Yeah, sharing is one of the big problems with Immich. You either need to expose your whole instance to make share links publicly accessible (and there's no way to secure user accounts with 2FA), or create user accounts for the people you want to share with and/or invite them and set up your VPN for them to access the shares.
There's https://github.com/alangrainger/immich-public-proxy, which is a weird workaround (it's a separate gallery frontend that syncs with your instance to only expose this specific service, so you can keep your Immich instane private).
Is there a way to sponsor a feature request to improve the sharing and user auth story? Passkeys could be used here (versus MFA/2FA), because if the passkey is lost, you can simply delete it and have the known user go through the onboarding flow again to create a new passkey.
I don't think the Fellowship qualifies as "working for STF". These people are being paid/supported for the work they do in their own projects. They're not being paid to work on/work for STF to help them steer/analyze/control the other projects that are being funded.
EDIT: OPs comment can be understood in multiple ways: Should STF hire Developers to work on their own projects, or should they hire Developers to understand/guide/manage the funded projects in a better way.
Another very american understanding of censorship. In Germany (where Codeberg is registered), censorship means censorship by a government. A private entity (Verein) deciding what and how they want to run their business is typically not considered censorship (maybe the term "private censorship" could work), but just "freedom of doing business". So if you want to argue FOR freedom, you could also turn this argument around and support the fact that a private entity should be able to decide whatever the fuck they want to do with their own service offering.
It has all the backends you want. Just a script rebuilds it. Then I just put it in a docker compose and whenever we want to update it, we pull the repo and rebuild it.
I just a models.ini like:
version = 1
```
[*]
n-gpu-layers = 999
threads = 28
flash-attn = on
mlock = off
mmap = off
fit = off
warmup = on
jinja = true
direct-io = on
cache-prompt = true
cache-reuse = 256
models-max = 1
models-autoload = 1
timeout = 600
ctx-checkpoints = 256
cache-ram = 0
mlock = on
kv-unified = on
no-context-shift = on
swa-full = on
[Q36-35B-A3B]
model = Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf
mmproj = mmproj-BF16.gguf
rope-scaling = yarn
rope-scale = 1.14441
yarn-orig-ctx = 262144
override-kv = qwen35.context_length=int:1000000
ctx-size = 300000
reasoning = auto
reasoning-budget = 421
reasoning-budget-message = you're reasoning too much; compress context and use subagents to investigate further
spec-type = draft-mtp
spec-draft-n-max = 3
batch-size = 16001
ubatch-size = 16001
cache-idle-slots = on
temp = 0.6
top-p = 0.95
min-p = 0.00
top-k = 20
repeat-penalty = 1.1
presence-penalty = 0.0
```
Using subagents, you keep the context size from bloating, and add tooling for the context compaction and it can keep looping. You probably want to clean up opencode's agent; back in the "Prompt Engineering" days there's a lot of unnecessary fluff. The goal of a prompt should be to get the model in the correct starting place, and not, try to give it everything it ever needs to do what you want it to do.
There's https://github.com/alangrainger/immich-public-proxy, which is a weird workaround (it's a separate gallery frontend that syncs with your instance to only expose this specific service, so you can keep your Immich instane private).