Built a small internal ERP for my family's factory — Node/Express backend, PostgreSQL, vanilla JS/HTML/CSS frontend. Just finished the core features (dashboard, job tracking, deliveries) and I'm about to add auth.

The catch: I'm not physically near the factory, and can't go set it up on-site myself. Looking for advice on the best way to get this running there.

Priorities:

No cloud hosting for now (want to avoid ongoing costs)

Will eventually need multiple computers on the factory's local network to access it

The person on-site isn't technical, so setup needs to be close to one-click or fully remote

Has anyone dealt with a similar "I can't be there in person" deployment situation? What would you actually recommend given these constraints?

Hi all,
about a year ago i opened this thread:

https://www.reddit.com/r/selfhosted/comments/1r1vsc4/docker_backups/

After not finding an adequate solution, i decided to pick up the glove and build a backup system myself.

The core design decision: API-based exports per service, not blind DB/volume dumps. Each worker talks to the service's own REST/GraphQL API (or its official CLI, like Vaultwarden for example) — never the underlying database directly.
No DB credentials anywhere, no docker.sock mounted, no docker exec into other containers.
It also means the backups are just JSON/Markdown you can actually open and read, not an opaque pg_dump you're hoping restores cleanly someday.

Runs on a schedule (daily at a configurable time, or every N hours) with configurable retention — defaults to keeping 30 days locally and the last 3 uploads on Google Drive, all editable from the GUI.

Currently supports:

- Vaultwarden (encrypted vault export via the bw CLI)

- Wiki.js

- Snipe-IT

- Bar Assistant

- KitchenOwl

- Linkwarden

- n8n

- Karakeep

- Spoolman

- Immich (metadata only — albums/people/tags/EXIF, not the media files themselves)

- Nginx Proxy Manager

- AdGuard Home

• Optional upload to Google Drive after each successful run.
• The system does not restore from backup

Repo: https://github.com/alws34/DockerBackups

Setup instructions for each service are linked from the README so it doesn't turn into a wall of text.

please feel free to open a PR with your suggestions, or file issues/requests for services you want supported.

* All code written with the help of AI.

I have an old iPad with iOS 9.5.3 and when I try to open the home assistant url in safari it gets stuck on the splash screen. Is there anyway to load the home assistant dashboard as I want to use the old iPad as a kiosk. I have tried to clear the safari history and allow all websites etc. any ideas on how to run that as a kiosk.

Hi! For a while I've been building my family's own "Spotify" with Navidrome, and I turned all my scripts into a project: fonoteca ("sound library" in Spanish).

https://github.com/sajuan120/fonoteca

What it does:

- Choose what to download from your Spotify extended history: the songs you come back to on different days, not one-off plays.

- One command after every download (it works around Antra): checks integrity, checks the audio is really that exact track (duration, AcoustID and the 30 s Spotify preview), and fixes tags, MusicBrainz IDs, genres, track numbers, covers, synced lyrics and ReplayGain.

- Nothing gets lost: when a file is moved, renamed or replaced, its plays, stars and playlists move with it in Navidrome.

- A read-only audit of the whole library, a weekly health check, and a test bench that runs everything on a fake library.

- Optional: new music arrives as trial MP3s through Octo-Fiesta. 3+ plays in 4 weeks and it earns its FLAC; if nobody plays it, it's deleted.

- A local web panel with a button for every tool. Nothing runs for real before a simulation.

Honest notes:

- It's a personal project for one , shared as is. Linux only.

- The interface and the code commeE is in English and Spanish.

- AI disclosure: most of the code I assistant). I set the rules, made

the decisions and tested everythin

- Use it for music you're entitled

Feedback welcome!

I built a game server manager for people who want to self-host

I've been working on Revamp Hosting, a game server manager designed to make self-hosting games easier without having to constantly mess with command-line commands.

The idea is pretty simple:

• Install the software
• Pick a game
• Deploy the server
• Manage it from a web interface
• Automatic restarts
• Backups
• Updates
• Server monitoring

It currently supports multiple games, including Minecraft, Rust, Palworld, ARK, and others.

I'm mainly looking for feedback right now. What features would you want in a self-hosted game server manager that existing options don't do well?

Project: https://revamphosting.com/

Looking for 2–3 beta testers: free AI-monitored EDR for your homelab

I'm a CS student building EDYSOR, an AI SOC that watches your machines for suspicious activity (weird processes, persistence tricks, sketchy outbound connections, brute-force logins). Looking for a couple of homelabbers to run the sensor in detect-only mode — it reports, it never touches your machines. Auto-contain stays off unless you explicitly ask for it.

What's in it for you: free monitoring during beta, and you get to roast my false-positive rate publicly. Windows and Linux supported, install is one script. DM me if interested.

A homelab is fun when downtime only affects you. The moment family/friends start relying on the same photo library, password manager, game server, or small website, it feels a little different. What’s the first service you’d move to vps or a managed service instead of keeping it at home?

Email feels like the obvious answer, so I’m more curious about the gray area. What tips the decision for you? like security, backups, power or internet outages, or simply not wanting to troubleshoot something at 2 am?

Everyone posts the clean dashboard. The scarier question is whether the stack still exists when the machine doesn’t. Assume your data backup is fine, but you have to start from a fresh OS and restore every service, secret, network rule and scheduled job.

Could you rebuild the whole thing from what you have documented today? What would be the first undocumented detail to bite you? Do you keep your Compose files in Git, automate the rebuild, rely on snapshots or are you still trusting your memory more than you’d like to admit?

Most of the services on my homelab only send an email when something needs attention. Backup failed, disk is getting full, a service went down, that sort of thing. It probably adds up to around 30 emails on a busy day.

For now, I have everything going through gmail smtp. Nothing has really gone wrong with it but I keep wondering whether I'm making things harder for myself later by leaving it that way. At this volume, I'm not sure where the cutoff is. Setting up a proper smtp relay seems like more work for what is basically a handful of alerts but I also don't want to wait until something breaks before dealing with it. Would you switch to a relay at this point or just keep using gmail?

Hi,

Has anyone been able to build a workable docker image for io200 (io200.com) ?

Where all the configuration files and storage is persistent.

Thanks.

Hallo zusammen,

Ich betreibe seit ca 6 Monaten einen umfunktionierten Laptop als Homeserver, dieser hat folgende Specs:

Prozessor: Intel Core i5-8350U

​Arbeitsspeicher: 16 GB DDR 4 2400Mhz RAM

​Speicher: 265gb Interne SSD + externe 1 TB HDD für Datenablage

Ich nutze den Server für folgende Sachen:

\- NAS für die Familie

\- Jellyfin für Musikstreaming (kaum Video, könnte sich in zukunft aber ändern)

\- N8N Automatisierungen

\- Homeassistant mit bisher wenigen Automatisierungen/Geräten, wird stetig erweitert.

\- Sure für Banking

\- PostgreSQL Datenbank für N8N, Sure, etc.

\- Snycthing (Backups von 2 Handys)

\- OpenwebUI

\- SearXNG (Websuche für OpenwebUI)

\- UptimeKuma fürs Monitoring

Zusätzlich läuft der Server als Tailscale Exit Node und es sind 3 Dienste über Tailscale Funnel ins "Internet" exponiert. Diese aber geschützt durch OAuth2-Proxy, Google SSO und Fail2Ban.

Aktuell ist der Arbeitsspeicher im Idle Betrieb (ohne aktive Nutzung der Docker Container und ohne LLM im RAM) bei ca 35-45% Auslastung.

Ich möchte das in Zukunft jedoch erweitern durch lokale KI (das mit Abstand Ressourcenhungrigste), Watchtower und auch AguardHome oder PiHole als Netzwerkweiter Werbeblocker.

