intent-gen

Natural language to IntentLang spec generation via LLM (Layer 0) for IntentLang.

Part of the IntentLang toolchain — a declarative specification language for human-AI collaboration.

What it does

Translates natural language descriptions into valid .intent specification files using an OpenAI-compatible LLM API. Implements a generate-check-retry loop: the LLM output is parsed and validated, and if validation fails, errors are fed back to the LLM for correction.

• Generation — describe what you want in plain English, get a valid .intent spec
• Editing — modify existing specs with natural language instructions
• Validation loop — auto-validates output via parser/checker, retries with error feedback
• Model-agnostic — works with any OpenAI-compatible API (OpenAI, Anthropic, Ollama, vLLM)
• Confidence levels — 1 (conservative) to 5 (comprehensive) controls generation scope

Usage

use intent_gen::{LlmClient, GenerateOptions, generate};

let client = LlmClient::from_env().unwrap();
let spec = generate(&client, "a user authentication system with login and logout", &GenerateOptions::default()).unwrap();
println!("{}", spec);

Environment Variables

Modules

Part of IntentLang

This crate is the natural language generation layer (Layer 0). Other crates in the workspace:

• intent-parser — PEG parser and typed AST
• intent-check — Semantic analysis and type checking
• intent-render — Markdown, HTML rendering, and formatting
• intent-ir — IR lowering, verification, audit bridge
• intent-cli — CLI binary (intent generate, intent check, etc.)