anaconda-auth

A client library for Anaconda APIs to authenticate and securely store API keys. This library is used by other client packages to provide a centralized auth capability for the ecosystem. You will need to use this package to log in to anaconda.com before utilizing many of the other client packages.

This package provides a requests client class that handles loading the API key for requests made to Anaconda services.

This package provides a Panel OAuth plugin called anaconda_auth.

Installation

conda install anaconda-auth

Usage

The primary usage of this package is to provide CLI actions for login, logout, user information, and api-keys to Anaconda API Services.

❯ anaconda auth

 Usage: anaconda auth [OPTIONS] COMMAND [ARGS]...

 Anaconda auth commands

╭─ Options ──────────────────────────────────────────────────────────────────────────────────╮
│ --version  -V                                                                              │
│ --help               Show this message and exit.                                           │
╰────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ─────────────────────────────────────────────────────────────────────────────────╮
│ api-key   Display API Key for signed-in user                                               │
│ login     Login                                                                            │
│ logout    Logout                                                                           │
│ whoami    Display information about the currently signed-in user                           │
╰────────────────────────────────────────────────────────────────────────────────────────────╯

Change site

To change the default site for all anaconda CLI plugins you can run

anaconda sites add --name <site-name> --domain <domain> --default [extra-flags]

For the list of extra flags see the next section. This will edit your ~/.anaconda/config.toml.

Configuration

You can configure anaconda-auth by either:

1. Setting parameters in the plugin.auth section of the ~/.anaconda/config.toml file.
2. Setting one or more sites in the [sites.<name>] section of the ~/.anaconda/config.toml file.
3. Setting one or more ANACONDA_AUTH_ environment variables or using a .env file in your working directory.

ANACONDA_AUTH_ env vars or a .env file take precedence over the ~/.anaconda/config.toml file.

Anaconda parameters

The following parameters in the plugin.auth section control the login actions and API requests to Anaconda Services.

| Parameter | Env variable | Description | Default value | | --------------------------- | ----------------------------------------- | ----------------------------------------------------------------------------------------------------------- | -------------------- | | domain | ANACONDA_AUTH_DOMAIN | Authentication and API request domain | "anaconda.com" | | ssl_verify | ANACONDA_AUTH_SSL_VERIFY | SSL verification for all requests | True | | preferred_token_storage | ANACONDA_AUTH_PREFERRED_TOKEN_STORAGE | Where to store the login token; can be "anaconda-keyring" or "system" | "anaconda-keyring" | | api_key | ANACONDA_AUTH_API_KEY | API key; if None, defaults to keyring storage | None | | extra_headers | ANACONDA_AUTH_EXTRA_HEADERS | Extra request headers in JSON format | None | | use_unified_repo_api_key | ANACONDA_AUTH_USE_UNIFIED_REPO_API_KEY | Configures anaconda-auth to use unified api key for conda repository access instead of default repo token | False | | use_device_flow | ANACONDA_AUTH_USE_DEVICE_FLOW | Configures anaconda-auth to use device flow grant | False | | disable_conda_auto_config | ANACONDA_AUTH_DISABLE_CONDA_AUTO_CONFIG | Disable automatic configuration of conda's channel_settings | False |

Example

Here's an example ~/.anaconda/config.toml where SSL verification is turned off for login and API requests and where the preferred token storage is anaconda-keyring.

[plugin.auth]
ssl_verify = false
preferred_token_storage = "system"

Multi-site configuration

This plugin supports configuration of multiple sites for use with anaconda login and BaseClient. Each of the above parameters can be set within a [sites.<name>] table within ~/.anaconda/config.toml. For example

[sites.on-prem]
domain = "my.site.foo.com"
ssl_verify = false

When utilizing a site with BaseClient(site="<name>") the global [plugin.auth] table acts as a fallback, the site config will utilize any value in [plugin.auth] that is not explicitly overridden in the [sites] table. Finally, all value of any configured site can be overridden with the ANACONDA_AUTH_* env vars above.

