Skip to content

najose/quetz

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

quetz header image

The Open-Source Server for Conda Packages

part of mamba-org
Package Manager mamba Package Server quetz Package Builder boa

Quetz

Documentation Status Docker Version

The quetz project is an open source server for conda packages. It is built upon FastAPI with an API-first approach. A quetz server can have many users, channels and packages. With quetz, fine-grained permissions on channel and package-name level will be possible.

Quetz has an optional client quetz-client that can be used to upload packages to a quetz server instance.

Usage

You should have mamba or conda installed.

Get Quetz sources:

git clone https://github.com/mamba-org/quetz.git

Then create an environment:

cd quetz
mamba env create -f environment.yml
conda activate quetz
ln -s "${CONDA_PREFIX}" .venv  # Necessary for pyright.

Install Quetz:

Use the editable mode -e if you are developer and want to take advantage of the reload option of Quetz

pip install -e .

Use the CLI to create a Quetz instance:

quetz run test_quetz --copy-conf ./dev_config.toml --dev --reload

Links:

Download xtensor as test package:

./download-test-package.sh

To upload the package, install the quetz-client:

mamba install quetz-client

To run the upload, you need to set environment variables for the quetz API key (which authenticates you) and the quetz server URL. As we passed the --dev flag to quetz, a testing API key can be found in quetz's output which you can use for this example.

export QUETZ_API_KEY=E_KaBFstCKI9hTdPM7DQq56GglRHf2HW7tQtq6si370
export QUETZ_SERVER_URL=http://localhost:8000
quetz-client post_file_to_channel channel0 xtensor/linux-64/xtensor-0.24.3-h924138e_1.tar.bz2
quetz-client post_file_to_channel channel0 xtensor/osx-64/xtensor-0.24.3-h1b54a9f_1.tar.bz2
quetz-client post_file_to_channel channel0 xtensor/osx-arm64/xtensor-0.24.3-hf86a087_1.tar.bz2

Install the test package with conda:

mamba install --strict-channel-priority -c http://localhost:8000/get/channel0 -c conda-forge xtensor

Output:

...
  Package  Version  Build          Channel                                                     Size
─────────────────────────────────────────────────────────────────────────────────────────────────────
  Install:
─────────────────────────────────────────────────────────────────────────────────────────────────────

  xtensor   0.16.1  0              http://localhost:8000/get/channel0/osx-64                 109 KB
  xtl       0.4.16  h04f5b5a_1000  conda-forge/osx-64                                         47 KB

  Summary:

  Install: 2 packages

  Total download: 156 KB

─────────────────────────────────────────────────────────────────────────────────────────────────────
...

Browse channels: http://localhost:8000/get/channel0/

S3 Backend

To enable the S3 backend, you will first require the s3fs library:

mamba install -c conda-forge s3fs

Then add your access and secret keys to the s3 section with your config.toml, like so:

[s3]
access_key = "AKIAIOSFODNN7EXAMPLE"
secret_key = "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
url = "https://..."
region = ""
bucket_prefix="..."
bucket_suffix="..."

Be sure to set the url and region field if not using AWS.

Channels are created with the following semantics:

{bucket_prefix}{channel_name}{bucket_suffix}

The s3 backend is currently designed for one bucket per channel. It may be possible to put all channels in one bucket, but that will require some code tweaks

If you're using IAM roles, don't set access_key and secret_key or set them to empty strings "".

Google OAuth 2.0 OpenID Connect

To enable auth via Google, you will need to register an application at: https://console.developers.google.com/apis/credentials

Then add the client secret & ID to a google section of your config.toml:

[google]
client_id = "EXAMPLEID420127570681-6rbcgdj683l3odc3nqearn2dr3pnaisq.apps.googleusercontent.com"
client_secret = "EXAMPLESECRETmD-7UXVCMZV3C7b-kZ9yf70"

PostgreSQL

By default, quetz will run with sqlite database, which works well for local tests and small local instances. However, if you plan to run quetz in production, we recommend to configure it with the PostgreSQL database. There are several options to install PostgreSQL server on your local machine or production server, one of them being the official PostgreSQL docker image.

