@agentlang/agentgen
v0.0.6
Published
AI-powered code generation for [Agentlang](https://github.com/agentlang-ai/agentlang) - a declarative language for building agentic applications. This package provides pre-configured AI agents, workflows, and instructions to generate Agentlang code from n
Downloads
101
Keywords
Readme
@agentlang/agentgen
AI-powered code generation for Agentlang - a declarative language for building agentic applications. This package provides pre-configured AI agents, workflows, and instructions to generate Agentlang code from natural language descriptions.
Overview
AgentGen is a meta-programming toolkit that uses AI agents to automatically generate Agentlang modules, including:
- Data Models: Entities, records, relationships, and attributes
- Workflows: Event-driven business logic and orchestration
- AI Agents: Intelligent agents with LLM integration
- Components: Individual entity, event, workflow, agent, and relationship components
Features
- Full Module Generation: Generate complete Agentlang applications from requirements
- Component-Level Generation: Create or update individual components
- Smart Requirement Analysis: Automatically distill and structure user requirements
- Validation & Retry: Built-in validation with automatic retry logic
- Metadata Generation: Auto-generate documentation for generated code
- Orchestration: Multi-step component creation with dependency management
Installation
npm install @agentlang/agentgenQuick Start
import { BUILDER_MODULE_DEFINITION, getAgentCraftConfigContent } from '@agentlang/agentgen';
// Get the AgentCraft module definition
const moduleDefinition = BUILDER_MODULE_DEFINITION;
// Get LLM configuration
const config = getAgentCraftConfigContent();Usage in Agentlang Applications
To use AgentGen in your Agentlang application:
- Load the module definition into your Agentlang runtime
- Configure the LLM settings using
config.al - Trigger generation events with your requirements
import { BUILDER_MODULE_DEFINITION, getAgentCraftConfigContent } from '@agentlang/agentgen';
// In your Agentlang runtime initialization:
// 1. Load the AgentCraft module
// 2. Load the LLM configuration
// 3. Trigger events like 'generateAgentlang' with user requirementsAPI Reference
Exports
BUILDER_MODULE_DEFINITION
The complete AgentCraft module definition as an Agentlang string. This contains all agents, workflows, events, and data models needed for code generation.
const moduleDefinition: string = BUILDER_MODULE_DEFINITION;getAgentCraftConfigContent()
Returns the LLM configuration content from config.al.
function getAgentCraftConfigContent(): string;Returns: Raw JSON configuration string compatible with Agentlang's loadConfigEntities function.
Utility Functions
escapeAgentlangString(str: string): string
Escapes special characters for safe embedding in Agentlang strings.
import { escapeAgentlangString } from '@agentlang/agentgen';
const escaped = escapeAgentlangString('My "string" with\nnewlines');Development
Build
# Build the package
nx build agentgen
# Build for playground
npm run build:agentgen-playgroundProject Structure
libs/agentgen/
├── src/
│ ├── index.ts # Main exports
│ ├── module.ts # AgentCraft module definition
│ ├── instructions.ts # Agent instructions and prompts
│ └── configLoader.ts # Configuration loader
├── config.al # LLM configuration
├── package.json # Package metadata
├── vite.config.ts # Build configuration
└── README.md # This fileRelated Projects
- Agentlang - The declarative language for agentic applications
Support
Built with ❤️ by the Agentlang team
