TL;DR
Average grade over long routes hides nasty short climbs. I wrote a script that calculates gradient every 50m and color-codes the elevation profile, so you can actually see the painful bits.

I’m training for the Camino de Santiago on a bike. I live in Barcelona and only have a single-gear bike, which means steep hills = death for my knees. So I’ve been looking for good route-planning tools that don’t lock everything behind paywalls.

This is where my research on tools and apps to plan routes began. I want to avoid the usual suspects with paywalls and have been looking for self-hosted, open-source alternatives. I found this amazing tool Wanderer, installed it on my home server, and started creating routes. This was my first attempt:

There’s a hill at the beginning, but it seemed doable — a maximum grade of 7.2%. Reality check: the hill in general was OK, tough but doable, but at the very end the climb gets insane. After doing the route, there were several things I wasn’t pleased with; for instance, one bit of the route seemed too dangerous. But for now the most worrying part was being able to plan longer routes without trusting the route profile.

Looking more closely at the hill, the last bit shows a 12.4% grade — not cool for a gear-less bike nor for my knees.

I started wondering: how can I prevent this? I dug into tools, blogs, and all kinds of information about bike climbs. After a lot of research, I couldn’t find a reliable way to know in advance whether I’d suffer or have to step off my bike and walk those bits — which, for me, defeats the purpose.

Even official routes weren’t enough. For example, EuroVelo gives official routes and categorizes them by how developed they are. That’s an awesome starting point, but still doesn’t tell me whether I could handle a long trip. I exported the GPX from the EuroVelo Catalonian Cross and plugged it into Wanderer.

The route looked fine, but the bit at the end (Spain/French border) seemed to have quite a climb. Looking at the elevation, it’s only 400 meters, and hovering the mouse showed a max grade of 4.9%. Sounds perfect. But now, having learned my lesson, I know that average grades over long distances are misleading.

I decided to take a closer look. The only problem: having to cut the route profile and extract the section of interest. So I wrote a Python script to give me all the details I needed without jumping between tools.

The script calculates the gradient every 50 meters and colors the area below depending on grade level. That way, I can now see that part of the route has several painful climbs — nothing like the misleadingly smooth 4.9% shown in Wanderer. There are even multiple red-zone chunks before the massive climb.

Zooming into the massive climb at km 255, I can see that for about 1 km the average is >10%. Definitely not doable on a gear-less bike, and probably even with the right bike my legs aren’t ready yet.

Now I know I can ride most of the Catalonian Cross except for this chunk. I also noticed that from km 80–120 there’s a nice hilly area that could be a good starting point for training before tackling tougher climbs.

By plotting positive and negative climbs, I can see what seems easier, and going backwards looks like the best option. Now I only need to get there. Using Wanderer, I found that the closest train station to km 120 is Caldes de Malavella. The whole route has train stations relatively close (in case 120 km is too much for me).

So far I’ve done 60 km in a day without problems, so doubling that could be ambitious but possible. Looking forward to seeing how the ride goes!

If you’re interested in the code, you can find it here.

If you found this useful, please follow me on the Fediverse: @luis@write.dadadock.com for more posts like this.

TL;DR
Definitely worth it. Even with a solid (not AI-focused) setup spending around €300, the return on investment is under a year, and the freedom you gain is hard to beat.

The journey begins

Sometime in mid-2022, I started playing around with the idea of reclaiming privacy and taking back ownership of my data and media. I began with a Raspberry Pi 4 and a 32GB microSD card. Now I’ve ended up with a mini PC, 16GB RAM, 2TB NVMe, and a solid home setup.

At some point recently, I wondered if I was putting too much money and time into all of this—so I decided to actually put numbers to it. Short answer: nope. Totally worth it.

My first experiments started with the Pi. Pretty simple to set up a headless local server with SSH. What wasn’t so simple was making it accessible from anywhere. And that’s when I hit the first real wall.

I didn’t know much about VPNs back then and didn’t want to pay for a solution. The idea from the beginning was to avoid managed services and go open-source where possible. I went down the rabbit hole—dynamic DNS, local tunnels, and even briefly considering a VPN provider—until I found Tailscale. Not open source, but their free tier was honestly too convenient to pass up. I could SSH into my Pi from my phone using Termius. That was huge.

Later, I came across NetBird, which offers pretty much the same thing but is fully open source. That made me feel a bit more at ease with the whole setup.

At this point, I had around 30GB of usable storage. Not groundbreaking—100GB from a cloud provider is like €2/month. But hey, I had full autonomy over those files. A Raspberry Pi 4 + 32GB SD costs ~€50, and electricity consumed by the Pi in Spain runs around €0.50/month, so I’d break even in under 3 years. Not amazing, but still fair.

Scaling up the setup

Of course, 30GB and a Pi only get you so far. Eventually, I wanted more storage and to try hosting more types of services.

After a bunch of trial and error (and some frustrating dead ends), I ended up getting a mini PC with 16GB RAM, an Alder Lake CPU, and 512GB M.2 SSD for about €170. Later I added a 2TB NVMe SSD for another €130. All in, it’s somewhere north of €300. and I got a 512GB external drive.

Turning it into a proper headless server wasn’t trivial. It came with Windows, which I briefly tested (WSL looked interesting), but eventually I shrunk the Windows partition and installed Debian as the main OS. Getting it to avoid sleep, auto-reboot after power loss, and behave like a proper server took some digging, but once done, things opened up fast.

Adding a small VPS

Eventually, I also got a small VPS (I’ve used both Scaleway and DigitalOcean—both have good budget options). Mainly to set up WireGuard, host a small proxy with Caddy, and expose a few things publicly, like this blog (which is running on WriteFreely).

Even though NetBird was working fine, I liked the idea of being fully detached—and having the option to tinker with a bit more robust infrastructure. (currently using free credits on GCP… but VM same size of DigitalOcean's budget droplet)

Current setup and services

Right now, this is what I’ve got running at home. It covers most of my entertainment and productivity needs, and I’m slowly dipping into Fediverse territory.

Services I’m self-hosting so far:

• Immich – photo management and phone backups
  
• Navidrome – music streaming from my local library
  
• Plex – for video content
  
• Affine – for notes and planning
  
• Rocket.Chat – team/friends messaging
  
• Filebrowser – for quickly sharing files
  
• ntfy – instant notifications from scripts

This blog was the first step toward exploring federated social networks, and I’ll likely look into more soon.

The reality check

There are tradeoffs. No five-nines uptime. No real redundancy. If the internet goes out, some services are unavailable. But for personal use—and a few friends—I’m okay with that.

The upside is clear: full control, privacy, no ads, and no monthly subscriptions.

Just looking at the services above, the equivalent cloud services would easily run me over €600/year. Music, video, storage, productivity tools, chat, VPN—it adds up fast.

With my current setup (around €300 for hardware + around €5/month in electricity and VPS), I’m already saving money after 10–12 months.

Also, I can finally justify buying physical media again—vinyls, DVDs, all that good stuff I hadn’t touched in years.

What’s next?

If you made it this far—thanks! I’ll be posting more detailed write-ups for each service I’m running, plus other bits from this self-hosting rabbit hole.

Expect posts about: – how I set things up technically – other open-source tools I’m trying – some woodworking or random DIY projects – and probably more Fediverse experiments

Appreciate you reading — and if you're into this kind of stuff, feel free to follow @luis@write.dadadock.com for updates.