VIZLY SPACE

01Context

A shared visual reference for prompt engineers and artists, where an image remains connected to the instructions and settings that produced it.

02Problem

Image-generation results often travel without their prompt or model settings, making a useful result hard to understand, credit, or reproduce. Vizly makes the recipe part of the entry and lets visitors browse before they create an account.

03Constraints

1. 01Keep public discovery open while gating personal saves, creator follows, and publishing behind sign-in.
2. 02Present long prompt text and image-generation metadata in a readable detail view.
3. 03Keep creator identity and actions attached to the prompt receipt.
4. 04Support a varied image gallery alongside search, tags, sorting, and prompt-specific actions.

04Architecture

The promptgallery source uses the Next.js App Router and a client-driven React interface. Typed Prompt and User models connect catalog data to reusable browse, prompt-detail, submission, profile, and account components. A Gemini service can expand prompt text and suggest tags; the live site presents discovery, public receipts, creator actions, and utility views as one archive.

Prompt catalog and types

Define the image, prompt, creator, tags, and source fields used throughout the browsing experience.

Archive and search

Combine a prompt grid, topic tags, search, trending and newest views, random discovery, and category filters.

Prompt receipt

Show the artwork with its full source prompt and available model, aspect ratio, seed, license, and creator details.

Creator and account flows

Separate public browsing from save, follow, profile, and publish actions that require a creator account.

Gemini prompt helpers

Use Gemini 2.0 Flash to expand a short prompt and suggest tags, while returning the original prompt if enhancement fails.

05Decisions

06Trade-offs

07Performance

• ↯Use browsing controls for search, tags, sorting, and random discovery to shorten the path from opening the site to finding a useful prompt.
• ↯Keep Gemini assistance scoped to explicit prompt-enhancement and tag-suggestion actions rather than making archive browsing wait on a model response.

08Challenges

• #Keep the prompt, image, creator, and generation settings legible together when a receipt contains a long recipe.
• #Make public discovery feel complete while clearly communicating which personal and publishing actions need an account.
• #Preserve useful prompt text when an optional AI enhancement call fails.

09Learnings

• →For generative artwork, the process details can be as valuable as the final image; provenance belongs in the content model and the UI.
• →A guest-friendly archive and account-based creator library can share one interface without making discovery feel gated.
• →AI helpers work best as optional, recoverable steps around a human-authored prompt.

10Retrospective

A next pass could version each prompt receipt with its model revision and generation parameters, giving creators a clearer path to reproduce results as image tools change.