Skip to content

README.md

Latest commit

 

History

History
159 lines (121 loc) · 5.43 KB

README.md

File metadata and controls

159 lines (121 loc) · 5.43 KB
Raw

docdrop

Drop a document. Get clean Markdown. No AI, no account, no history.

docdrop is a small, self-contained web application. It embeds a drag-and-drop browser interface and a JSON conversion endpoint over the Rust @firecrawl/anydoc engine, which turns office documents into consistent GitHub-Flavored Markdown.

There is no private service, database, login, or AI dependency. Conversion runs entirely in-process.

Browser → Node web server → anydoc (Rust) engine → Markdown response

docdrop ships in two forms that share the same UI:

  • Server build (src/, public/, Dockerfile) — a self-hostable Node app that runs anydoc server-side. Documented below.
  • Browser build (docs/) — a fully static site that runs anydoc as WebAssembly in the browser, so files never leave the visitor's machine. It is deployed to GitHub Pages and needs no server. See Browser build (GitHub Pages).

Run

Requirements: Node.js 20 or newer.

git clone https://github.com/iam-niranjan/docdrop.git
cd docdrop
npm install
npm start

Then open http://127.0.0.1:3000.

Set PORT or HOST to change where it listens (default 127.0.0.1:3000).

Run with Docker

docker build -t docdrop .
docker run --rm --name docdrop \
  --read-only \
  --tmpfs /tmp:rw,noexec,nosuid,nodev,size=64m \
  --cap-drop ALL \
  --security-opt no-new-privileges:true \
  --memory 512m --memory-swap 512m \
  --cpus 1 --pids-limit 100 \
  -p 127.0.0.1:3000:3000 \
  docdrop

For public deployment, place it behind an HTTPS reverse proxy or managed load balancer with request-rate and concurrency limits.

Browser build (GitHub Pages)

docs/ is a fully static version of docdrop that runs the anydoc engine as WebAssembly directly in the browser via @firecrawl/anydoc-wasm. There is no server and no upload — the document is converted on the visitor's own machine. It is published to GitHub Pages by .github/workflows/pages.yml on every push to main.

The ~6 MB WebAssembly binary is not committed; the workflow fetches it from npm and drops it beside the page at deploy time. To build and preview it locally:

npm install --no-save @firecrawl/anydoc-wasm@0.1.5
cp node_modules/@firecrawl/anydoc-wasm/anydoc_wasm.js docs/
cp node_modules/@firecrawl/anydoc-wasm/anydoc_wasm_bg.wasm docs/
python3 -m http.server -d docs 8000   # then open http://127.0.0.1:8000

(The copied anydoc_wasm* files are gitignored.)

Supported formats

Category Extensions
Word .doc, .docx, .docm
PowerPoint .ppt, .pps, .pot, .pptx, .pptm, .ppsx, .ppsm
Excel .xls, .xlsx, .xlsm, .xlsb
OpenDocument .odt, .ods, .odp
Rich Text Format .rtf
EPUB .epub
CSV .csv
PDF .pdf

How it works

The Node web server serves the embedded HTML, CSS, and JavaScript interface and exposes a single conversion endpoint:

  • POST /api/convert — multipart upload with one file field; responds with { "markdown": "…", "filename": "name.md" }.
  • GET /api/health — returns { "status": "ok" }.

anydoc detects the format from the file's content markers (not just its extension), so mislabeled files still convert correctly. For markerless formats such as CSV, docdrop falls back to the file extension. Every format is rendered through anydoc's single Markdown serializer for consistent headings, tables, lists, and footnotes.

Uploaded files are processed in bounded memory for the current request. The application does not create a conversion history or store uploaded documents.

Save AI tokens with Markdown

Before giving a PDF, Word document, or presentation to an AI assistant, convert it to Markdown and provide the .md file. Markdown preserves useful structure without much of the presentation and layout noise in office-document formats. This can reduce unnecessary input tokens and lets you remove irrelevant sections first. docdrop itself does not use AI.

Privacy and security

  • No login, database, cookies, analytics, telemetry, AI, or external converter calls.
  • Uploads are limited to 20 MB and held in memory only for the current request.
  • Raw ZIP archives are rejected outright, including a disguised ZIP uploaded under another file extension.
  • File extensions are allowlisted, and each conversion is bounded by a timeout.
  • Every response sets a strict Content-Security-Policy and related hardening headers, and POST /api/convert rejects cross-origin requests.
  • Filenames and document contents are not written to application logs.

Tests

node --test
docker build --target test -t docdrop-tests .
docker build -t docdrop .

Project layout

src/server.js    Node HTTP server, static UI, and the /api/convert endpoint
src/convert.js   anydoc wrapper: upload limits, format gating, timeout
public/          Server build's browser interface (HTML, CSS, JS)
Dockerfile       Multi-stage build with a test stage and runtime image
docs/            Static browser build (anydoc-wasm) deployed to GitHub Pages
.github/         Pages deploy workflow

Credits

Conversion is powered by @firecrawl/anydoc (MIT).

License

docdrop is released under the MIT License. Third-party dependencies retain their own licenses.