This website was a project on its own, with the following objectives and/or motivations:

• It was necessary and long overdue as the existing Wordpress CMS just did not do the job, in particular for a multilanguage website;
• Irrespectively of the website upgrade, we needed to structure the knowledge-base to give more flexibility around retrieval. We expect retrieval to become a big theme in the future, and we believe that a curated knowledge-base will be a prerequisite to maximize the potential of the next generation of AI tools;
• We wanted to allow basic search through the entire base from anywhere (and by anyone);
• We did not want to rely on third-party platforms, such as social media, to manage bookmarks and search. Firstly because they are mostly underwhelming in these areas (Twitter/X is just terrible). And secondly because of a concern that they will limit access to historical data at one point : Twitter's introduction of a cap preventing from seeing older posts was a warning that we took very seriously.

For each project, we try to choose different platforms, tools and languages in order to acquire new skills and knowledge as we build. That said, it is hard to turn your back on Vercel (hosting Off_the_charts). If anything, we are moving more services to Vercel and we went with them again for the frontend (Nextjs app hosted on Vercel). Given our objectives, we wanted a headless CMS and Strapi was a choice we did not regret. We found ways around limitations (WYSIWYG or search), it's very flexible and it's free , so we will likely use Strapi again for another projet. The backend is hosted on Digital Ocean (Strapi app + PostgreSQL database + bucket for media), which we would recommend too. Take this feedback with a pinch of salt, or at least put it in perspective with the fact that we are not developers.

As the secondary objective the all our projects is to learn, we try to refrain from using third-party plugins if it is clear that we can build a feature ourselves. Finding a plugin for a usecase is reasonably easy, but finding code examples to build a new one can be more difficult. To address cases where we felt that frustration, we saved components in this public repo if someone needs inspiration...

Next steps:

• AI-powered categorization;
• Assisted drafting of excerpts using open-source LLM;
• Knowledge graph.