[Tool] SFOS Dev Reference Skill - Validate Harbour Compliance, Catch Mistakes, Follow Silica Rules

Platform Development
21

Thanks for commenting, is nice to see people with your way of thinking, should be the normal of course.

But there is something is important, the dev world is not really like this, usually people help each other, motivates (you can see the comments of other devs here a lot more positive, and is obvious he is a experienced sfos dev), more external people they always have their opinions on if they like X or Y, but is not very usual to see devs hating each other in public, or talking sarcastically like I saw here, we’re devs, people is usually supportive, and in general the most seniority level is always helping the people with less experience or trying to open the development to more people, this has been a foundamental, specially on Linux, like is the case.

I don’t have too much time working with SFOS or participating in this enviroment, but in that very short of time I saw this show happening 3 times, not only in the forum but in another channels, people attacking each other, not to me, but to some new people trying to do something for the first time, this is basically gatekeeping, hidden as “AI hate”, that is not really that because they don’t even know what was made with AI in a project, and what not, is the tool/excuse they use to release hate and frustration, fear of being replaced, etc.

Nowadays in almost all the medium/big size companies the usage of AI is promoted, this could be good or bad, but is becoming a new standard for increase productivity, detect issues before even the software reaching the QA phase, etc. Any dev will know this, using AI as a tool is not a direct mean of vibecoding or that you hate traditional devs/methods/practices, so, judging people because they use AI or not, sounds a little bit weird, specially coming from a dev, right?

Maybe they have some benefit gatekeeping? maybe they have some projects, they receive donations, they like to see their world being small and continue being theirs? small world small mind? Can you imagine thinking so small and at the same time creating apps for the masses? that’s the case of Android or iOS for example, do you really think devs have time to hate like this to each other? people is out there hunting for ideas, improving their skills, using all the tools they have in their hand to make software better and deliver faster.

In any single enviroment, specially on mobile, the diversity is a benefit, even if you’re that special kind of hater, having more people doing more, creating more, give you space for show your work to them, to the new users, to highligh your praised work in comparison to them, and of course, more devs, more projects, will always attract more casual users, is a win-win situation for everyone, for jolla, for SFOS, for the devs, for newcomers, for literally everyone here.

Instead of that, they prefer to release hate publicly like this, so, in a personal level, I’ve a work that take a lot of my time, I have a HW Startup that takes a lot of my time, I have some software projects that usually take less time, but I work on there for fun and of course I like helping the community, because I love being a dev, specifically mobile dev, since Symbian, and all of the projects I work on, produce benefit for me, in the free ones, people always support and send positive vibes, in others, I receive a profit.

If I need to chose between continue collaborating here, or continue working for Android/iOS/Web and be surrounded by positive thinking or receiving a profit, is a no brainer election for me…

The post had 50 views or less when I received the first hate comment, with less than 100 views it was 3 different people. Thankfully there is also great people who also commented, I lose my faith in the project not because of the hate, but because I think they’re too much people hating for a very tiny pool of devs and users.

I see hate in another projects I work on, is normal, is internet, is expected, but I can count them with my hand, and they’re projects with 60.000 active users, put that in the balance and you will understand why I halted the development of everything I was doing here, I deleted my first app on open repos, make the repo private for this, for that project, and for the Threema client we was working on, people would have enjoyed those projects I am sure, but I am not here to be surrounded by hate, this is my hobby, creating is what I choose to do with part of my free time, I don’t need money, I don’t need literally anything, luckily I have a very nice and comfy life, a beautiful family, I do this because is what I love to do, there is nothing more.

Why I am explaining myself so much, and writting so much? easy, if the mods don’t delete my post, then I will not let the haters get the highligh, even if they try to reduce the importance of they’re damage with “wow” or “you’re overeacting”, I will write like this so the normal people will read and understand what’s happening, maybe the mods/sfos people here will see the consequences of the hate, hidden attacks between lines, and will change something in the future? maybe yes maybe not, I don’t really care, but I did my part exposing the consequences, yesterday I took back what I was offering, this and all my other projects, we (my team and I) are not continue with our developments for SFOS, the people will have less projects to see, SFOS lose some devs, Jolla will lose money because we already sent emails cancelling our pre-orders (i really hope they cancel them and we can get a refund), and I will be more happy surrounded by more positive vibes and dedicating my free time to a more greatful and welcoming ecosystem.

