I packaged Forgejo.
My git repo : https://git.cloudron.io/timconsidine/cloudron-forgejo
not added to CCAI catalogue yet
Cloudron App Packagers
I packaged Forgejo.
My git repo : https://git.cloudron.io/timconsidine/cloudron-forgejo
not added to CCAI catalogue yet
@LoudLemur said in Packaging Applications for Cloudron Using AI:
I often run into the situation where so much of the AI's context has been consumed that its capabilities really start deteriorating.
This is certainly true - you're right ! 
Excellent thread.
I look forward to hearing others' thoughts, as I sure want to improve my process.
Personally, I'm not convinced about some points.
It's just my 2p, and others may have a different view.
@LoudLemur said in Packaging Applications for Cloudron Using AI:
One key thing that team cloudron could do to help with a modular step one is to provide the specifics that would remain the same for packaging in every project.
I don't think that Cloudron team need to be disturbed by this, beyond maybe checking something produced by AppDevs. It's all in the docs, barring some hard-won tips and tricks. AI could summarise it if it was asked. Or an AppDev dives in with their understanding. Maybe a Fider instance could be a good format.
@LoudLemur said in Packaging Applications for Cloudron Using AI:
To Do list.
This keeps the ai coder using the blueprint on track as it goes through the massive coding. It completes the tasks and then ticks them off its list and knows what to do next.
I find this is not needed.
Any decent dev assistant does this anyway as part of its workings. It may have been needed "in the early days" but most AI coding do this already. Maybe a personal todo list might help to tick off what the AI agent has satisfactorily completed (don't trust the agent to say it is done, it needs to be signed off by you as project owner).
@LoudLemur said in Packaging Applications for Cloudron Using AI:
Interview
You dump your initial idea and then ask the ai to ask you a single question at a time until it is satisfied it has the basics to create a spec sheet.
Hmm, seems a long winded approach. What I do is either :
I don't think we will add a new container runtime. gVisor needs one - https://gvisor.dev/docs/user_guide/install/ . We have to get grist working with pyodide somehow.
BTW, is https://www.nestybox.com/ dead? Gives me bad cert error.
@robi cool !
Reluctant to switch dev setup mid-project but next one will try it out.
Hoping to get Appflowy and ZeroNet out the door soon.
@nebulon said in Ollama + Claude Code making coding free.:
VPS with some nvidia card
I use Koyeb for this kind of thing.
It's a good proposition but I am not needing it much, so may discontinue. Others might find it helpful.
Generally I am happy with my TRAE and desktop Ollama client using one of their cloud models (to avoid pulling it locally).
But everyone raves about Claude Code, so will check them with an Ollama out.
My quick look suggested you still need a subscription for Clause Code even if you hook up Ollama model. Did I misunderstand this ?
Looks like pyodide wants deno (?) There is some confusing error.
In the meantime, I am trying to make some sense of gVisor . If we want this, it will require support from the platform via some extension to the manifest.
@girish I will check it out
Thanks for taking a look at it
Just trying to get AppFlowy and ZeroNet out the door.
Will then re-check Grist package
Pretty sure it is pyodide from memory
It has something to do with grist sandboxing. GRIST_SANDBOX_FLAVOR can apparently be pyodide or gvisor. Not sure which one we have to use, but the Dockerfile has some python install, so we are going for pyodide is it?
https://support.getgrist.com/self-managed/#how-do-i-sandbox-documents
@timconsidine I tested this package (2-3 days ago) and it installs and runs well! When I create a new spreadsheet , clicking on a cell throws some error . Do you see the same?