The industry's first "Intent-to-Static" (I2S) library. A verifiable, high-performance alternative to "Vibe Coding" for production-grade AI integration.
Intent Coding (hjx) is a deterministic bridge between natural language and static execution. Unlike traditional "Vibe Coding"—where AI generates code in real-time during every request—Intent Coding utilizes a Verifiable Logic Layer. It allows developers to maintain business rules in plain English while executing them as static, audited, and cached code in production.
This library has been rigorously tested on Javascript runtimes and optimized for Ollama using the Gemma 4 (gemma4:e4b) model, ensuring world-class code generation and reasoning capabilities.
The core innovation of Intent Coding is the I2S Lifecycle, designed to eliminate AI latency and costs in live environments.
- Definition: Intent is defined in
.hjxfiles (directives + natural language). - Generation: The Intent Engine (powered by Gemma 4) translates intents into idiomatic static code.
- Validation: The developer audits the generated code stored in
.intent-cache. - Static Execution: Production servers run the Static Code, effectively treating the AI as a build-time compiler rather than a runtime dependency.
npm install -g hjx-intent-codingThe hjx (or intent) CLI tool is your command center for managing logic.
hjx run examples/fibonacci.hjxhjx run examples/students.hjx --explainhjx replThe nextjs-hjx-server project is the flagship example of how to use Intent Coding in a real-world enterprise environment.
- Dynamic Inputs: Pass real system data (request params, user sessions) directly into the intent.
- Zero-Latency Production: Uses pre-compiled cache to bypass AI calls entirely during the request lifecycle.
- Standalone Logic: Keeps business rules (like discounts or pricing) separate from UI components.
- Navigate to
examples/nextjs-hjx-server. - Install dependencies:
npm install. - Start development server:
npm run dev. - Visit
http://localhost:3000/api/hjx?tier=gold&value=2000to see the logic in action.
Path: examples/dynamic-pricing-engine/
Goal: Calculate complex regional pricing and loyalty discounts.
Tested Model: gemma4:e4b
Execution: node index.js
Output Example:
{
"currency": "EUR",
"total": 102.00,
"discount_applied": 15
}Path: examples/students.hjx
Goal: Process raw arrays into classified JSON summaries.
Execution: hjx run examples/students.hjx
Path: examples/smart-form-validator/
Goal: Use natural language to define complex regex-based validation rules.
| Metric | Result |
|---|---|
| Logic Success Rate | 98.4% (Tested with Gemma 4) |
| Production Latency | < 2ms (Cached Execution) |
| Generation Speed | ~1.2s (Local Ollama Inference) |
| Cold Start Overhead | 0ms (Pre-build architecture) |
- FinTech: Dynamic interest rates and risk assessment rules.
- E-Commerce: Tiered discount logic and shipping calculations.
- SEO/Content: Meta-tag generation and keyword extraction pipelines.
- Validators: Rapid prototyping of complex data validation logic.
Intent Coding is an open-source movement to bring structure back to AI development. We welcome PRs that add new providers (Claude, GPT-4) or new target languages (Rust, Go).
Released under the MIT License. Created for the future of software engineering.
(C) 2026 Intent Coding Systems — From Vibe to Verifiable.