@finqu/aptl

v1.0.0-alpha.27

Published

a month ago

A template engine for rendering human-readable templates with simple conditionals and easy data injection.

0High
0Medium
0Low

ozmium

wreakz

template engine rendering conditional data-injection ai

APTL (AI Prompt Template Language)

⚠️ Active Development Notice This project is under active development and not yet production-ready. APIs and features may change as we iterate toward a stable 1.0 release. Please expect potential breaking changes until we reach 1.0.0.

A modern template engine designed specifically for AI system prompts.

Stop wrestling with string concatenation and messy JSON. Write clean, maintainable prompt templates with inheritance, conditionals, and type-safe data injection—compile to optimized output for any LLM.

Why APTL?

Building AI prompts shouldn't feel like writing assembly code. APTL brings modern templating to AI development:

🎯 Purpose-Built for AI - Designed for LLM system prompts, not HTML pages
📝 Human-Readable - Clean, indented syntax that makes sense at a glance
🏗️ Template Inheritance - DRY principles with @extends and modular sections
🔄 Dynamic & Adaptive - Conditionals, loops, and context-aware rendering
🛡️ Type-Safe - Full TypeScript support with detailed error messages
📦 Production-Ready - Used in production AI systems at Finqu

Quick Start

Installation

npm install @finqu/aptl --registry=https://npm.pkg.github.com

Your First Template

Before APTL:

const systemPrompt =
  "You are " + (agentName || "AI") + ", a " + role + ".\n" +
  (hasCredentials ? "Credentials: " + credentials.join(", ") + "\n" : "") +
  "Your goal is to " + goal + ".\n" +
  (examples ? "Examples:\n" + examples.map(e => `- ${e}`).join("\n") : "");

With APTL:

import { APTLEngine } from '@finqu/aptl';

const template = `
@section identity
  You are @{agentName|"AI"}, a @{agentRole} specialized in @{domain}.

  @if credentials
    Credentials:
    @each credential in credentials
      • @{credential}
    @end
  @end
@end

@section objective
  Your primary goal is to @{primaryGoal}.

  @if examples
    Examples of great responses:
    @each example in examples
      Input: @{example.input}
      Output: @{example.output}
    @end
  @end
@end
`;

const engine = new APTLEngine('gpt-4');
const output = await engine.render(template, {
  agentName: 'CodeAssist Pro',
  agentRole: 'senior software engineer',
  domain: 'full-stack development',
  credentials: ['10+ years experience', 'TypeScript expert'],
  primaryGoal: 'write clean, maintainable code',
  examples: [
    { input: 'Optimize this loop', output: 'Use map() instead of forEach for transformation' },
    { input: 'Fix memory leak', output: 'Remove event listener in cleanup function' }
  ]
});

console.log(output);

Output:

You are CodeAssist Pro, a senior software engineer specialized in full-stack development.

Credentials:
• 10+ years experience
• TypeScript expert

Your primary goal is to write clean, maintainable code.

Examples of great responses:
Input: Optimize this loop
Output: Use map() instead of forEach for transformation

Input: Fix memory leak
Output: Remove event listener in cleanup function

What Makes APTL Different?

🔤 Variables with Defaults

Never crash on missing data—gracefully fallback to defaults:

Welcome, @{user.name|"Guest"}!
Timeout: @{config.timeout|30} seconds

🔀 Conditionals & Loops

Build adaptive prompts that respond to context:

@if userLevel == "expert"
  Use technical terminology freely.
@end

@each task in tasks
  • @{task.name} - @{task.priority} priority
@end

🏗️ Template Inheritance

Build sophisticated prompts from reusable components:

@extends "base-agent.aptl"

@section identity(override=true)
  You are @{agentName}, an expert software developer.
@end

📦 Modular Snippets

Share common patterns across templates:

@section guidelines
  @include "snippets/ethical-guidelines.aptl"
  @include "snippets/output-format.aptl"
@end

Real-World Example

See a complete production-ready example in the /demo directory with 5+ AI agent templates including template inheritance, reusable snippets, and dynamic content generation.

Documentation

For comprehensive documentation, examples, and API reference, visit our documentation site:

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.