enconvert

---

EnConvert reads two kinds of input through one API: files, which it converts between formats, and web pages, which it renders and returns as text or images.

This repository is the API under AGPL-3.0: request routing, the conversion pipelines, storage, background jobs, and a self-hostable render path. The hosted service at enconvert.com runs this code together with additional closed components — the multi-engine render ladder, anti-bot rendering, render-quality scoring, LLM-based extraction, and semantic diffing. The table under Open source vs Cloud lists exactly which parts are in this repository and which are not.

Endpoints

Verb		What it does
perceive	URL → MD/JSON/PNG/PDF	Renders a URL once and returns Markdown, HTML, a screenshot, a PDF, extracted links, and structured data.
convert	FILE → FILE	Converts between 46 file types, including PDF, DOCX, XLSX, PPTX, HEIC, SVG, and CSV.
discover	SITE → URL MAP	Lists a site's URLs from its sitemap and an HTTP crawl, without rendering pages.
lookup	QUERY → RESULTS	Queries web, news, scholar, and maps search.
distill	SCHEMA → JSON	Extracts structured data from a page against a schema you supply.
ingest	SITE → CHUNKS	Crawls a site and returns JSONL chunks suitable for RAG indexing.
watch	PAGE → DIFF	Checks a page on a schedule and sends a webhook when it changes.

Quick start (hosted)

Create an API key at enconvert.com, then:

# perceive: a live page → Markdown
curl -X POST https://api.enconvert.com/v2/perceive \
  -H "X-API-Key: YOUR_API_KEY" -H "Content-Type: application/json" \
  -d '{"url": "https://example.com", "outputs": ["markdown"]}'

# convert: a URL → PDF
curl -X POST https://api.enconvert.com/v1/convert/url-to-pdf \
  -H "X-API-Key: YOUR_API_KEY" -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'

Self-hosting

git clone https://github.com/enconvert/enconvert.git
cd enconvert
docker compose up

The self-hosted build performs the file conversions and renders pages with a plain headless browser. See SELF_HOST.md for the full capability list and configuration.

Render quality scoring

A page fetch can succeed at the HTTP level and still return something other than the page — an anti-bot challenge, a login wall, a cookie interstitial, or an unhydrated SPA shell. Downstream consumers usually cannot tell the difference from the response body alone.

The hosted service scores every read as render_quality (0.0–1.0) and flags a blocked or login-walled result rather than returning it as ordinary content. This scoring is not part of the open build; a self-hosted render returns whatever the browser produced, unscored.

Open source vs Cloud

	Self-hosted (this repo)	Cloud (enconvert.com)
46-format file conversion	✅	✅
perceive: URL → Markdown / screenshot / PDF	basic render	Chromium engine, scored
discover: site → URL map	✅	✅
lookup: live web search	results only	+ render, extract & cited answers
ingest: site → RAG chunks	✅	✅
watch: change monitoring	content-hash	semantic diff
Anti-bot / stealth rendering	—	✅
render_quality read scoring	—	✅
distill: schema → structured JSON	CSS pass	+ LLM extraction
Managed dashboard, teams, billing	—	✅

Rows marked — are absent from this repository. The self-hosted build is functional without them: it converts files, maps sites, crawls to chunks, and renders pages with a plain headless browser.

SDKs

Official SDKs are MIT-licensed and default to the cloud API:

• Node / TypeScript SDK
• Model Context Protocol server — one-command install for Claude, Cursor, and Windsurf

License

EnConvert is open source under AGPL-3.0 — see LICENSE. The SDKs and MCP server are MIT. The hosted service at enconvert.com includes the additional components listed above.

---