Bonkers by Foyer
Bonkers by Foyer is a creative production system rebuilt from v2 to v3 with reusable templates, multi-model routing, and faster workflows.
Overview
Bonkers is a creative production system for visual assets. It moves beyond one-off image generation to reusable workflows where users discover styles, start from templates, refine outputs, and manage history in one place.
Bonkers is a creative production system for visual assets. It moves beyond one-off image generation to reusable workflows where users discover styles, start from templates, refine outputs, and manage history in one place.
I led the v2 to v3 rebuild, adding template-driven workflows, image-to-image remix and editing, and a multi-engine generation layer.
How It Was Built
The main technical choices behind the product, from system design to the parts that make it work day to day.
- Redesigned the generation pipeline around reusable templates with structured inputs, parameterized defaults, and version control.
- Built a unified routing layer across 22+ image models from Replicate, Fal AI, Ideogram, OpenAI, Google Vertex, GoAPI, and Midjourney, with cross-provider fallbacks.
- Redesigned the generation pipeline around reusable templates with structured inputs, parameterized defaults, and version control.
- Built a unified routing layer across 22+ image models from Replicate, Fal AI, Ideogram, OpenAI, Google Vertex, GoAPI, and Midjourney, with cross-provider fallbacks.
- Implemented parallel batch generation of up to 4 images per request via Promise.all, replacing sequential provider calls.
- Extended the feature set beyond text-to-image into remix, inpaint, upscale, background edit/erase, and background removal.
- Built Templates as a first-class product surface with authoring tools, discovery, and one-click instantiation rather than treating them as saved prompts.
Impact
- Daily active usage grew by 50% after the v3 rebuild around template-first, reusable workflows.
- Parallel batch generation cuts wall-clock time from the sum of per-image latencies to the slowest call in a 4-image batch.
- Daily active usage grew by 50% after the v3 rebuild around template-first, reusable workflows.
- Parallel batch generation cuts wall-clock time from the sum of per-image latencies to the slowest call in a 4-image batch.
- 22+ model routing gives users engine choice per task while cross-provider fallbacks keep generation resilient when a provider fails.
- Templates drove 20K+ generated images in the first month by lowering the barrier to a useful starting point.
Highlights
- Template-first workflows replace one-off prompting with reusable, versioned recipes.
- Image generation and editing tools are unified in one interface instead of scattered across services.
- Template-first workflows replace one-off prompting with reusable, versioned recipes.
- Image generation and editing tools are unified in one interface instead of scattered across services.
- Parallel batching and provider fallbacks reduce wait times and keep generation resilient.
Tech Stack
More Projects
Additional work across AI products, developer tooling, and full-stack systems.
Fieldcraft
A careful Chrome extension copilot that reads the open job tab, evaluates fit against your profile, researches the company, and drafts truthful answers for review before any form is filled. It never auto-submits.
Inkdown
Markdown editor and sharing platform with folders, live preview, and public links.
DeployNinja
GitHub-native deployment platform for automated builds, live logs, and repeatable releases.