Softr Vibe Coding, Docker’s AI Fix, & The Rise of Dify (Week 49)

Key Takeaways

All right, week 49 [music] of the no-code news. We're going to go over Softr's vibe coding block, which actually surprised me. I thought it would just be another vibe coding thing. [music] Deep Seek math model and other models that we should pay attention to for local LLM in building things. Docker in its work with MCPs, really cool feature here with Docker. Give them a clap for doing something that might solve a problem here. And then the Dify tool I didn't know about. Let's just touch on that for a moment as we share what's trending on GitHub. All right, so the first one is N8N the leap to version 2.0. But anyways, like it's coming. I think I said this last week and you don't have to rush into it. It's there. If you're already have what you have and it's working, I wouldn't sweat it. I asked them on their community forum like when is the end date for or end of life for 1.x or whatever. And that's when you really have to start thinking about things in your company. So yeah, 2.0's coming, some cool features and updates. Nothing to jump on in my opinion. And I mentioned Xano in the last video and their 2.0 is just mind-blowing. So I'm like see what happens here. The next one just to bring up make.com, they've released an HTTP app. Remember when you're doing no-code or any of these integrations, the particular tool you're using for no-code might not have a node or step for your API. So this is really key. By using this you can connect almost any API and then get your work done to get that data to paginate. So it's nice that it's here or here and made better. I will someday check out make and do some videos on that. All right, Softr vibe coding block. I thought this was just going to be another no-code sorry, vibe coding thing. And those are fine, but I don't need to code now with Softr and other tools or if I do it's very small. But what they did is interesting because they've given you the full context of everything around the block, the ACLs, the permissions, the data and all that stuff so you can start using the particular surrounding context to make this bit of code really work well. So I'm excited about this cuz I have some customers who want more complex reporting and charts. Like and wow, I could just use like a nice tool like Nano Banana to make some really fun charts for the customer. All right, so Softr vibe coding block, check it out. There's a YouTube channel link there as well. It's really worth looking at their product. Deep Seek math model. Again, I only talk about AI a little bit and cuz there's so many channels on it, but when it comes to local stuff we can run, stuff we can use in what we're building is really important. So Deep Seek's math V2 and just more of these models out there that can really do a lot. And I have a link below to join my channel to support the long-tail content I have on this where we'll be building our own on-prem system and we'll be using models like this to get the job done. And you'll be surprised. A lot of work can get done with the models we have already. All right, some videos that caught my attention. This one, these guys are always good and this is a building database agents. They cover this stuff a lot. Their videos are just so well done and worth watching. There's a lot of times that I've approached the database with AI and I get stuck. I just think things like this help me to see the bigger picture and how to do it. So this is a big help here. And remember, no-code database queries, there's a lot of power there and it can be done well and it can be done safely. This is another big video that I caught my attention this week is Docker with their MCP hosting pattern. And the big problem here is these MCPs are key. And but if you have too many of them or too many tool calls in them, then the AI context gets filled up so you can't get as much done or the AI could potentially get confused or whatever. And with the Docker MCP protocol or whatever you want to call it, it helps the AI to use the Docker tool to get the right MCP going so it can get just what it needs from it and not fill up all the context. So it's an interesting, really good idea and I'm glad it's nice to see Docker just really getting ahead on this. All right, some trending GitHub and Hugging Face news. Browser use, please go check that one out. That can help you do automatic QA for your websites. It can help you do more complex integrations or data scraping from other websites. Run it locally, run it on our on-prem system, run it in the cloud. It's a lot there that you can do and it's really worth keeping an eye on. Another one is Firecrawl. Firecrawl is an open source solution. They have a paid one. I use the paid one, but I'm going to start running this one in my my training sessions because I do need to do stuff where I get content from websites, scraping them for proposals or other information. And it's just does a really good job. The it might not have all the features of the the paid one, but it's just a really good job at getting the websites, using the whole JavaScript version of it so you can actually get to the website and doing other things like making it markdown and whatnot. So it's just a really good library here and I got to start checking out. This next one, I don't know. Maybe I have amnesia, but I didn't know about this one and it's Dify. So Dify AI is just an another good example of a no-code builder. This one's for Agentic. So it wouldn't replace N8N, but it could complement N8N or ActivePieces by being more of the AI place for the work we're doing. So you could have those other systems be the APIs and the scheduling tasks and the more complex database interactions. You could throw them the particular request here and then come back out the other end with your answers. So I'm going to really dig into Dify. It's something I can run in that particular system I was talking about as I start to build on-prem systems for customers. And again, there's always [clears throat] ComfyUI. That's maybe my things I use to find what's trending, what's big or not is a little bit too maybe have to tweak the prompt a bit. But ComfyUI has been there for a while now and you can use it to make no-code node-based image generation processing systems. It's nice. It's intimidating just like N8N, but I think what I see people do with it is amazing and it would be worth I think for me to get used to it as I try to move more in-house local AI to do images and other things. Mostly for companies want more privacy, but then you get just cost factors and things. And once you dial it in, you can actually you don't need all the upgrades to the AI at that point. And lastly, just another model that's out there. It's the something MAI Z image turbo Tony. It's again, it's image generation. It'd be interesting to see how good it is. It's about speed in their case. But Nano Banana's hard to beat. But when you need an image generation and you're worried about something, these models are there to help do that work. But right now Nano Banana has stolen the show. There's some open source competitors there, but I don't think they're anything that good. But okay, there's a model that seems to be trending. Not a lot in the news. I will continue to improve my prompts. I have three different particular systems looking for data for me and then my own looking. [music] But share comments below if you have any ideas or projects you want me to keep an eye on. I'll just do that. Remember, join below. Not only subscribe and share, but join the channel. I think it's like three bucks a month and it's going to help me start building [music] 2026 year-long content on-prem self-hosting no-code solutions and AI. [music] And track number two is just how do I build? And I'll show you all the tips and tricks there for [music] building with no-code in a way that scales and is just really easy. Relatively easy. [music] Thanks for joining week 49 of 2025. Thank you.