stakeholders trust results thanks to the “show reasoning” toggle.
What we learned
Guardrails matter more than fancy prompts; domain context boosts LLM accuracy; and transparency is key to user confidence.
What's next for Automated Requirement Generator
- Voice capture during stakeholder calls.
- Instant cost estimates beside each story.
- Open-source template packs so other industries can plug in their own rules.
Log in or sign up for Devpost to join the conversation.