generating-apps.md

Generating Applications with AI

Learn how to generate complete ObjectQL applications from natural language descriptions. This guide covers zero-shot microservice generation, prompt templates, and best practices for AI-powered app creation using LLMs.

The Workflow

1. Prompt: You ask the AI to design a system (e.g., "A Property Management System").
2. Generation: The AI outputs a set of *.object.yml files.
3. Execution: You drop these files into your project, and the API is instantly live.

Prompting Guide

To get the best results, use a prompt that enforces the ObjectQL schema format.

Template

Act as an ObjectQL Architect.

I need a backend for [APP_DESCRIPTION].

Please generate the object definitions in YAML format. Follow these rules:

1. Use standard types: text, number, boolean, select, date, lookup.
2. For relationships, use type: lookup with reference_to: <object_name>.
3. Include required: true for important fields.
4. Suggest 2-3 essential objects.

Output the YAML for each object in a separate code block.

Example: Real Estate CRM

User:

"I need a backend for a Real Estate Agency to track Properties and Agents."

AI Output:

# agent.object.yml
name: agent
fields:
  name:
    type: text
    required: true
  email:
    type: email
    required: true
  phone:
    type: phone
  license_number:
    type: text
  status:
    type: select
    options: [active, inactive]

# property.object.yml
name: property
fields:
  address:
    type: textarea
    required: true
  price:
    type: currency
  bedrooms:
    type: number
  bathrooms:
    type: number
  listing_agent:
    type: lookup
    reference_to: agent
  status:
    type: select
    options: [for_sale, sold, pending]

Automating the Process

You can build tools to automate this loop.

1. The Generator Script

Imagine a simple CLI tool that takes a user description and writes files to disk.

import { generateSchema } from './my-llm-service'; // Wrapper around OpenAI

async function main() {
    const description = process.argv[2];
    const schemas = await generateSchema(description);
    
    for (const schema of schemas) {
        fs.writeFileSync(`${schema.name}.object.yml`, yaml.dump(schema));
    }
    
    console.log("App generated! Starting server...");
}

2. Hot Reloading

Since ObjectQL can load metadata at runtime, you can build Self-Evolving Apps.

1. The App receives a request: "Add a 'renovation_date' field to Property."
2. The App calls an LLM to update the YAML.
3. The App reloads the metadata registry.
4. The new field is immediately available via API.

Summary

ObjectQL turns software development into a Content Generation task. Instead of generating complex imperative code (which is brittle), you generate simple declarative configurations (which are robust).