Install Python packages from our internal Pipy registry

To use the package from within another project, you need to install it from our internal registry. For this you need to obtain a personal access token in gitea, which you can find under Settings → Applications, or directly via https://code.rekonas.com/user/settings/applications

For adding the registry to your poetry, there are two option. We recommend to use the second one with with devcontainers.

With Poetry (local)

  poetry source add --priority=supplemental rekonas https://code.rekonas.com/api/packages/rekonas/pypi/simple
  poetry config http-basic.rekonas username personal_access_token
  poetry add --source rekonas <your_package>

With Poetry (devcontainer without compose file) [Recommended]

In a devcontainer, we need to make sure to pass the secrets for the private PyPy registry to the container environment.

Create two files containing only your Gitea username and personal-access token under ~/.config/.rekonas-code-user and ~/.config/.rekonas-code-pat

Mount the credentials files in your devcontainer.json and export them as environment variables in the container:

{
"mounts": [
    "source=${localEnv:HOME}/.config/.rekonas-code-user,target=/workspace/.rekonas-code-user,type=bind",
    "source=${localEnv:HOME}/.config/.rekonas-code-pat,target=/workspace/.rekonas-code-pat,type=bind"
],
"postCreateCommand": "export REKONAS_CODE_USER=$(</workspace/.rekonas-code-user) && export REKONAS_CODE_PAT=$(</workspace/.rekonas-code-pat)"
}

Add the repository as in the local option above and set the credentials with poetry config http-basic.rekonas $REKONAS_CODE_USER $REKONAS_CODE_PAT

UV authenticates pipy packages authenticates to it from these env vars:

  export UV_INDEX_REKONAS_USERNAME=<username>
  export UV_INDEX_REKONAS_PASSWORD=<token>

When requesting an internal package (like 'rkns'), the uv runtime discovers it automatically.

uv add 'ml-sleepyland-onnx[cli]' sleepyland-weights-usleep-2024-eeg   # one model
uv add rkns

Or run it once, installing nothing, straight from the index:

uv run --with rkns python