Cloudflare Workers AI toMarkdown API & SDK Generator

REST API Contract (POST /api/convert) & wrangler.jsonc Binding

The Worker exposes a free, keyless public conversion endpoint backed by Cloudflare Workers AI env.AI.toMarkdown(). Submit a multipart/form-data payload containing any 0-Neuron document (up to 10 MB) under the file field:

HTTP/1.1 200 OK
Content-Type: application/json

{
  "data": "# Architecture Specification\n\nConverted Markdown content...",
  "name": "architecture-spec.pdf",
  "size": 248910,
  "mimetype": "application/pdf",
  "tokens": 1842
}

Response Format Negotiation & Guardrail Status Codes

• JSON Metadata Mode (format=json or Accept: application/json): Returns the extracted Markdown string in data alongside name, size, mimetype, and tokens.
• Direct Markdown Stream (format=markdown or Accept: text/markdown): Returns Content-Type: text/markdown; charset=utf-8 with an RFC 5987 Content-Disposition: attachment; filename="document.md" header.
• Guardrail Semantics: Returns 400 Bad Request when file is missing or empty, 413 Payload Too Large above 10 MB, 415 Unsupported Media Type for image/SVG or unsupported formats (protecting 0-Neuron billing), 429 Too Many Requests above 15 requests/minute per IP, and 500 Internal Server Error on upstream failure.

Required wrangler.jsonc Configuration

{
  "name": "everything-to-markdown",
  "main": "src/index.ts",
  "compatibility_date": "2025-04-02",
  "ai": {
    "binding": "AI"
  }
}