I’d like to thank everyone for contributing to this thread.
For the most part, I’ve avoided “AI”. A new release of Xcode taps into it, so I started investigating locally hosted models (too cheap to pay subscriptions).
Thanks to this thread, (other than an issue I’ve mentioned here) I was able to get Pinegrow connected to the models of my choice (have gone for devstral as it’s coding orientated) and doing my bidding.
Hi there Adam.
With regards to the Chatty deepseek thing…
this is interesting.
I didnt realise that… you can create a model, (in this instance using Ollama,) and… change how it behaves!
so, you could , following this video example… create
a Deepseek R1 model that follows the Claude prompt protocol, which is passed to the model with each request.
wow.
So bye bye Chatty R1, Hello Claude Style Replies… running on the R1 model.
… or… you COULD make it speak PiRaTe!
not sure how that would work out for coding but…
This video shows how to do it…
And its that simple.
I might try that later, Tonight.
and here is a link to a smaller model of
Llama 3.1 …using the Claude Sonnet Prompt which another user has created and uploaded based on this video.
Note: Llama model is currently 3.2 and this user created model is 9 months old. Also, the Antrhopic model isnt open source (apparently - correct me if wrong)
So creating an updated version of this from 3.1 to the 3.2 instead of this 3.1 older LLM model I was using (not much at all laterly) should be easy.
Note.
The text to be copied for this can be found on the Anthopic site here;
within the
<claude_info>
</claude_info>
Section
So, based on what you experienced…
, if you could find the System prompts used in those 2 models you favour, you could do the same, and create your own personality or Styled model of DeepSeek R1, to see how it delivers code to be parsed by Pinegrow.
Challenge accepted?
PS.
I just tried Devstral’s own online chat agent to ask about their own System prompt…
… it had no idea…
Suspicious!
it DID…however have the exact same info for … Claude.. so yeah, maybe dont bother
Thanks for the info. There is so much to learn here and my old-man brain is having a hard time keeping up. I do like the idea of coding in pirate-speak, though. Maybe they’ll start embedding some ANSI art like the we saw back in the BBS/Warez days 
Oh that brings back memories. I ran my BBS on an Amiga 500+ and the home phone. We sure must have annoyed people in the household back then.
Trying to recall but I think it ran on Max’s BBS software. 
. Simpler times.
I think it needs more “pirate” 
_ _ _ _
___ _ __ | | ___ _ __ __| (_) __| |
/ __| '_ \| |/ _ \ '_ \ / _` | |/ _` |
\__ \ |_) | | __/ | | | (_| | | (_| |
|___/ .__/|_|\___|_| |_|\__,_|_|\__,_|
|_|
_ _
_ __ ___ | |_(_) ___ _ __
| '_ \ / _ \| __| |/ _ \| '_ \
| | | | (_) | |_| | (_) | | | |
|_| |_|\___/ \__|_|\___/|_| |_|
_
_ __ ___ __ _| |_ ___ _ _
| '_ ` _ \ / _` | __/ _ \ | | |
| | | | | | (_| | || __/ |_| |
|_| |_| |_|\__,_|\__\___|\__, |
|___/
Oh man, I wish I had an Amiga 500. I used to love using the Amiga Video Toaster machines back in college and always lusted after them.
I’m pretty sure my first BBS experience with was Wildcat! Those really were simpler times…
I miss my Amiga. I also had a 4000 at one stage, wish i has kept it. But i sold it to the animation school i was at, for a line tester. The old days of hand drawn animation.
Good points! LocalAI looks like the most practical way since it mimics the OpenAI API, so PG could potentially connect without much hassle. Modelz LLM is also promising, but I haven’t seen a clear PG integration guide yet.
Curious if anyone here has already tested PG with Llama 3 locally?
@Zahra Well… yeah. that’s kind of what started this whole thing.
here you go 
Its written about here….
unless…..shudders you’re some form of bot.. creating random links on here…
That was simple, I used Ollama
Now need to find a good model to use, any suggestions?