Self Hosting Issues Getting Started - 404 (Sign up not going through)

[doronkatz]

Hi folks
Starting out new here. Trying to install on my mac mini via docker. My stack also includes caddy and tailscale, both of which I run using local standalone instead of docker versions. They work fine with my other apps.

When i go to sign up and enter my details it wont let me continue. It gives me a 404 error.

My .Caddyfile is:

...
## Rybbit Analytics
:443 {
    bind tailscale/analytics
    tls {
        get_certificate tailscale
    }
    tailscale_auth
    
    handle /api/* {
        reverse_proxy localhost:8101
    }
    
    handle {
        reverse_proxy localhost:8102
    }
}

My .env is:

# Variables configured by setup.sh
DOMAIN_NAME=analytics.tail7c7620.ts.net
BASE_URL=https://analytics.tail7c7620.ts.net
BETTER_AUTH_SECRET=<REDACT>
DISABLE_SIGNUP=false
USE_WEBSERVER=true
HOST_BACKEND_PORT="127.0.0.1:8101:3001"
HOST_CLIENT_PORT="127.0.0.1:8102:3002"

docker-compose not altered..

Logs are:

 docker logs client 
   ▲ Next.js 15.3.1
   - Local:        http://localhost:3002
   - Network:      http://0.0.0.0:3002

 ✓ Starting...
 ✓ Ready in 144ms

 docker logs backend
Running database migrations...

> [email protected] db:push
> drizzle-kit push --config=drizzle.config.ts --force

Reading config file '/app/drizzle.config.ts'
Using 'pg' driver for database querying
[⣷] Pulling schema from database...
[⣯] Pulling schema from database...
[⣟] Pulling schema from database...
[✓] Pulling schema from database...
[i] No changes detected
npm notice
npm notice New major version of npm available! 10.8.2 -> 11.3.0
npm notice Changelog: https://github.com/npm/cli/releases/tag/v11.3.0
npm notice To update run: npm install -g [email protected]
npm notice
Starting application...
[2025-05-13T04:52:14.987Z][WARN][@clickhouse/client][Config] "host" is deprecated. Use "url" instead.
2025-05-13T04:52:15.023Z WARN [Better Auth]: Social provider google is missing clientId or clientSecret
2025-05-13T04:52:15.024Z WARN [Better Auth]: Social provider github is missing clientId or clientSecret
GeoIP database loaded successfully
Starting server...
Loaded 0 sites into site config cache
Initializing cron jobs...
Cron jobs initialized successfully
2025-05-13 04:52:15.286+0000 - info - Server listening at http://127.0.0.1:3001
2025-05-13 04:52:15.286+0000 - info - Server listening at http://192.168.164.4:3001

Attached is UI error

[Ultragian]

Same issue, only I don't use Caddy nor Tailscale, just accessing the Web-UI directly (doesn't work through NPM either).
There is nothing in the logs about it either.

[smileBeda]

Can you check if #135 solves your issue?
You need to rewrite /api to /

[doronkatz]

I haven't tried it yet. Is there a branch or tag I can look at

[smileBeda]

No, you've to use a proxy and rewrite as of now.

[lunibo]

Rewriting /api to / does not seem to fix this problem:

[smileBeda]

That's a different error, actually.
This one for me never happened

Fairly sure it's also related to some routing issue though, and should be somewhere in your setup, since I can't see that problem.

Could you maybe share the exact way you set this up? We can try to replicate and find solutions / fix

Thank you!

[doronkatz]

Hmm. I've posted env and docker compose. I use caddy and so I posted caddy file. I also use Tailscale.

What can I post to help with understanding the setup ?

[smileBeda]

Can you share your current caddy file please?
it should look somehow like this:

{$DOMAIN_NAME} {
    # Enable compression
    encode zstd gzip

    handle_path /api/* {
        // REWRITE RULE FOR /API to / HERE ONLY

        reverse_proxy backend:3001
    }


    # Proxy all other requests to the client service
    handle {
        reverse_proxy client:3002
    }

} 

I notice you map Docker to host port in your env. Could you also try to remove that, and instead map directly via caddy to the docker container?
This should work, since docker opens the port anyway on the host

For reference my NGINX backend block:

location /api/ {
    rewrite ^/api(/.*)$ $1 break;
    proxy_pass http://rybbit_backend:3001;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}

(not sure abut the exact equivalent caddy syntax but I could imagine the trailing slashes are as well important in Caddy as they are in nginx). Not sure caddy can process the same type of regex, but it should help to understand where and how to rewrite)

[doronkatz]

My Caddyfile is above in my post, the first thing.
Can you explain to me the map docker to host part? Where is that?

[smileBeda]

If the Caddy file you shared is still as in the first post, it is not rewriting /api to / as it should, due to #135
That is a issue that eventually will be fixed, but right now, it appears required
And, that rewrite should be only in the backend block, not in the front end block.

