📑 Table of Contents
⚡ The Vision
Every AI coding assistant today answers the same question: > "How should this look?"
Color palettes. Font pairings. Useful—but fundamentally shallow. Real, industry-grade design requires asking harder questions before a single line of code is written:
- Who is this for, and what do they actually need?
- What does this specific industry strictly forbid?
- What cognitive psychology drives this user's decisions?
- Will this actually convert?
Forma is built to answer all of them. Automatically. In sequence.
💎 Core Features
Forma operates at the absolute pinnacle of AI software engineering. It is not just a prompt file—it is a secure, multi-language execution environment powered by Python and Go.
🧠 12-Layer Reasoning System
Before any UI is generated, Forma forces your AI through a sequential gauntlet. It analyzes the UX Pre-Flight, queries Industry Intelligence, cross-references Cognitive Psychology, applies Component Best Practices, and performs an Accessibility Audit. Only then does it write code.
🩺 AST-Driven Auto-Healer (Zero-Token Healing)
Instead of forcing the AI to waste thousands of tokens rewriting an entire React component just to fix a missing aria-label, the Forma Critic agent triggers the Go-based Auto-Healer (forma heal <file>). The CLI physically parses your frontend AST and surgically injects missing ARIA labels, alt texts, and SVG properties in milliseconds.
📐 Mathematical W3C Tokens
Every generated design system is backed by advanced mathematics (OKLCH Color Spaces, Fluid clamp() Typography, and Spring Physics) and exported directly into the global W3C Design Token format.
🛡️ Graceful Halting
Forma will never catastrophically crash on a developer's machine. The forma doctor command instantly diagnoses your environment. If it detects a system failure, it triggers a Graceful Halt rather than hallucinating bad code.
🏗️ System Architecture
Forma is powered by a dual-engine architecture: a high-speed Go CLI that manages system states, and a Python Intelligence Engine that processes complex reasoning matrices.
graph TD
A[User Prompt in IDE] --> B{Forma Agent}
B -->|Calls| C[Python Intelligence Engine]
C --> D[12-Layer Reasoning Matrix]
D --> E[Industry Rulesets]
D --> F[Psychology Triggers]
D --> G[Accessibility Audit]
G --> B
B -->|Generates| H[Frontend Code]
H --> I[Forma Critic]
I -->|Detects WCAG issues| J[Go CLI: forma heal]
J -->|AST Injection| H
classDef primary fill:#6366F1,stroke:#fff,stroke-width:2px,color:#fff;
classDef secondary fill:#8B5CF6,stroke:#fff,stroke-width:2px,color:#fff;
classDef engine fill:#10B981,stroke:#fff,stroke-width:2px,color:#fff;
class B primary;
class C engine;
class D,E,F,G secondary;
🔬 Research & Data Methodology
Forma is built on a foundation of peer-reviewed UX research, mathematical color spaces, and cognitive psychology. It is not an LLM hallucinating designs; it is an LLM querying a strict, deterministic database.
The Automated Reasoning Sequence
When you prompt your IDE, Forma executes a massive background sequence before writing any code:
