Self-hosted pub.dev alternative.
The root of this repository contains a docker-compose.yml
, which contains three services used to aid in debugging PubNet
:
database
: A postgres database (user: 'pubnet', password: 'pubnet')adminer
: A webinterface for managing the databaseseq
: A useful logging service
The only service you need to start is database
, though the other services help during development.
Note
You can also use a database you set up on your host, but currently only postgres is supported.
The solution consists of a few projects, three of them compose the whole PubNet service:
PubNet.API
: provides the API used by thedart pub
command line tool to fetch and upload new packagesPubNet.Frontend
: contains a Blazor WebAssembly project to act as a frontend for the APIPubNet.Worker
: executes tasks (scheduled and unscheduled) to analyze packages and do general housekeeping
You can use any IDE you want, as long as it supports debugging .NET 8 or higher.
For the PubNet.Frontend
project, it is recommended to run it using dotnet-watch
.
This project uses Entity Framework Core with the code-first approach, so migrations are added using dotnet ef migrations add <name>
and executed using dotnet ef database update
in the PubNet.Database
project folder.
Currently, the database project expects the credentials and database name to all be equal to 'pubnet'.
If you want to contribute improvements or bugfixes, fork this repository, create a branch, commit your changes to it and open a pull request here on GitHub.
Note
In case you only need a simple, privately hosted pub package API, the API project alone is sufficient.
When you are ready to deploy PubNet
, you may want to review the OpenRegistration
setting in the backend-appsettings.json
to toggle whether anyone is able to register an account.
This setting can be changed at runtime.
docker-compose.yml
template
Create a docker-compose.yml
with the following contents:
version: '3.9'
volumes:
postgres_data:
pubnet_packages:
caddy_data:
caddy_config:
services:
database:
image: postgres:latest
restart: always
environment:
POSTGRES_USER: "pubnet"
POSTGRES_PASSWORD: "pubnet"
volumes:
- postgres_data:/var/lib/postgresql/data
backend:
image: ghcr.io/ricardoboss/pubnet/api:main
restart: always
volumes:
- "./backend-appsettings.json:/app/appsettings.Production.json"
- "pubnet_packages:/app/packages"
depends_on:
- database
- caddy
worker:
image: ghcr.io/ricardoboss/pubnet/worker:main
restart: always
volumes:
- "./worker-appsettings.json:/app/appsettings.Production.json"
- "pubnet_packages:/app/packages"
depends_on:
- database
frontend:
image: ghcr.io/ricardoboss/pubnet/frontend:main
restart: always
depends_on:
- backend
- caddy
# you can choose any reverse proxy you want, Caddy is not required
caddy:
image: caddy
restart: always
volumes:
- "./Caddyfile:/etc/caddy/Caddyfile"
- "caddy_data:/data"
- "caddy_config:/config"
ports:
- "80:80"
- "443:443"
- "443:443/udp"
Note
In this example, caddy is used as a reverse-proxy.
You can also host the backend and frontend on different ports, and publish them directly, removing the need to configure a reverse proxy entirely.
Reverse proxy configuration (Caddyfile
)
In case you want a reverse proxy, configure it appropriately (in this case using a Caddyfile):
*:80, *:443 {
reverse_proxy /api/* backend:80
reverse_proxy /* frontend:80
}
backend-appsettings.json
template
{
"AllowedOrigins": [
"https://localhost"
],
"ConnectionStrings": {
"PubNet": "Host=database;Database=pubnet;Username=pubnet;Password=pubnet"
},
"Jwt": {
"Issuer": "https://localhost",
"Audience": "https://localhost",
"SecretKey": "GenerateASecureKey!"
},
"PackageStorage": {
"Path": "./packages"
},
"OpenRegistration": true
}
worker-appsettings.json
template
Add a worker-appsettings.json
:
You could use the same appsettings for the worker and the backend as their options shouldn't interfere.
{
"ConnectionStrings": {
"PubNet": "Host=database;Database=pubnet;Username=pubnet;Password=pubnet"
},
"PackageStorage": {
"Path": "./packages",
"PendingMaxAge": "00:05:00"
}
}
Finally, start your own PubNet
using
docker-compose up -d
and access it at https://localhost
.
It should be possible to host the API using IIS, but it is not supported.
Same goes for the Worker: it should be possible to run it as a Windows service/systemd unit, but is not supported.
The frontend can be hosted from anywhere, as long as the backend-appsettings.json
contains the domain to allow CORS.
This project is licensed under the Apache 2.0 license. For more information, see LICENSE.
This is a screenshot of how a package looks like after uploading and analysis:
This screenshot shows different versions in a table: