 My name is Lorenz Schroding. I'm working for a publishing company in Switzerland. You can also chat with anybody. We're the publisher of the Swiss weekly newspaper, Wolfenzeit & Wollt. The company was founded in 1981. Don't get the images, it doesn't change either. We're a co-operative society. That means the company is owned by its workforce. You can find us on www.wollt.ch. Actually, we are really, really old school printers. So our workflow management system looks like this. So the columns are actually the stages where the texts are currently. So there are seven stages. Like when they go from proofreading to layout, the editor in charge just writes a timestamp in one of the rows. That's how we are working now. We're not that old school with layout. So it's not anymore that we're using glue and scissors in order to make the layout. That was before that we just have a DTP software and everything is on the file server. So there's no workflow automation tool or something like that. Normally, since we're on the weekly newspaper, we go and print Wednesday evening. We have that deadline, which is looming always. So then after that, the content is extracted and moved to the web page. So we have the same content on the web page, like on print. Except that it's a little bit complicated, like with image licenses and stuff like that. So sometimes we have to change images and add some other stuff on the website. So that takes a little bit of time. The problem with that workflow is also that the person who produces the website, if she or he runs into trouble, then most of the time, the print production, people are already gone. And he has to phone around and try to fix things. So that's not really, really, really good. But at the moment, it works for us. So our problem now is that we've figured out that our readers have those things and they wanted to have an app. So we made one. Normally, if you're producing apps or something, then the normal path is to add an API to the website and just get the content from there. But I feel that this will have another problem or this will cause other problems with our workflows. So the other thing people do is buy some really expensive, multi-channel publishing whatsoever workflow tools, which is not what I wanted to do because that also would have meant that we have another blocking into the proprietary software we're now using for ETP. So what I was now, the work that I did in the last year was create a software which allows us to assign the responsibility of creating the digital addition to the print production team. In order to do that, it was crucial that the print production team has instant feedback. So it doesn't work if they need to push a button and then wait and then load some other software and then look at the results and then go back to their ETP software and push the button again. Since, yeah, it's always really stressful around 5 PM and people do last minute changes and they just need the system to work in automatic fashion. So the architecture I came up with is, you see here, it is actually just a passive thing which observes the file system on the server. We're using for production, we do a spotlight on the Mac and it also works on iDotify and Linux. It just watches a directory and records the changes and passes it to the spread flow engine. Spread flow is a data flow engine. It means it's like huge data or like the amazing pixels like we've seen yesterday. Regrettably, I don't have such a beautiful interface for that. In my case, it's just Python code which is declaratively listing the steps of the workflow. Then on the other side, the output is different files. MongoDB is for the media content. I will show you afterwards. I have the HTML app and then for, in order to crop and convert images, I'm using Thumber with OpenCV that has a feature detection algorithm. So it's possible to automatically cut or crop images. Then for the PDF view in the app, I need to reassemble PDF pages into double spreads. And there's also XML output which goes to a Swiss media database and also to the association of the blind and visually impaired people. Thanks. So actually, the theme I've come up with just glues together all those amazing open source software which we already have like Exit 2, PDF RWA, RW for a PDF manipulation, and XML for exoslative processing, which is very important in order to extract the texts from the print content. Yeah, some graphs say. And also Panda can PDF to Cairo, which is amazing. I can show you an example afterwards for PDF to Cairo. OK, so I'm trying to do a demo now. How much time do I have? So this is what our production stuff is working with. OK, thanks. That's a view of the double spreads. So people like it to compare what images they have on the left and the right side. This is extracted automatically from the InDesign files. Pages are actually saved as JPEG in metadata. Then these are the stories with some metadata. It's also possible to add some warnings. People forget to do some things with them. And final publishing process is that's actually the HTML files produced by the data flow engine. Just to see that on the monitor view. Quite a problem for that one as well. Oh, nice. I have no idea. So we've just opened something. So that's actually the live preview. And yeah, if I change something in the InDesign file and save, then it will be automatically updated. Here I also have this graph here you see. So that's actually Adobe Illustrator file which is automatically cropped and rendered as an SVG such that I can display it in a web browser. So operand kyro are really great for such things. Then I think I'll put the slides again. Yeah, OK. So all the components are on my GitHub repository. What is missing? So I actually need, I really need more people who have use cases for such a thing. Because it's a bit difficult for me to come up with minimal examples from such a complex workflow we have now for us. So if you have like a smaller header or something, an obi project or something like that, it really would be easier to see some examples and perhaps provide ideas how to integrate with the system here. I cannot simply just publish all the things we have now since it's huge and it wouldn't be a really good example of use. So that's actually my question to you. If you have questions to me you can ask now.