At the end, maybe SFOS doesnt need more people, maybe it is good as it is right now?

2 Likes
22

1080×1485 348 KB

I’ve written a couple of sfos apps for my own niche use. One of them asks for the battery voltage of my eSkate over BLE, so suuuper niche. What do I know about BLE? Exactly f@ck-all.

But I’m a good manager, so I guided ChatGPT through the VESC open source code, via the sfos dev environment and through to an app that uses pulley menus, doesn’t break sandbox and just works and feels very native.

I would have loved to have a guide to show ChatGPT so I wouldn’t have had to explain so much about sfos myself. At the end of the day, it comes down to caring to get a good result and not a terrible LLM copy/port of someone else’s work.

Sorry you got flamed.

7 Likes
23

Does that OpenCode skill work with OpenCode running qwen coder via ollama? I only do LLM local stuff. EDIT: I think it does work with local models, but have no need for the skills since they are basic scaffolding which I’m familiar with.

2 Likes
24

I respect if you want to stop working on with Sailfish, it is your choice, but would you consider someone else to continue your work with Threema so it wouldn’t go to waste? That is definitely one quite big gap in sfos ecosystem and has been wished quite many times. I know, android version apparently works, but native is always native. I probably would have switched to Threema if native client would have been there when I made the jump to sfos. As I’m not dev, I can’t finish it and don’t really know who could, I just would hate to see work going to waste (meaning in general level, you probably still learned new things and got things out of the work you put in to it)

1 Like
25

Is compatible yeah, but is basic and probably you are a good candidate to improve it even :rofl: that was the initial idea of sharing it here at least :grinning_face:

2 Likes
26

Yeah, of course I don’t mind doing a hand off of the project to someone else, is an actually very capable client with a good future, just require some work for group chats/calls etc but the base is there and is totally functional for activating a license, adding contacts, individual chats etc, very raw in terms of UI/UX of course since is not ready, the plan was releasing it as experimental in a couple of weeks only with activation and individual chats and a more polished UI.

3 Likes
27

Where the fridge is this hate you are referring to?
The two threads about slop-forks received rightful hate. All you ever got was a “Are you sure you are not over-doing it a bit? Would you not be better off sharing effort with what already exists?”.

6 Likes
28

It might be the amount of ‘hate’ that AI gets in some domains that makes people sensitive. I’ve noticed on mastodon that the anti-AI (and it’s not even real AI!) sentiment get’s really into the hostile end of things. So much so that it reminds me of why I never used twitter.

I’m a skeptic, but I’m a tool user, so I think it’s wise to pick ones words carefully. Of course, I also ‘own’ my code, since I actually make it available to others. I don’t see a contradiction in using tools including LLMs and being personally responsible for a code base. But, I also believe that many people don’t get that. They assume that use of a tool automatically makes the author less responsible. I have to admit I’m very conservative with LLMs (only local, analysis, no direct code gen.) and I think the USERS would probably benefit from my being less conservative.

6 Likes
29

Maybe, seeing as we’ve put in some effort and you’re still here, you will re-consider withdrawing. I don’t expect you to get a ‘thick skin’ over night, but it’s a quality of the users here. We have very thick skins since we’ve been doing ‘off-boadway’ mobile OS stuff for a long time. But we mean no harm :slight_smile:

4 Likes
30

+1

Can you show us where? I can only speak for my own posts, but there is nothing implied between the lines. Of course it’s sometimes difficult to judge tone via text, but instead of going off like this, you could also just have answered my questions. Given that you’re still here, maybe there’s still time for that?

6 Likes
31

Totally, and I understand, there is actually some AI Fatigue from the community, users in general, but there is something I think is bigger, and as you say, is the anti-AI fatigue, bad redirected, from people that don’t even understand what are they looking at.

While I use AI at work, and in a lot of areas, I’m not pro AI, I wish we can come back to Symbian S60 v3 and nothing else, I don’t need anything else :rofl: and you can see this very clearly from all the disclairmers and alerts I wrote on the first post, but in a competitive market, out there people is using these tools to be competitive and fierce, companies use them to be safer, respond faster to the market, I kind of understand that the people that is not expose to that world don’t agree or they assume the worst all the time, but if wanna be surrounded by that I will be on X/Reddit and not here of course.

