With App Connect getting it's major overhaul in Wappler 6, what will be the main focus in Wappler 7? Not asking for a complete roadmap as that can be very fluid, just asking what the main focus will be.
Well we have quite a lot on our todo list for Wappler 7 but the main focus will be:
- improve targets and deployments with new centralized target manager
- improve visual workflows editors with more functionality and maybe real time inline debugging
- improve design view with more visual aids
- improve the property UI panels with more richer controls on demand rendering and filtering, combination of static and dynamic controls instead of two separate sets and dynamic attributes
- visualize more event bindings per components in design view
- introduce building tools for publishing where code gets compiled, minified and css cleaned. Tree shaking
- maybe add some AI integrations
- and much, much more
14 Likes
Thanks George!
Am glad that is a 'maybe' as focus elsewhere is a priority for us. A.I is everywhere and if anyone wants to use it (Prompt Engineers hahahaha) then they have plenty of resources for it outside of Wappler. Can see the Community filling up with questions about JavaScript that A.I has generated, or code that is outside of Wappler's jurisdiction, ie non standard implementations... Anyway just my thoughts, but as for the rest it all sounds good to us!
5 Likes
a new text editor? just spent the last two days implementing editorjs... wish there was another option haha
4 Likes
Looks like it is mostly Wappler UI changes. I'm also not sure about AI. I'd have to see it in action. Not sure how it would remain Wappler compatible.
1 Like
There’s a feature request for TinyMCE, get your vote in for it.!
1 Like
Nice...please consider zero downtime docker options.
Anything that helps us support ourselves is welcome...debug improvements, Sentry integration, advanced logging, etc.
I'm with Cheese...no thanks.
My personal opinion is that the dynamic bindings is what saves me the most time every day...when it works. I'm happy to participate in a sprint of bug fixes in this area as to this point it feels like one thing gets fixed, and another breaks, so I stopped reporting bugs.
2 Likes
We have used it to do on the fly translations to multiple languages with API calls to ChatGPT. Primarily as one of our team is away for several months who usually do this for us. But once they are back we are reverting to good old human input. Outside of that we don't use it to do our UI nor code. Our reason is we like to understand what we are doing so we can explain it when requested to do so, and fix it if it breaks. A.I in software and Operating Systems is going to be everywhere. I don't think Wappler needs to jump onboard the bandwagon personally. It won't dissuade us from using Wappler at all if it is integrated, we'll just find the off button and disable it. Do believe it is a waste of effort though, and that effort is far better directed at improving Wappler itself and what it is already capable of, along with improvements as George has outlined.
I'm hoping that's because it's coming before 7 since it was supposedly 80%+ complete.
2 Likes
In my personal opinion, integrating a GPT should be a high priority, especially around giving it all the knowledge in the community and all the common files used to build in Wappler (i.e., knowledge of appconnect.js and its extensions, knowledge of the dmx framework, and knowledge of how custom extensions are created). This should greatly improve the speed at which we can build.
Currently, the GPTs do not have enough combined knowledge of Wappler, so using it to stub out a custom extension or do anything other than simple coding or writing SQL statements is challenging. However, I can tell you that it has saved me many multiples of time and helped me solve complex SQL queries that I wouldn't have been able to write myself.
Even if this means a custom GPT that we can use that isn't baked into the UI, it would be a great step in the right direction.
We do not want our code base used as training data for any A.I so an opt out is essential.
User-agent: GPTBot
Disallow: /
I was only talking about common Wappler files and community posts used in a custom GPT that could help guide us in building apps.
Yes that is exactly why I said “maybe”
Because we had a very intensive research on the latest AI technology and even had a university graduation done with research dedicated about AI integration in Wappler.
However all those research pointed out that the current state of AI, even the most advanced LLMS coming out now every day, doesn’t solve any problems for Wappler and do not improve productivity.
This is because they are not there yet. And the knowledge they mis is being fit by “logical” hallucinations.
So while the AI systems can pretty easy generate a template for a basic web sites and apps ( boy Claude AI even knows Server Connect code and App Connect components and expressions! ) they generate a 80% valid code.
So you spend like 10x more time afterwards trying to convince the AI with follow up prompts to generate the right code.
So currently the Wappler own visual tools produce much more productivity than the AI in the next few years.
That is for coding of course.
Regular text generation and writing skills of the current AIs are already very good! So we can use those.
4 Likes
I actively use this tool in my favorite application:
If Wappler generates logic in the context of the Wappler in visual workflows editors functionality, this will be a good help.
Another thing is that there are more important things to do!
I m working on normalization of some municipalities’ archives. AI can help when the street is written in different ways like (in italian)
Via D’ Amelio, which can be written as:
Via D’Amelio, Via d’Amelio, Via D’amelio, Via D Amelio, D’Amelio Via etc
So for example some AI function can help this kind of job. I’m trying ChatGPT and seems giving good result.
AI is good for such kind of use and of course text generation. I think that AI integration on Wappler should be done on this side. Not to generate code. But to give some kinds of tools or possibility to create the tool to the user of Wappler
1 Like
“I suggest that the team at Wappler take a good look at Astro, which is currently highly regarded. It is an excellent tool for frontend, optimizing everything to make it faster. They could take some tips from Astro to create a Webpack or something similar.”
3 Likes
Have you tried fine-tuning a model with Wappler knowledge and examples?
2 Likes
We did fine-tune it with a lot Bootstrap knowledge, it did work but the more we trained it the dummer it got. So it was a fine line to follow.
Our hope is for new models like Claude 3.5 to be fine tuned and given exact context of all the app connect and server connect rules that we already have, so the AI knows exactly what it can produce and doesn’t hallucinate any more.
However the current context window is just 200K tokens so that will fit the rules but they will be no room left for the generation.
So still waiting for larger context and improved fine-tuning possibilities.
4 Likes