Serializable JSON Rules
Store, version, transmit, and audit business logic as data instead of burying decisions in brittle if/else branches.
SOURCE_RELEASE :: v0.7.0
Neuron JS
AI-friendly TypeScript rules engine for serializable JSON business rules and deterministic workflow decisions.
Use Neuron-JS when hardcoded conditions are too rigid, but a full workflow platform is too heavy. Define approved rules as data, validate them, execute them predictably, and explain the result.
RULE_ENGINE_SURFACE
Core Capabilities
Validation Before Runtime
JSON Schemas and validation helpers let teams and AI agents reject malformed generated rules before execution.
Deterministic Decisions
Run the same pricing, eligibility, routing, policy, or feature-targeting contract in Node.js or the browser.
REGISTRY_ENGINE_PATTERN
Neuron Registry + Synapse Runtime
Neuron owns the approved vocabulary: actions, conditions, parameters, rules, and lifecycle hooks. Your application decides what generated or stored JSON is allowed to do.
Synapse executes the serializable script against a runtime context, then returns a result that can be summarized, validated, and explained for logs, tests, or human review.
// pricing-rules.script.json
{
"id": "pricing-decision",
"rules": [
{ "type": "simple_rule", "conditions": [ ... ], "actions": [ ... ] }
]
}
VALIDATE: validateScript(script)
EXECUTE: new Synapse(new Neuron()).execute(...)
EXPLAIN: explainExecution(...)
// deterministic decision boundary
DEFINE: DecisionDefinition + approved registry
VALIDATE: context before component execution
EVALUATE: evaluateDecision(definition, context, neuron)
RETAIN: receipt + replay evidence
SIDE_EFFECTS: host application only
DECISION_RUNTIME :: v0.7.0
Deterministic Decision Runtime
Define a reviewed DecisionDefinition, validate its caller-supplied context before any component runs, and evaluate against an immutable snapshot. The result includes diagnostics and a receipt that the host can retain or replay.
Neuron-JS returns decision evidence, not side effects. Your host, n8n workflow, or agent graph owns context resolution, receipt storage, approvals, and the downstream action.
RELEASE_SURFACE :: v0.7.0
Release Highlights
- 0.7.0 added an opt-in deterministic decision runtime: validated decision definitions, immutable context evaluation, canonical receipts, replay, and portable test vectors.
- 0.5.0 added schemas, validation helpers, normalized execution output, and explainability contracts.
- 0.5.1 added llms.txt, llms-full.txt, assistant instructions, and the official Neuron-JS AI skill.
- 0.5.2 added package-level llms/llmsFull autodiscovery fields, Claude guidance, richer workflow-agent docs, and corrected AI-skill Markdown examples.
AI_READABLE_DOCS
Built for Humans and Coding Agents
The public docs expose runnable examples, schema contracts, decision-runtime and agentic-boundary guidance, reproducible benchmarks, `llms.txt`, `llms-full.txt`, and an official skill so coding agents can use Neuron-JS without inventing APIs.