OpenForm is documents as code. It lets developers and AI agents define, validate, and render business documents using typed, composable schemas. This eliminates template drift, broken mappings, and brittle glue code — while giving AI systems a reliable document layer they can safely read, reason over, and generate against in production workflows.

Package overview

Framework-neutral AI tool definitions for OpenForm. Provides Zod input schemas, execute functions, and an HTTP registry client that any AI framework adapter can wrap.

• 5 tools - validateArtifact, fill, render, getRegistry, getArtifact
• No framework dependency - Pure tool protocol, no AI SDK lock-in
• 3 render modes - Render from artifact JSON, URL, or registry lookup
• Registry client - Fetch artifacts from any OpenForm registry over HTTPS
• Edge-compatible - Optional proxy text renderer for edge runtimes

This package is the foundation for @open-form/ai-sdk (Vercel AI SDK) and @open-form/tanstack-ai (TanStack AI). Advanced users can build custom adapters for other frameworks.

Installation

npm install @open-form/ai-tools

Usage

Direct usage (no framework)

import {
  executeValidateArtifact,
  executeFill,
  executeRender,
  executeGetRegistry,
  executeGetArtifact,
} from "@open-form/ai-tools";

// Fetch available artifacts from a registry
const registry = await executeGetRegistry({
  registryUrl: "https://public.open-form.dev",
});

// Fetch a specific artifact
const { artifact } = await executeGetArtifact({
  registryUrl: "https://public.open-form.dev",
  artifactName: "pet-addendum",
});

// Validate the artifact schema
const validation = executeValidateArtifact({ artifact });

// Fill with data and check for errors
const fillResult = executeFill({
  artifact,
  data: {
    fields: { petName: "Buddy", species: "dog", weight: 45 },
    parties: { tenant: { id: "t1", name: "Jane Doe" } },
  },
});

// Render to markdown (inline)
const rendered = await executeRender({
  source: "registry",
  registryUrl: "https://public.open-form.dev",
  artifactName: "pet-addendum",
  data: fillResult.data,
  layer: "markdown",
});

Render modes

The render tool supports three input modes via discriminated union:

| Mode | Fields | Resolution | |------|--------|------------| | source: 'artifact' | artifact, baseUrl? | Direct - no fetch needed | | source: 'url' | url | Fetches artifact JSON, derives baseUrl | | source: 'registry' | registryUrl, artifactName | Fetches registry.json, then artifact |

Configuration

import type { OpenFormToolsConfig } from "@open-form/ai-tools";

const config: OpenFormToolsConfig = {
  defaultRegistryUrl: "https://public.open-form.dev",
  proxyTextRenderer: {
    url: "https://your-documents-service.example.com",
    apiKey: "your-api-key",
  },
  fetch: customFetchWithAuth,
};

Edge compatibility

• PDF and DOCX renderers work everywhere (pure JS)
• Text renderer (Handlebars) requires Node.js. On edge runtimes, configure proxyTextRenderer to delegate to an HTTP service

Tools

| Tool | Input | Output | Network? | |------|-------|--------|----------| | validateArtifact | {artifact, options?} | {valid, detectedKind, issues?} | No | | fill | {artifact, data} | {valid, artifactKind, data?, errors?} | No | | render | 3 modes (see above) | {success, content, encoding, mimeType} | Modes 2+3 | | getRegistry | {registryUrl} | {name, items[]} | Yes | | getArtifact | {registryUrl, artifactName} | {artifact, artifactName} | Yes |

Changelog

View the Changelog for updates.

Related packages

• @open-form/ai-sdk - Vercel AI SDK adapter
• @open-form/tanstack-ai - TanStack AI adapter
• @open-form/sdk - OpenForm framework SDK

Contributing

We're open to all community contributions! If you'd like to contribute in any way, please read our contribution guidelines and code of conduct.

License

This project is licensed under the MIT license.

See LICENSE for more information.