Glyphcast is an alchemical webservice that transmutes your DOCX and SVG files into beautiful PDFs. Harnessing the power of CairoSVG and unoconv, Glyphcast is optimized for blazing-fast in-memory conversion of small files (think <100mb).

Glyphcast currently supports conversions from SVG, DOCX, and HTML files to PDF.

See client.py for example client usage, or see the examples using HTTPie and cURL below.

HTTPie

http -d GET http://localhost:5000/convert?from=svg&to=pdf @sparkfish.svg --output sparkfish.pdf

cURL

curl -vX GET http://localhost:5000/convert?from=svg&to=pdf -d @sparkfish.svg --header "Content-Type: application/text" > sparkfish.pdf

Development

Copy example.env to .env and ensure that the environment variables are properly set (the parameters are thoroughly documented in the .env file comments).

Then, to install dependencies, pip install -r requirements.txt and simply do python main.py to start the server.

In production you'll want to use a production grade server such as gunicorn:

gunicorn --workers=4 main:app

Docker image

To build the Dockerfile, do docker build . -t glyphcast, and then docker run -p 5000:5000. This will start a container with the app running behind a gunicorn server listening on port 5000.

Caveats and Roadmap

tmpfs storage

Presently Glyphcast uses /dev/shm to store intermediate files during document conversion in volatile memory (in /dev/shm/glyphcast). Intermediate files are unlinked immediately after conversion. Note that different Linux distributions have different default size limits on /dev/shm, but it tends to be about half the memory on the system. Check the amount available on your distribution of choice with df -h /dev/shm. Resizing can be done with e.g. mount -o remount,size=16G /dev/shm (the previous command remounts /dev/shm with 16GB). Further there are no guarantees that data written to /dev/shm will not be written to disk in a system under heavy load. Future versions of Glyphcast will create and mount a dedicated tmpfs filesystem with configurable sizing.

Scalability

At the moment, Glyphcast handles requests sequentially, meaning it is unsuitable for heavy load. We intend to support a pub/sub architecture in the future for better scalability.