model-language (Ruby)

Ruby bindings for model-language — a typed, safe template language for AI-agent prompts. Runs the exact same engine as the JavaScript/TypeScript package via an embedded WebAssembly module: byte-for-byte identical output, guaranteed by a shared conformance suite.

▶ Live demo — try the language in a Tiptap editor (live validation, autocomplete, real-time render), powered by the same engine.

Filters format values — dates, numbers, text:

{{ user.created_at | date: "MMM D, YYYY" }}            →  Jul 5, 2026
{{ user.created_at | date: "h:mm A", "Europe/Kyiv" }}  →  5:37 PM
{{ user.last_seen  | time_ago }}                       →  3 days ago

require "model_language"

engine = ModelLanguage::Engine.new
out = engine.render(
  "Hi {{ user.name | default: 'there' }}!",
  data: { "user" => { "name" => "Vasyl" } }
)
puts out["text"] # => "Hi Vasyl!"

Directives embed machine-readable constraints in a prompt. They are stripped from the rendered text and returned in directives:

src = <<~SRC
  Help with billing.
  {{verify_before: payments}}
  {{identity: contact.email == payment.email}}
  Greet {{contact.first_name | default: "there"}}.
SRC

out = engine.render(
  src,
  data: { "contact" => { "first_name" => "Vasyl" } },
  schema: [
    { "path" => "contact.email",      "type" => "string" },
    { "path" => "contact.first_name", "type" => "string" },
  ]
)
puts out["text"]                             # => "Help with billing.\n\nGreet Vasyl."
puts out["directives"].map { |d| d["name"] } # => ["verify_before", "identity"]

• render(template, data:, schema:, options:) → {"text", "warnings", "resolvedBranches", "directives", "tokenEstimate"}
• validate(template, schema:, options:) → {"diagnostics", "maxTokenEstimate"}
• parse(template) → {"ast", "diagnostics"}

Never raises for template problems — they degrade to empty output plus a warning. Pass options: { "now" => epoch_ms } for datetime filters.