Running PostgreSQL server with docker

You can the PostgresSQL image from the docker hub and start the server with the commands:

docker pull postgres
docker run --name some-postgres -p 5432:5432 -e POSTGRES_PASSWORD=mysecretpassword -d postgres

This will start the server with the user postgres and the password mysecretpassword that will be listening for connection on the port 5432 of localhost.

You can then create a database in PostgreSQL for quetz tables:

sudo -u postgres psql -h localhost -c 'CREATE DATABASE quetz OWNER postgres;'

Deploying Quetz with PostgreSQL backend

Then in your configuration file (such as dev_config.toml) replace the [sqlalchemy] section with:

[sqlalchemy]
database_url = "postgresql://postgres:mysecretpassword@localhost:5432/quetz"

Finally, you can create and run a new quetz deployment based on this configuration (we assume that you saved it in file config_postgres.toml):

quetz run postgres_quetz --copy-conf config_postgres.toml

Note that this recipe will create an ephemeral PostgreSQL database and it will delete all data after the some-postgres container is stopped and removed. To make the data persistent, please check the documentation of the postgres image or your container orchestration system (Kubernetes or similar).

Running tests with PostgreSQL backend

To run the tests with the PostgreSQL database instead of the default SQLite, follow the steps above to start the PG server. Then create an new database:

psql -U postgres -h localhost -c 'CREATE DATABASE test_quetz OWNER postgres;'

You will be asked to type the password to the DB, which you defined when starting your PG server. In the docker-based instructions above, we set it to mysecretpassword.

To run the tests with this database you need to configure the QUETZ_TEST_DATABASE environment variable:

QUETZ_TEST_DATABASE="postgresql://postgres:mysecretpassword@localhost:5432/test_quetz" pytest -v ./quetz/tests

Frontend

Quetz comes with a initial frontend implementation. It can be found in quetz_frontend. To build it, one needs to install:

mamba install 'nodejs>=14'
cd quetz_frontend
npm install
npm run build
# for development
npm run watch

This will build the javascript files and place them in /quetz_frontend/dist/ from where they are automatically picked up by the quetz server.

Using quetz

Create a channel

First, make sure you're logged in to the web app.

Then, using the swagger docs at <deployment url>:<port>/docs, POST to /api/channels with the name and description of your new channel:

{
  "name": "my-channel",
  "description": "Description for my-channel",
  "private": false
}

This will create a new channel called my-channel and your user will be the Owner of that channel.

Generate an API key

API keys are scoped per channel, per user and optionally per package. In order to generate an API key the following must be true:

  1. First, make sure you're logged in to the web app.

  2. The user must be part of the target channel (you might need to create a channel first, see the previous section on how to create a channel via the swagger docs)

  3. Go to the swagger docs at <deployment url>:<port>/docs and POST to /api/api-keys:

    {
      "description": "my-test-token",
      "roles": [
        {
          "role": "owner",
          "channel": "my-channel"
        }
      ]
    }
  4. Then, GET on /api/api-keys to retrieve your token

  5. Finally, set the QUETZ_API_KEY environment variable to this value so you can use quetz-client to interact with the server.

Create a proxy channel

A proxy channel "mirrors" another channel usually from a different server, so that the packages can be installed from the proxy as if they were installed directly from that server. All downloaded packages are cached locally and the cache is always up to date (there is no risk of serving stale packages). The reason to use the proxy channel is to limit traffic to the server of origin or to serve a channel that could be inaccessible from behind the corporate firewall.