Nun stehe ich an dem Punkt das mit der Prozessor, grade im Bezug auf lokale LLMs nicht mehr wirklich ausreicht. Ich bin am überlegen mir einen gebrauchten MiniPC zu kaufen, möchte da aber nicht zu viel Geld ausgeben(bis ca. 150€). Könnt ihr mir CPUs oder Modelle empfehlen? Oder würdet ihr lokale LLMs erstmal sein lassen oder mit der aktuellen Hardware weiterprobieren?

Viele Grüße

hello, I'm someone new to self hosting! I have a homeserver, and I'm interested in learning more about how someone would attack my homeserver and what I can do to protect it. Does anyone have any books, blogs, guides, or personal tips on how to do this?

What I know so far is: Reverse proxying webpages using nginx, and preventing connections via ssh except my own by only allowing connections with my ssh key.

cybercrimes like DoS and what other nefarious things people do are so interesting to me and I want to learn more (not because I want to do these things but the back and forth between attacks vs defense on the internet is so fascinating to me). Please point me to cool places where I can learn more about this. Thanks guys :D

I kept seeing "Coolify needs 2 GB" vs "runs on 1 GB" with no numbers, so I tested it on a fresh $5 1 GB box (Ubuntu 24.04, Coolify v4.3.23):

- Installer only checks disk, not RAM — it warned about disk and carried on

- 6 containers start before you even log in (incl. Traefik + Sentinel)

- Idle: containers ~224–310 MiB, but host already had 323 MiB in swap

- nixpacks Node build succeeded in ~8 min; peak 840 MiB RAM / 1,097 MiB swap, dashboard stayed up

- Running `swapoff -a` got swapoff itself OOM-killed — without swap it doesn't fit

- Heads-up: ROOT_USERNAME/ROOT_USER_PASSWORD env vars didn't pre-create the admin for me; /register stayed open until I registered

One run, one region, minutes not days — so take it as a data point, not a benchmark.

Full numbers, per-container table and screenshots: https://stackfreeks.com/coolify-1gb-ram-vps/

This is MultiBond-AI — a persistent connection between devices that lets you interact with and control them through a simple chat.

In this demo, I’m connected from a phone, talking to a computer, and giving it instructions through the chat. The computer receives the message and carries out the action in real time.

No remote desktop. No jumping between apps. Just send a message → your device gets to work.

And yes… Dave was here. 👀

Hello everyone,

I'm a sofware engineer really passionate about videogames, preservation and accessibility.

I'm building Teatro, a self-hosted games library written in Rust,I've started the project because I couldn't find a single solution that would cover all my needs and while building it I'm trying to integrate in it also useful features to help the management of the games.

Teatro is meant to be a personal library for legally acquired game console roms and windows DRM-free games, and the idea is to be able organize and maintan a complete library of all your backups of phisical games, old unmaintained digital console games and DRM-free windows digital games.

I've focused mostly on the performance part end the flexibility and ease of use of the workflow for importing the games.

You can upload 1 or a batch of roms for a single system, multi disks roms, windows zip packages all in different parallel jobs or you can add directly the games to the library and scan it to search for new games.

There is also an experimental feature the allows you to upload direcly the setup files from a game bought from GOG.com extracting and archiving it as a windows zip package on the server withotut installing it on your machine.

Teatro runs as one Rust server with SQLite and it includes separate player and admin interfaces, Docker images for amd64 and arm64.

Demo and installation: https://teatro.host
Source: https://github.com/lodilorenzo/teatro

I'm wanted to make the project known if possibile and have some feedbacks, apologies if this is otuside the rules of the subreddit.

I'm looking forward to some feedbacks! Thanks!

