AI Generated README
sfpm-js is a lightweight, dependency-free, forward-chaining inference engine written in modern JavaScript. It is designed for managing complex state and logic in a declarative way, making it ideal for game development (especially for AI and quest/dialog systems), interactive narratives, simulations, and other dynamic applications.
The core principle is simple: you define a set of Rules, provide the current state as a collection of Facts, and the engine finds and executes the most specific rule that matches the current situation.
The system is built around a few key ideas:
- Facts: Simple key-value pairs that represent the current state of your world or application (e.g., PlayerLocation: 'Dungeon', HasKey: false). The collection of all current facts is managed by a FactSource.
- Criteria: The building blocks of rules. A single criterion is a condition that checks a fact against a value (e.g., "is the PlayerLocation fact equal to 'Dungeon'?").
- Rules: A rule is a collection of Criteria and a payload (a function to execute). A rule is considered a match only if all of its criteria are met by the current set of facts.
- Query: An object that holds the current facts and provides a simple interface to trigger the rule-matching process.
- Specificity-Based Matching: When multiple rules match the current facts, the engine automatically selects the most specific one to execute. Specificity is determined by the number of criteria in a rule—more criteria means more specific. If multiple rules have the same highest specificity, one is chosen at random.
You can install the package using npm or your favorite package manager like bun or yarn.
npm install sfpm-jsOr with Bun:
bun add sfpm-jsHere’s a step-by-step guide to using sfpm-js.
import { Query } from './src/Query.js';
import { Rule } from './src/Rule.js';
import { Criteria, Operator } from './src/Criteria.js';The Query object holds the state of your application. You can initialize it with a Map of facts.
// The world state is a collection of facts.
const worldState = new Map([
['playerHealth', 80],
['playerLocation', 'Forest'],
['isRaining', true],
['hasMagicSword', false],
]);
// The Query object is our interface to the rule engine.
const query = new Query(worldState);Rules consist of criteria and a payload. The Operator enum provides different ways to compare fact values.
- Operator.Equal
- Operator.NotEqual
- Operator.GreaterThan
- Operator.LessThan
- Operator.GreaterThanOrEqual
- Operator.LessThanOrEqual
- Operator.Predicate (for custom logic via a function)
// This array will hold all the logic of our application.
const gameRules = [];
// Rule 1: A general rule for being in the forest.
gameRules.push(new Rule(
[
new Criteria('playerLocation', 'Forest', Operator.Equal)
],
() => {
console.log("It's a bit dark in the forest.");
}
));
// Rule 2: A more specific rule for when it's raining in the forest.
// This rule will be chosen over Rule 1 if it's raining because it has more criteria.
gameRules.push(new Rule(
[
new Criteria('playerLocation', 'Forest', Operator.Equal),
new Criteria('isRaining', true, Operator.Equal)
],
() => {
console.log("The forest is dark and wet from the rain. You should find shelter.");
},
"Rainy Forest Rule" // Optional name for debugging
));
// Rule 3: A rule that uses a predicate for a custom check.
// This rule is even more specific.
gameRules.push(new Rule(
[
new Criteria('playerLocation', 'Forest', Operator.Equal),
new Criteria('isRaining', true, Operator.Equal),
new Criteria('playerHealth', (health) => health \< 50, Operator.Predicate)
],
() => {
console.log("You're cold, wet, and injured. Your chances don't look good.");
}
));
// Rule 4: A rule that modifies the world state.
gameRules.push(new Rule(
[
new Criteria('playerLocation', 'Forest', Operator.Equal),
new Criteria('hasMagicSword', false, Operator.Equal)
],
() => {
console.log("You find a glowing sword stuck in a stone! You pull it free.");
// The payload can modify the facts for subsequent matches.
query.add('hasMagicSword', true);
},
"Find Sword Rule",
10 // A higher priority can break ties among rules with the same specificity.
));Call query.match() with your rules to find and execute the best matching rule.
console.log("--- First Turn \---");
query.match(gameRules);
// Expected Output: "You find a glowing sword stuck in a stone\! You pull it free."
// Why? "Find Sword Rule" (2 criteria) and "Rainy Forest Rule" (2 criteria) both match.
// "Find Sword Rule" has a higher priority (10 vs 0), so it wins.
console.log("\\n--- Second Turn \---");
// The state has changed (\`hasMagicSword\` is now true). Let's match again.
query.match(gameRules);
// Expected Output: "The forest is dark and wet from the rain. You should find shelter."
// Why? "Find Sword Rule" no longer matches. "Rainy Forest Rule" (2 criteria) is now the most specific valid rule.
// Now let's change the health to trigger the most specific rule.
console.log("\\n--- Third Turn (Health is low) \---");
query.add('playerHealth', 40);
query.match(gameRules);
// Expected Output: "You're cold, wet, and injured. Your chances don't look good."
// Why? The rule with 3 criteria now matches and is the most specific.This repository includes more advanced examples in the /example directory:
- textAdventure.js: A complete, playable mini text adventure game that showcases how to structure game logic, locations, and item interactions using sfpm-js.
- dialog.js: A sophisticated dialog system built on top of the engine. It demonstrates how to create branching conversations and manage character memory with a declarative API.
The project uses bun:test for testing. To run the test suite:
bun test
A benchmark suite using mitata is included to measure the performance of rule evaluation and matching under various conditions.
To run the benchmarks:
bun run ./bench/sfpm.bench.js
Contributions are welcome! If you find a bug or have a feature request, please open an issue. If you'd like to contribute code, please fork the repository and submit a pull request.
This project is licensed under the MIT License. See the LICENSE file for details.