To create a proxy channel use the properties mirror_channel_url=URL_TO_SOURCE_CHANNEL and mirror_mode='proxy' in the POST method of /api/channels endpoint. For example, to proxy the channel named btel from anaconda cloud server, you might use the following request data:

{
  "name": "proxy-channel",
  "private": false,
  "mirror_channel_url": "https://conda.anaconda.org/btel",
  "mirror_mode": "proxy"
}

You may copy the data directly to the Swagger web interface under the heading POST /api/channels or use the cURL tool from command line. Assuming that you deployed a quetz server on port 8000 (the default) on your local machine, you could make the request with the following cURL command:

export QUETZ_API_KEY=...
curl -X POST "http://localhost:8000/api/channels" \
    -H  "accept: application/json" \
    -H  "Content-Type: application/json" \
    -H  "X-API-Key: ${QUETZ_API_KEY}" \
    -d '{"name":"proxy-channel",
         "private":false,
         "mirror_channel_url":"https://conda.anaconda.org/btel",
         "mirror_mode":"proxy"}'

where the value of QUETZ_API_KEY variable should be the API key that was printed when you created the quetz deployment or retrieved using the API as described in the section Generate an API key.

Then you can install packages from the channel the standard way using conda or mamba:

mamba install --strict-channel-priority -c http://localhost:8000/get/proxy-channel nrnpython

Create a mirroring channel

A mirror channel is an exact copy of another channel usually from a different (anaconda or quetz) server. The packages are downloaded from that server and added to the mirror channel. The mirror channel supports the standard Quetz API except requests that would add or modify the packages (POST /api/channels/{name}/files, for example). Mirror channels can be used to off load traffic from the primary server, or to create a channel clone on the company Intranet.

Creating a mirror channel is similar to creating proxy channels except that you need to change the value of mirror_mode attribute from proxy to mirror (and choose a more suitable channel name obviously):

{
  "name": "mirror-channel",
  "private": false,
  "mirror_channel_url": "https://conda.anaconda.org/btel",
  "mirror_mode": "mirror"
}

or with cURL:

export QUETZ_API_KEY=...
curl -X POST "http://localhost:8000/api/channels" \
    -H  "accept: application/json" \
    -H  "Content-Type: application/json" \
    -H  "X-API-Key: ${QUETZ_API_KEY}" \
    -d '{"name":"mirror-channel",
         "private":false,
         "mirror_channel_url":"https://conda.anaconda.org/btel",
         "mirror_mode":"mirror"}'

Mirror channels are read only (you can not add or change packages in these channels), but otherwise they are fully functional Quetz channels and support all standard read (GET) operations. For example, you may list all packages using GET /api/channels/{channel_name}/packages endpoint:

curl http://localhost:8000/api/channels/mirror-channel/packages

If packages are added or modified on the primary server from which they were pulled initially, they won't be updated automatically in the mirror channel. However, you can trigger such synchronisation manually using the PUT /api/channels/{channel_name}/actions endpoint:

curl -X PUT localhost:8000/api/channels/mirror-channel/actions \
   -H "X-API-Key: ${QUETZ_API_KEY}" \
   -d '{"action": "synchronize"}'

Only channel owners or maintainers are allowed to trigger synchronisation, therefore you have to provide a valid API key of a privileged user.

Plugins

Quetz offers plugins in the plugins folder of this repo as well as via standalone installs. The following plugins are currently available:

Plugin Description
quetz_conda_suggest Generate .map files for conda-suggest
quetz_content_trust Generate signed repodata files
quetz_current_repodata Trim the repodata to only include latest package versions
quetz_dictauthenticator Demo for creating new authenticators
quetz_harvester Extract additional metadata from packages using the libcflib harvester
quetz_mamba_solve Export a specific set of package versions for reproducibility
quetz_repodata_patching Repodata patching
quetz_repodata_zchunk Serve repodata using zchunk
quetz_runexports Extract and expose run_exports from package files
quetz_sql_authenticator An authenticator that stores credentials in the Quetz SQL database using passlib.
quetz_tos Enforce signing the terms of service for Quetz users
quetz_transmutation Convert packages to .conda format

License

We use a shared copyright model that enables all contributors to maintain the copyright on their contributions.

This software is licensed under the BSD-3-Clause license. See the LICENSE file for details.

About

The Open-Source Server for Conda Packages

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 96.7%
  • C 1.8%
  • Jinja 0.5%
  • HTML 0.5%
  • Dockerfile 0.2%
  • Shell 0.1%
  • Other 0.2%