Can you explain to me the map docker to host part? Where is that?

In your .env, you have:

HOST_BACKEND_PORT="127.0.0.1:8101:3001"
HOST_CLIENT_PORT="127.0.0.1:8102:3002"

that is not how it is documented in the .env here: https://github.com/rybbit-io/rybbit/blob/master/.env.example

So, I am suspecting you did that to map host:docker ports, however, I can imagine this creating some conflicts this way.
Try to directly route the traffic in caddy as the caddy file shows: https://github.com/rybbit-io/rybbit/blob/master/Caddyfile
Of course, instead of container names, you can use HOSTIP:Port. This should work, since Docker exposes the port on the host in all cases, unless you do not expose ports at all in docker, however, since you said you did not alter docker-compose, it fairly certain does expose it.

Hope that makes sense!

[doronkatz]

I cant figure it out :( Ill wait till the fix I suppose. I tried cleaning up Caddyfile according to the link but nothing. Happy to send mt docker compose again and see if there's anything there that I can do.

[smileBeda]

Try this caddy file:

{$DOMAIN_NAME} {
    # Enable compression
    encode zstd gzip

    handle_path /api/* {
        # Strip /api from the path
        rewrite * /{path.*}

        reverse_proxy 127.0.0.1:3001
    }

    # Proxy all other requests to the client service
    handle {
        reverse_proxy 127.0.0.1:3002
    }
}

(this is assuming you run everything on the same machine - caddy, rybbit back and front end)

Then, change your .env to this:

# Variables configured by setup.sh
DOMAIN_NAME=analytics.tail7c7620.ts.net
BASE_URL=https://analytics.tail7c7620.ts.net
BETTER_AUTH_SECRET=<REDACT>
DISABLE_SIGNUP=false
USE_WEBSERVER=true
HOST_BACKEND_PORT="127.0.0.1:3001"
HOST_CLIENT_PORT="127.0.0.1:3002"

Then, restart the docker (I am not sure if you use pure docker compose or the setup.sh scripts, both offer ways to restart / rebuild the containers)

Let me know if that works!

[doronkatz]

Yep tried same thing no luck. Really odd. I didn't edit docker compose though not sure if it's needed

[smileBeda]

Sorry, I notice something else in your env:
USE_WEBSERVER=true

If you do that, I think you need to use their Caddy - that is, it appears to me you probably started off with the setup.sh script, as shown here? https://www.rybbit.io/docs/self-hosting#4-run-the-setup-script

That creates an .env file that will set the server to true, since it includes Caddy.
If you want to change that and user your own caddy, you have to use what they describe here https://www.rybbit.io/docs/self-hosting-advanced#using-your-own-web-server
Otherwise you basically get 2 caddies

What I suggest:

1. Decide on which method you want to use (sounds like you want to NOT use their caddy, but your own, so use https://www.rybbit.io/docs/self-hosting-advanced#using-your-own-web-server)
2. Make sure your caddy uses the proper config and rewrite
3. Delete all env, and generated folders/etc and start anew.

Honestly, if you do not want to use their inbuilt server, I suggest https://www.rybbit.io/docs/self-hosting-advanced#docker-compose-setup-with-npm-without-exposing-ports - but that will require you to use the proxy in docker, so you can have it in the same network.

[antoniolago]

A suggestion would be to add an environment variable to backend "PREFIX_PATH=/api" and add this to all routes (if not provided follow the current behavior) instead of needing to rewrite paths (which can get complicated depending on environment)

[doronkatz]

A suggestion would be to add an environment variable to backend "PREFIX_PATH=/api" and add this to all routes (if not provided follow the current behavior) instead of needing to rewrite paths (which can get complicated depending on environment)

Thanks how do I do that

[antoniolago]

A suggestion would be to add an environment variable to backend "PREFIX_PATH=/api" and add this to all routes (if not provided follow the current behavior) instead of needing to rewrite paths (which can get complicated depending on environment)

Thanks how do I do that

Actually that suggestion was meant for the maintainers as it would be necessary to implement into the backend's code itself, or maybe you can try if you're feeling like it via Pull Request, anyway sorry for the confusion lol

[smileBeda]

A suggestion would be to add an environment variable to backend "PREFIX_PATH=/api" and add this to all routes (if not provided follow the current behavior) instead of needing to rewrite paths (which can get complicated depending on environment)

Rest assured this is definitely a BUG, not something that is intended to be "this is how it has to be done"
The workaround here discussed simply avoids everyone waiting until any eventual fixes are implemented.

[doronkatz]

Thanks I’m not in a rush so I can wait for the fix unless someone wants to create a PR

[smileBeda]

In this case, @doronkatz - you should be able to follow #135 as it would be the main/same issue as this one.
Also note, there has just been a release to Caddyfile right now, that however will not work, see #250 :)

So don't copy that just yet :D

[github-actions]

Assuming the original need was handled, this will be automatically closed now.