Working AI Surfaces
Seven functional demo surfaces — ThreadStudio, PRD Studio, contact AI, and more — enabled out of the box.
Start from a working application: Laravel auth, typed routes, structured AI workflows, admin screens, ontology tooling, and local verification commands already wired together.
View on GitHubcomposer create-project xuple/evolayer-base-starter my-appThe split
It turns agents, structured output, streaming, provider config, ontology, and React surfaces into a reproducible Laravel project you can run, inspect, resync, and eject from.
Application outcomes
Seven functional demo surfaces — ThreadStudio, PRD Studio, contact AI, and more — enabled out of the box.
Real-time token streaming mapped directly into your React components, not raw SSE plumbing.
A compiled ontology gives the LLM↔frontend channel a strict, type-safe event vocabulary.
Gemini and OpenAI are runtime-verified for ThreadStudio; the laravel/ai SDK reaches many more via one env switch.
A committed lock and an exact-pinned framework mean your setup never randomly drifts.
Keep pristine examples updated via resync, or eject any surface to own the code entirely.
First-class Laravel Boost wiring for Claude Code, Codex, OpenCode, and Cursor.
Testable, not magical
EvoLayer Base proves your AI logic works before you ship it. With Pest-first tests, runtime doctor checks, and a strict SSE vocabulary, you aren't guessing whether your agent's structured output will parse.
$ composer test$ php artisan evolayer:doctor --strict$ php artisan evolayer:ontology:compileStart in the README when you want to run it. Move to the docs when you want to understand, customize, update, or eject it. Managed surfaces stay updated via resync; eject any surface to own it outright.