I really thought it was time for an european OS to do a big comeback so we stop depending of american/chinese technology so much, instead of being only for tech enthusiast on a niche small market, but if from inside there is no vision of this and we still having this kind of discussions instead of discussing ideas, tools, innnovation (at least is what I see) then maybe is not meant to be? the world out there is running, if we’re sitting maybe this our dream is not possible.

2 Likes
32

Although I understand your sentiment, I am careful about this kind of economic drive determining the rate of change. I’m one of those who did NOT go under in the Dotcom era precisely because I avoided investor driven hype behavior. Literally every major competitor of the time (1992-2002) is gone. My old company still exists having chosen to be more conservative. I think that we as a community have a conservative but stubborn streak and that is one key to survival.

That doesn’t mean that I won’t use the available tools for a good cause. I have accepted submissions made with claude. Half of them I had to discard, but the other half were beneficial for the users. To date with most projects I’ve only asked to input where I was lacking something but I will use Opencode like I use grep. I do NOT want to encourage people to ‘one-shot’ or vibe code (unless it’s just for personal use). I also would prefer to convince people to use local models (where the weights are available) so that we build out a FOSS ecosystem of tools.

But I also do not like the constant ‘slop, slop’ chants and the aggressive nature of some critics. I think the criticism is valid and necessary, but it should not discourage people from getting chops with the technology. And working to make it open and free.

6 Likes
33

Yeah those seem pretty basic, a nice tool/skill would be forcing local llm to search through the silica reference docs instead of generic ‘sailfish qml whatever github’, that’s usually not useful, glm4.7-flash wastes so much time on that. Ideally a local llm on phone would just search through /usr/share/{harbour-*, jolla-*, sailfish-*} so much useful examples there how to properly use silica components. Not sure stuffing the whole docs into context is the right way to go, maybe rag from all kinds of sfos apps would work, ideally then the llm can just pull examples it needs for given task and not bloat the context with garbage (btw seen qwen3.6 attempt a tool call (search?) to forum.sailfishos.org and got ‘no readable text’ in reply, are we blocking AIs, this is a great source of knowledge for them, would be a shame)

34

Hmmm, glm4.7, qwen3:30b are both loquacious and wast a lot of time thinking which can be muted, but they are slow even on an RTX4900 … the worst thing is that they may be able to do something immediately. But give them something simple and they fall over. Today I wrote the awk needed to transform a field in a tab delimited file while qwen 30b was ‘thinking about it’. What it came up with was also just wrong. This is what I wrote.

awk '{print substr($3,1,1) tolower( substr($3,2) ) "\t" $4 "\t" $5 "\t" $6}'

That’s shorter than the prompt and context I supplied the stupid llm.

35

Try the flash version of glm4.7, should fit whole on 4090 in reasonable quants and be pretty fast (not full context probably, but even with ram offload you should get double digit tk/s), if it only would make more sensible tool calls

36

Yeah, that is a big issue. I’m screwing about a bit now, but I generally just do it myself. It’s far too slow a process and I’m not senile yet.

37

Question, are you using OpenCode? I’ve dropped back down to ollama since opencode does not do proper copy/paste properly (ah, yeah in an ‘AI’ product. no copy paste. no x bufer. stupid shit).

38

Currently just playing with unsloth studio (I think it uses llama-server as backend rather than ollama), should work with opencode How to Run Local AI Models with OpenCode | Unsloth Documentation, I do want to test it with pi

39

You’ve got my interest here. You shall blog about your setup.

1 Like
40

Not worth a blog post it’s so basic

  1. download the ollama installer shell file, read it.
  2. Small mods (don’t install to /usr/share)…
  3. run ollama pull glm-4.7-flash:latest and then
  4. ollama run …

If you add OpenCode it’s not much more, though I can’t recommend it at the moment. Maybe one of the other tools … I’m going to try GitHub - gergap/vim-ollama: Vim plugin for integrating Ollama based LLM (large language models) · GitHub since I do everything in vim …
EDIT, the vim plugin is quite good. I need to try some reall work with it. Let’s make it write Brainfuck.
EDIT: sigh, ok, it’s NOT quite prime time. The vim plugin would be nice to use for Prompted tasks (a feature it has), but that is really buggy.

1 Like