ChatGPT

[eureka]

The ChatGPT post I placed in French forum yesterday (and now pushed back to English forum) only made sense and context when attached to the French OP discussion.
Therefore to avoid confusion I withdraw my pennyworth of experience on using ChatGPT to generate Actiona scripts. If anyone is interested I can restart the discussion afresh.

[Jmgr]

Sorry for that but that message was reported as off topic by the creator or the topic and it seemed to be the case to me. You could have left the content of the topic I moved though, I think it had value on its own also.

I've tried asking ChatGPT to generate a script for Actiona but it was unable to do so and somehow mixed inputs for different automation tools, like AutoIt and Autohotkey.

[eureka]

"that message was reported as off topic by the creator or the topic and it seemed to be the case to me"

But I only referred to ChatGPT because it was mentioned in French forum! Perhaps in a sibling thread. So it was not off topic in my opinion. Some OP's leave their topic dangling with no feedback on ideas. Why not report back to the contributor so tha tmore context can be added?

Never mind, I press on. When experimenting with ChatGPT with questions about an unknown tool such as Actiona I find that it helps to introduce (train) ChatGPT to corpora such as Actiona legacy sites and the forum. Indeed you can add such custom links to give extra context to your query.

I have posed questions on Actiona (and similar tools and approaches) through Phind.com (but there are other tools to try). As an example I pointed to the old book "Actionscript, the Definitive Guide by Colin Moock, publisher O'Reilly". After this homework ChatGPT converges on more relevant answers. Look at the Settings options in Phind.com (hamburger menu).

So training the particular model is key to avoiding foolish answers.

If you raise queries in French the session will be in French. But as has been pointed out in French forum there is a broader training corpus in American English. I just use DeepL to switch between languages since I am not proficient in French.

There is much potential in using Phind.com (and other like tools) as programming assistants. Just query the output by follow up questions. Phind.com should now be aware of queries I for one raised recently.

Experiment. I now use Actiona to orchestrate multiple tools. The simple "baton" used by conducter (user) to conduct an orchestra of AI and other tools. I think that users miss this new usage of Actiona.

P.S. on your point about ChatGPT confusing Actiona with other tools the user must insist that the focus be only on Actiona. You must add training context such as "I am aware of AutoIT and AutoHotKey but I insist that you focus on Actiona and Qt and OpenCV. Also consider mixing Actiona with resources such as Python. Do I make this clear?". Act as a bossy professor training ChatGPT as a student.

[Jmgr]

But I only referred to ChatGPT because it was mentioned in French forum! Perhaps in a sibling thread. So it was not off topic in my opinion. Some OP's leave their topic dangling with no feedback on ideas. Why not report back to the contributor so tha tmore context can be added?

That's fair, maybe I was a bit hasty in moving that message, sorry for that.

There is much potential in using Phind.com (and other like tools) as programming assistants. Just query the output by follow up questions. Phind.com should now be aware of queries I for one raised recently.

I do use ChatGPT every day now, but I hadn't heard about Phind.com. Looks very useful indeed.

As a side-note I am thinking of adding API actions, with one of them being an action allowing users to query the ChatGPT API (after entering an API key of course). That could allow people to copy data locally, process it in a complex way using AI and paste the result. Or generate an image depending on an input query.

[eureka]

Jmgr, if you are planning to reawaken Actiona from deep sleep then that is good news and I would urge that there be some explanation of how to use the Debugger. I have never figured it out. Could the code be debugged elsewhere?

Perhaps consider Actiona as a plugin to a supported co-editor such as Sublime Text and/or VS code.
I use Sublime Text for writing code in preference to the tiny fonts (for my aged eyes) in Actiona code editor, then copy and paste into Actiona. Sublime Text has its own syntax checking rules. Actiona just comes up with either "Syntax OK" or some syntax expectations (but not pointing out where, as does Sublime Text very well).

[Jmgr]

Jmgr, if you are planning to reawaken Actiona from deep sleep then that is good news and I would urge that there be some explanation of how to use the Debugger. I have never figured it out. Could the code be debugged elsewhere?

Well I have a "good" news: the debugger will be gone soon. It has been removed from Qt6, so when Actiona migrates to that version of Qt that feature will disappear. Someone has created a debugger for the scripting system of Qt6, but it requires private modules. I'll have another look at this once Actiona has migrated to Qt6.

Perhaps consider Actiona as a plugin to a supported co-editor such as Sublime Text and/or VS code.
I use Sublime Text for writing code in preference to the tiny fonts (for my aged eyes) in Actiona code editor, then copy and paste into Actiona. Sublime Text has its own syntax checking rules. Actiona just comes up with either "Syntax OK" or some syntax expectations (but not pointing out where, as does Sublime Text very well).

Maybe not "as a plugin", but rather "with a plugin". The issue is that my time is very limited and I barely have the time to work on Actiona itself, so learning how to write a plugin for an IDE is a bit too much for me. Contributions would be very welcome though. In the meantime you can always copy and paste the JS code into that IDE and then copy it back.

[eureka]

Referring to my earlier point that when ChatGPT gives a wrong code syntax (such as mixing up Actionscript with AutoIT, AutoHotKey, Javascript, the way around this is to train ChatGPT not to make that mistake in future sessions. I have just taught ChatGPT to remember that in Actiona the syntax Console.print should be used rather than console.log.

ChatGPT soon adapts. So persevere when you receive wrong answers and ChatGPT improves after several correcting quereis.

Answer from ChatGPT after this training:

Remember, these corrections will be remembered for future queries about Actiona.

So when first presenting a new (or older) syntax make the point to ChatGPT.

When I started, Actiona was unknown and links to website had to be given. Perhaps it might be an idea to have a webpage which is designed to train ChatGPT, such as wiki page on full scope. Or simply one master page with links to all other pages, tutorials, tips. references such as Qt. Then just point ChatGPT to this index page designed to train ChatGPT.