Poll: AI Tools to help us coding in B4X

Which AI tool do you find most useful in writing and editing code in B4X?


  • Total voters
    31

Cableguy

Expert
Licensed User
Longtime User
I like Copilot.... I can have what is the closest to a self brainstorming session with it
 

Daestrum

Expert
Licensed User
Longtime User
I use Copilot - as a reference book, not to produce complete code. Also use a local Phi-3 (like a baby Copilot)
 

stevel05

Expert
Licensed User
Longtime User
I used to use Perplexity, but Copilot seems to have the edge now. It generally gives enough information to work out a solution, even if it's answer is not 100% correct, but it's getting closer all the time. I've given it permission to remember what has gone before, which helps too.
 

yiankos1

Well-Known Member
Licensed User
Longtime User
Lately, on my main job we use extensively VS code in order to write some JS. Copilot auto completion is mind blowing! I think that our IDEs have lost the "train" about AI live assistant.

I have used the open-source project with b4x copilot, but i would prefer something native from the IDE not a custom tool. Specifically, a live assistant not a "chat bot".

I think i will raise it as a WISH.
 

hatzisn

Expert
Licensed User
Longtime User
Lately, on my main job we use extensively VS code in order to write some JS. Copilot auto completion is mind blowing! I think that our IDEs have lost the "train" about AI live assistant.

I have used the open-source project with b4x copilot, but i would prefer something native from the IDE not a custom tool. Specifically, a live assistant not a "chat bot".

I think i will raise it as a WISH.

I like this idea too, but you have to take under notice that money makes the world go round. As far as my knowledge goes all AI APIs are chargeable. Unless you run the models locally which needs many resources most computers do not have to serve multiple contexts of multiple programmers. It would not be viable to @Erel. Perhaps a solution with some prepaid tokens would be viable. An alternative would be an 'add your api key' solution. There are some solutions proposed already in this forum. Check my signature (although I am now recreating my site and it is not accessible the download page shown in the video).
 
Last edited:

Magma

Expert
Licensed User
Longtime User
I am using Perplexity AI now (because i have it for free for 1 year subscription) - i was using chatgpt...
But not exactly for coding but sometimes for optimizing or getting some ideas for flow...

I think perplexity has better search and based on internet web pages...

All AI are useful...
 

LucaMs

Expert
Licensed User
Longtime User
All AI are useful...
They are very helpful in getting you to this point...

1753086602282.jpeg


😂
 

stevel05

Expert
Licensed User
Longtime User
all AI APIs are chargeable.
I only use the free version of copilot and don't have too many problems. The occasional "Can't do that right now" response, but usually when the conversation is getting long and I guess it would use too many tokens to process when it's busy. Then I condense the conversation into a new question and it's happy to continue.
 

stevel05

Expert
Licensed User
Longtime User
Another plus for MS Copilot, I have been playing with an incubator module (Pre-preview code) in JavaFX 25 and found something that appears not to be fully implemented. Copilot offered to monitor release docs and let me know if anything changes. I definitely didn't expect it to be able to do that.
 
Last edited:

Cableguy

Expert
Licensed User
Longtime User
I specifically like the fact that I can come back to old discussions and just pick it up where I left it
 

epiCode

Active Member
Licensed User
I only use the free version of copilot and don't have too many problems. The occasional "Can't do that right now" response, but usually when the conversation is getting long and I guess it would use too many tokens to process when it's busy. Then I condense the conversation into a new question and it's happy to continue.
Gemini and Qwen are coming up with 1M tokens where everyone gives 16k, 32k and 128k. It is huge and enough for most work. Qwen Instruct even runs offline if you have enough juice in your GPU.
 

epiCode

Active Member
Licensed User
No unfortunately integrated graphics, but I have played with ollama a bit, just for the experience.
you can use huggingface spaces / google colab for free (though its a bit slow for free version) but capable of running really huge LLM
 

Peter Simpson

Expert
Licensed User
Longtime User
I don't really use them, but I do play about with this particular GPT because @Alexander Stolte kindly uploaded the pdf files that @klaus created to it. When this GPT replies, not only does it use information from the forum, but it also uses the information from the pdf files.


I have played about with Gemini (it's officially the worse when it comes to creating B4X code), and also Copilot (which is very good at creating B4X code).

AS GPT is very good though, really good in fact.
 

Peter Simpson

Expert
Licensed User
Longtime User
Ilan Ilan Ilan,
i love copilot,
Don't let your wife hear you say that about a bit of tech, you could end up in a load of trouble my friend ;)

Yes Copilot is very good, I just don't really use any of them for coding, just for playing about with. But saying that, I have used the following GPT to reoptimised some extremely complicated stored procedures once for a client of mine. Their previously developer complicated things, I optimised them, but then I got the following GPT to optimise them even more. Yes I could have used Copilot to do the same thing, but I already have this MySQL GPT saved so I just used it.


@Alexander Stolte GPT that uses information from the forum and @klaus's PDF files, is also very good too. Gemini is just terrible at creating B4X code, but oddly enough it's really good at fixing broken B4X code.


Enjoy...
 
Top