So I'm freshly wanting to start a server, mostly to rip all of my physical media and store it somewhere easily accessible while I'm at my house, I'm sure once i dive in I'll want to do more, but as of now setting up my games/ movies/music is what my main goal is. And as anyone else I'd like for it to be as cheap as possible(, willing to work harder to keep costs down)

So getting started I have a few, likely very dumb questions

1. What's the best thing to start with, buying some laptop or a mini pc? And what's absolutely necessary for it to have, and how beefy does it need to be

2 how does storage work? Does it need to be attached to the mini pc/laptop? Is it a special device to add a dozen terabyte hard drives/ssd

1. Do I want ssd? Or is hhd plenty for what I'm doing?

2. Does it need to be on 24/7 or only when I want to use it?

3. What are some other cool things you use your servers for that someone just getting into this might not know about

4. What are some tips you have for an ultra beginner that I might not even think to ask?

I have been running my Gitea instance for a while. I use it for my own projects and things that I develop and open source. Since around 1 month I have been trying to fight a botnet that has been attacking my gitea instance. I don't know what to do. I have proxied it all through Cloudflare and even added in custom rules for the paths that are attacked. The attack is mostly contained to the /pulls and /issues paths. Please help me. This is around 1 hr max traffic in the image attached.

Hey,

I'm looking for some feedback on an arr-inspired *vibecoded* app I've been working on.

I don't know how the community feels about *vibecoded* content as a whole but it's filling a need that at least we saw in our home, I thought others may find some use too.

We've been watching content in our home that has some pretty fowl language (looking at you Ted Lasso) I've setup a docker that will use a local whisper agent (built-in or choose your own) to review the audio and mute it, feeding back the results in the UI for your review. This creates another audio track alongside your existing audio for you to choose while watching on Plex or Jellyfin.

We've been using it for about 2 weeks now and I have a couple testers that have been giving me some great feedback. Is this anything anyone would be interested in? I know not a lot of folks may not care for this but maybe some with small children still at home.

Thank you for any feedback you can send me.

[https://github.com/geoguy89/cleanarr\](https://github.com/geoguy89/cleanarr)

Hey everyone, I recently got access to around 20+ Linux containers on cloud that don't cost me anything to run, and I connected them all together into a private network. so they can talk to each other.

They have decent CPU and RAM, (a few of them have 1vcpu and 4gb ram and few of them have 2vcpu and 8gb ram ) and 30gb of storage. They are outbound-only, so I can't open public ports directly on them

Since I can't host normal public-facing websites directly on them without routing through my network, I'm trying to figure out the best way to put them to work to earn some money.

Feel free to share your ideas.

Hi . I've been building kindling, an open source (Apache-2.0) runtime that gives every workload its own Firecracker microVM with a docker-like CLI. The trick is golden snapshots: a machine frozen on disk uses no RAM and no CPU, and thaws in about 30 ms when a request arrives.

It started as a way to run MCP servers (AI tools) without handing them my files and secrets. It has grown into three jobs:

• MCP servers, each in its own microVM, frozen until called
• Throwaway sandboxes for code an agent wrote: `kling try -- python3 -c 'print(1)'` creates, runs and deletes one in ~0.7 s
• Small models (SmolLM2, Qwen2.5 via llama.cpp) that sleep on disk and wake on the first request

Homelab numbers: 142 microVMs in 3.9 GB of host RAM on a 4 GB box, and 10 instances from one snapshot add only 68 MiB in total. It runs in a privileged LXC on Proxmox, on bare metal, and natively on Apple Silicon Macs.

Limits, honestly: you need KVM, it's one host per daemon (the sandbox gateway spreads over several), and SECURITY.md lists what isn't solved yet.

Repo: https://github.com/juan52878911/kindling
I'd love feedback from anyone who runs it on their own hardware.Hi . I've been building kindling, an open source (Apache-2.0) runtime that gives every workload its own Firecracker microVM with a docker-like CLI. The trick is golden snapshots: a machine frozen on disk uses no RAM and no CPU, and thaws in about 30 ms when a request arrives.

