Show HN: VZparse human/LLM readable format with built-in schema validation (github.com)

VZparse is a new, minimal configuration language and TypeScript/JS parser that blends YAML-style readability with JSON-like structure and adds built-in schema validation, template strings, imports, multiline strings with metadata, comments, and detailed error reporting. The format surfaces a compact schema syntax (e.g., id: "int(min:1,max:1000)!" and string(format:email)!), a ${variable} template interpolation system, and an import mechanism for modular configs. The project ships with a full v1.0 specification (ABNF grammar), a vzparse JS/TS SDK, a VS Code extension, documentation site, and examples — all written in TypeScript and built with Bun, under an MIT license. For the AI/ML community this matters because experiment configs, hyperparameter sets, and deployment manifests benefit from strong, composable validation and reproducibility. Built-in type constraints and expressive validation reduce silent misconfiguration of training pipelines or model serving stacks; template resolution and imports make it easier to share, extend, and parameterize experiment definitions across teams. The parser API supports resolving templates with a context (VZ.parse(..., { resolveTemplates: true, context: {...} })), and the ecosystem includes linting/formatting (Biome), tests and build scripts, plus a VS Code extension to improve developer ergonomics.