SpecDrop AI

​## Inspiration

​Product managers and founders spend hours rewriting messy thoughts, notebook sketches, and slack messages into structured tickets. I wanted to build a frictionless workspace where you can dump a raw stream-of-consciousness and instantly get deployment-ready GitHub Issues without any administrative bloat.

​## What it does

​SpecDrop AI takes raw feature specifications or PRD text directly from the user. Powered by the Gemini API, it structurally parses the text into explicit User Stories, Acceptance Criteria, and priority tags. The user can review the parsed cards, copy the Markdown instantly, or push them live as tracked issues to a target GitHub repository.

​## How we built it

​The application was built entirely in public using Bolt.new for full-stack rapid scaffolding. I integrated the Google Gemini API directly on the frontend for dynamic processing based on user-provided API keys. GitHub integration is handled via the Octokit REST API, and product performance monitoring is captured using Novus.ai behavioral mapping.

​## Challenges we ran into

​Handling AI configuration parameters directly via frontend user input variables without relying on a static backend server initially caused silent code blocks and mock data fallback loops. I had to aggressively strip down standard Node.js structures to ensure the Gemini SDK initialized dynamically on click. Synchronizing the Novus.ai telemetry platform across sandboxed Bolt domains also required precise tracking script placement.

​## Accomplishments that we're proud of

I shipped a fully functional micro-utility from a completely blank canvas in less than 3 days. The UI is clean, incredibly intuitive, and securely manages sensitive API tokens entirely within the user's active session without storing data on a database.

​## What we learned

​i learned how to leverage rapid AI generation workspaces like Bolt to collapse the time between ideation and deployment. I also gained deep insight into autonomous product telemetry by working with Novus.ai to map out user interaction journeys directly from codebase changes.

​## What's next for SpecDrop AI

​I plan to introduce automated context mapping, allowing SpecDrop to read an existing GitHub repository structure first so it can suggest contextual code file references directly inside the generated acceptance criteria. I also want to support direct sync capabilities to Jira and Linear.

Built With

• bolt.new
• gemini
• typescript