Introduction
We’re thrilled to share the Design System Concierge with you! You can reach him here as of today: https://www.kickstartDS.com/concierge/
Watch a small dramatization of what interacting with our Concierge looks like:
Support us on Product Hunt for our launch:
Why build the Design System Concierge?
As Design Systems gain more and more traction these days, the amount of available content can be overwhelming, making it difficult to know where to start. Our AI-powered Design System Concierge can help by providing expert guidance and pointing you towards the right resources, whether it’s implementation guides, architecture documents, adoption strategies, or other content.
Talking to the Design System Concierge
To get the best possible help from our Design System Concierge, we recommend asking detailed questions instead of engaging in a chat. The Concierge provides quick and transparent answers by summarizing the most relevant content sources. Every answer includes underlying sources that you can easily check and visit. Trust us to deliver accurate and reliable information for all your Design System-related queries.
It’s important to note: The final answer is generated by Open AI GPT-3, and even if we take great care to reduce wrong answers you should always dive deeper. The answer given is not really meant as the end of your interaction. You can, and should, dive deeper into the sections most relevant to your question… as the Concierge is great at giving directions, but not really an expert himself (just like in real life!).
Why use it
Using it is free
Asking questions to the Concierge is completely free of charge for you. We’ve developed, and are now hosting, it, because we want to help lower barriers of entry when it comes to Design Systems. After all our mission is to “Democratize Design Systems”.
Additionally it’s a great opportunity for us to learn what you, as a user, actually want to know more about when it comes to Design Systems… or features that you’d expect from a Design System. This in turn can inform us on features to develop for kickstartDS, articles to write, or documentation gaps to fill in.
It’s specialized for Design Systems
Though the final answer is generated by Open AI GPT-3, all input given to it as context is completely controlled by us, and our integration. Only carefully selected sources (= domains with content relevant to Design Systems) are added to the Design System database, backing the Design System Concierge. This ensures answers will be relevant to the topic, and hallucinations are minimized. If there’s no relevant content in the database, no answer will be generated!
Additionally, by providing the exact sources the answer by the Concierge was based on, you should always verify it for yourself by diving deeper.
It promotes great content & creators
Sometimes it’s tough to just “google” your answer, especially when faced with highly individual and specialized questions… the ones you’d typically have when starting out, or when facing new challenges. Good content gets drowned out by a sea of content farm spam. By asking the Concierge instead, you get a highly curated set of content as the backbone. Best case: you discover a new blog author, content creator, or a specific Design System example to follow, helping inform your Design System journey.
How to use it
For a quick overview, have a look at the following, short animation:
1. Ask a question
Enter detailed, natural questions instead of trying to engineer your way out of a thick jungle of content spam, only armed with the waning powers of Google-Fu:
2. Get initial answer
Receive an initial answer instantly by the Concierge. This should be taken as a first summarization by an enthusiastic, helpful assistant… not as the final expert answer to end all answers:
3. Dive into sources
Even more useful then the initial answer are the sources provided by the Concierge. Those are the sections of text in our database that were found to be most relevant to your question. This should provide full transparency, and enables using the Concierge as a great guide-post to the real content:
How it’s done
Database of expert content
The Concierge is backed by a vast database of content curated, prepared and catalogued specifically by us:
For a full list of sourced domains, go to the Concierge. All domains are listed in the section “Knowledge database” over there.
Behind the scenes
The implementation is mainly following the great blog post (and example code) provided by Supabase. Their post was also the main reason for me to use Supabase with their PostgreSQL + Edge Function for the main hosting in the end. It’s a super nicely packaged solution, was just loving the DX!
Going into all the nitty gritty details would go beyond the scope permitted by such a blog post as this one (Spoilers: it includes non-Open AI SBert embeddings, a Fly.io Python- / Flask-Webservice for runtime embedding of questions, spliced and transformed Server Sent-Events streams, pgvector for database support of embeddings, etc, etc). Feel free to join us on Discord to discuss further… and for all the quiet explorers among you, I’ve made a Miro board outlining how everything works, over here:
https://miro.com/app/board/uXjVMdpB0ao=/
Alternatively, you can just go to the Github repository for the Concierge. All the code is public and Open Source here:
https://github.com/kickstartDS/concierge
Conclusion
We really hope the Design System Concierge can help you along your personal journey through the Design System space. We’re thrilled to see all the questions you’ll try on him, and what the overall reaction will be. We already have a truckload of ideas to explore further, if this resonates with the community!
If you want to give feedback to us, have ideas on what to improve or add, or just want to tell us what cool answers you’ve received… we’d love for you to join our small Discord community. Or maybe the Concierge was completely missing the mark for you? We’d especially like to hear about those cases, too!
Top comments (2)
Congrats!
Thanks, Ben! :)