Shapeshift
One morphing text box becomes events, checklists, timers, and converters: parallel Jev intent fan-out plus deterministic value parsing, offline keyword fallback.
Overview
Shapeshift (github.com/anishfn/shapeshift, shapeshiftui.vercel.app) is Anish Gupta's Bun app where a single input becomes the right UI as you type. TypeSafe Jev answers fourteen typed questions in one parallel call (which card type plus signals like video call or urgency) while deterministic code parses dates, amounts, units, and math from the same string. README documents an offline keyword classifier by default so the demo runs without an account; optional TYPESAFE_API_KEY on the server route /api/intent switches the latency badge from jev-offline to pinned jev-1.13.0 with quiet fallback when the API is unreachable.
Problem: Multi-step forms force users to pick a mode before they type, so quick notes become three screens of empty fields.
Built for: Product builders and demo hackers who want one morphing text box that becomes event cards, checklists, timers, and converters without a chat model filling every field.
First indexed on Jev Directory: 2026-09-24
Creator and team
- Name
- Anish Gupta
- Handle
- @anishfn
How Jev is used
- Role in the product flow
- Parallel intent and card-type classification; code fills structured field values
- Primitives
- ChoiceNoul
- State in
- User utterance in the morphing text box; fourteen decomposed questions per README diagram (card type plus boolean signals).
- Decision out
- Selected card template and feature flags consumed by deterministic parsers for values.
- User types natural language in the single input
- Online path: one Jev fan-out answers fourteen questions in parallel
- Offline path: built-in keyword classifier when no key or mock flag
- Deterministic code computes dates, splits, conversions, and math
- Rendered card persists in browser localStorage until deleted
Shapeshift is the UI morphing counterpoint to marketing personalization listings like ploy-ai: Jev classifies intent and card shape, not audience segments. The fan-out pattern keeps latency predictable because every question shares one HTTP round trip while parsers stay testable TypeScript. That split matches the directory mantra of facts in code and judgments on Jev. Press slash to browse card types; optional keys never reach the browser per README server-only routing.
Sourced performance claims
- README states one Jev call answers fourteen typed questions in parallel while code handles values.Source: github.com/anishfn/shapeshift README
- Public GitHub repo anishfn/shapeshift had about five hundred ten stars when this listing was drafted.Source: GitHub star count September 2026
- Live demo hosted at shapeshiftui.vercel.app with Bun 1.2+ dev setup documented in README.Source: github.com/anishfn/shapeshift README
Features and stack
Features
- Sixteen plus card types from events to polls and time zones
- Offline keyword mode without API keys
- Optional TypeSafe server route with offline fallback
- Keyboard shortcuts and localStorage history
- Bun 1.2+ dev workflow
Stack
- Bun
- TypeScript
- TypeSafe Jev
- Vercel deploy
Pricing: Open source; TypeSafe usage bills to your key when online intent is enabled.
Links
FAQ
- Does Shapeshift need a TypeSafe key?
- No for the default experience. README documents offline keyword classification until you set TYPESAFE_API_KEY in .env.local for the server intent route.
- How is this different from ploy-ai?
- Shapeshift morphs a single input into structured personal UI cards. ploy-ai targets marketing segment personalization, not inline card typing.
- Where is my data stored?
- README states saved cards live in browser localStorage until you delete them.
Related learn guides
Original Jev guidance that pairs with this product pattern.
- Jev use cases
The patterns builders actually search for: moderation, routing, triage, RAG verify, and agent gates.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.