Forma — AI-Assisted Product Intake Workflow by 會飛的 香蕉Forma — AI-Assisted Product Intake Workflow by 會飛的 香蕉

Forma — AI-Assisted Product Intake Workflow

會飛的 香蕉

會飛的 香蕉

Forma — front-facing product presentation.
Forma — front-facing product presentation.
Fictional request — structured output in labelled local parser mode.
Fictional request — structured output in labelled local parser mode.
Review and edit — user decisions stay beside the original request.
Review and edit — user decisions stay beside the original request.
Human-confirmed brief — JSON export includes review notes.
Human-confirmed brief — JSON export includes review notes.
Responsive mobile interface — fictional demo content.
Responsive mobile interface — fictional demo content.
Personal Project / Demo · Role: Product UX, workflow design, full-stack implementation
Forma is a personal product intake demo built around human review. A user starts with a rough event or product request, then receives an editable brief containing a title, goal, audience, date, location, format, deliverables, tone, and open questions. The original request stays beside the working draft so decisions can be checked against the source. Missing details become visible questions; users can save answers as review notes, edit fields, and confirm the final brief before exporting JSON. I designed a product presentation with a scroll-driven words-to-fields story, a focused working surface, clear state changes, and English copy. The Node.js server supports schema-constrained output through the OpenAI Responses API. Without a model key, an explicitly labelled local parser demonstrates the review flow. The screenshots show that parser mode. This is a generic portfolio concept, with no client-specific implementation or production-use claims.
Screenshots show fictional requests and labelled local parser mode. A live model call is not verified without an API key.
Like this project

Posted Sep 28, 2026

Personal demo: rough requests become structured, editable briefs with human review, saved decisions, confirmation, and JSON export.