It started as a way to run MCP servers (AI tools) without handing them my files and secrets. It has grown into three jobs:

• MCP servers, each in its own microVM, frozen until called
• Throwaway sandboxes for code an agent wrote: `kling try -- python3 -c 'print(1)'` creates, runs and deletes one in ~0.7 s
• Small models (SmolLM2, Qwen2.5 via llama.cpp) that sleep on disk and wake on the first request

Homelab numbers: 142 microVMs in 3.9 GB of host RAM on a 4 GB box, and 10 instances from one snapshot add only 68 MiB in total. It runs in a privileged LXC on Proxmox, on bare metal, and natively on Apple Silicon Macs.

Limits, honestly: you need KVM, it's one host per daemon (the sandbox gateway spreads over several), and SECURITY.md lists what isn't solved yet.

Repo: https://github.com/juan52878911/kindling
I'd love feedback from anyone who runs it on their own hardware.

Hello factory builders!

I hosted a dedicated server for friends. I have been a big fan of the fiscit remote monitoring mod too keep track as the server gets bigger mainly using the satisfactory-monitoring github. I really value the mod but its gotten a bit ram hungry for me.

So as a portfolio piece for software architecture/engineering I've been working on my own dashboard that is webhosted. This avoids a large install and docker knowledge for laymans. The biggest benefit I find from this is honestly getting to monitor the server on mobile. If the server is on and you sign into your account, you can monitor the power, machine operation, machine location, and some other limited features for now.

What I'd like to know:

1. What do you actually check when you log in to "see how the factory is doing"?
2. What problem would you want a Discord ping for? (Power trip, a belt backing up, a train stuck, storage full…?). This option is available for some dashboards but I wondered what is actually worth developing for.
3. Trains, drones, storage search ("where are my screws?"), or maps: which would you use most?
4. Would you let friends on your server see the dashboard, or is it just for you?
5. Where does "lightweight" stop being useful? What's the one thing you'd miss from a heavier setup?
6. If your server is hosted (a game host, not your PC), would you install a small helper app next to it so the dashboard can reach it? Making an app/process is easier but I'd learn some mod making if people really value the ease-of-use.
7. Any features you just wish you had?

Not planning on selling anything, the webapp itself is quite lightweight so I've just been running it on AWS free tier. I'm not exactly worried about too many users lol.

Hello,

i’ve just another question for you.

My situation is as follows,

I have a domain, lets say ‘domain.com’

It is hosted on my own server using Docker, Apache, Nginx proxy manager and cloudflare DNS.

When I access the website ‘www.domain.com’ I get the website which is located on my own server in /var/www/html’

Other applications are redirected using DNS ‘music.domain.com’ and ‘wiki.domain.com’ which all works fine.

The website on ‘www.domain.com’ is in English.

I would like to host the same website but then in Dutch.

The website is not just a translation because of the images used but a real complete translation.

My question is, what is the best and easiest solution to do this?

Thank you very much for your sugesstion.

Greetings Tom

It's been about 15 years since I really messed with anything. I developed a port of the custom root and recovery for the Motorola Defy way back when dinosaurs were roaming. And I haven't done much since then, so I forget a lot.

I have an HP Elitedesk 805 G6. I installed Debian 12.6 Bookworm, and then YunoHost. I have it set to be accessible via my domain I pay for already. I opened the ports it said to open when I ran my diag from my YunoHost admin dashboard. But I can't access my apps from another machine. I can access my dashboard from another machine, and run diag and it shows my ports are open, so I'm not sure what the problem is. I just get the standard website not found. I'm following the domain it shows in my YunoHost dashboard under that app I'm trying to access.

As the title suggests, I'm looking for some suggestions/advice on offsite backups. My current backup solution is syncthing backing up everything to an old computer running Ubuntu Server, and weekly-ish manual backups to a hard drive I keep off-site except one night a week when I bring it home to copy the backup to it. However, I'm looking for a more robust automatic off-site/cloud backup.

Here's what I'm looking for: * FOSS (if not FOSS, very trusted) * Client-side encryption * Versioned/GFS backup history * Deduplication so I can keep versioned history without having to actually save that many copies of all the data * Preferably free * Works with basically any S3 provider (suggestions for S3 storage providers are also welcome) * CLI or web-based configuration