You can have any number of this tables within the config file. To set a site as default use default_site = "<name>" where the <name> is the name of one of the sites tables. For example, you can have an extra site in addition to the default anaconda.com site make it the default for BaseClient() and CLI operations.

default_site = "on-prem"

[sites."anaconda.com"]
# Empty table means use default values

[sites.on-prem]
domain = "my.site.foo.com"
ssl_verify = false

CLI commands

A top-level --at <site-name> flag is added to all anaconda CLI commands to switch the site.

anaconda --at <site-name> <cmd> [flags]

This works across all Anaconda CLI plugins and subcommands.

Multi-site configuration can be controlled using the anaconda sites subcommand to list, add, modify, and remove sites.

❯ anaconda sites

 Usage: anaconda sites [OPTIONS] COMMAND [ARGS]...

 Manage your Anaconda site configuration

╭─ Options ─────────────────────────────────────────────────────────────────────────────────────╮
│ --help  -h        Show this message and exit.                                                 │
╰───────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ────────────────────────────────────────────────────────────────────────────────────╮
│ list     List configured sites by name and domain.                                            │
│ show     Show the site configuration for the default site or look up by the provided name or  │
│          domain.                                                                              │
│ add      Add new site configuration to ~/.anaconda/config.toml                                │
│ modify   Modify site configuration in ~/.anaconda/config.toml                                 │
│ remove   Remove site configuration by name or domain.                                         │
╰───────────────────────────────────────────────────────────────────────────────────────────────╯

API Keys and tokens

When you login with anaconda-auth, an auth token is stored in the preferred keyring storage location and is deleted when you run logout. The auth token will need to be renewed once a year.

The preferred_storage configuration parameter in the plugin.auth section of the config.toml file takes two possible values:

| Storage location | Description | | -------------------- | ----------------------------------------------------------------------- | | "system" | Use the system keyring, if available. Otherwise, use anaconda-keyring | | "anaconda-keyring" | A file-based keyring at ~/.anaconda/keyring |

"anaconda-keyring" is the default value.

Non-interactive use

If you want to utilize Anaconda Services on a system where you do not have interactive access to a browser to use the login command, you have two options:

1. On a system where you can log in, use anaconda auth api-key command to print the API key to the terminal. You can then utilize the API key on the non-interactive system with the ANACONDA_AUTH_API_KEY env var (or in a .env file) or set the key parameter in the plugin.auth section of the ~/.anaconda/config.toml file.
2. With preferred_token_storage set to "anaconda-keyring", run the login command to create the ~/.anaconda/keyring file. Then, copy ~/.anaconda/keyring to the non-interactive system.

Python API

from anaconda_auth import login

login()

The login() function initiates a browser-based login flow for your default site. It will automatically open your browser and, once you have completed the login flow, it will store an API key on your system.

login() takes an optional keyword argument site="<name or domain>" to login to another configured site in your ~/.anaconda/config.toml.

Typically, these API keys will have a one year expiration, so you will only need to log in once and requests using the client class will read the token from the keyring storage.

If you call login() while there is a valid (non-expired) API key, no action is taken. You can replace the valid API key with login(force=True).

To remove the API key from your keyring storage, use the logout() function.

from anaconda_auth import logout

logout()

logout() takes an optional keyword argument site="<name or domain>" to logout from another configured site in your ~/.anaconda/config.toml. Without the site= argument the default site from your configuration is used.

API requests

The BaseClient class is a subclass of requests.Session. It will attempt to load the API key from the keyring on each request unless overridden by the api_key argument.

The BaseClient class can be used for non-authenticated requests even when the user has not logged in or provided an API in the request.

from anaconda_auth.client import BaseClient

client = BaseClient()

response = client.get("/api/<endpoint>")
response.raise_for_status()
print(response.json())

BaseClient accepts the following optional arguments.

• site: Site name from ~/.anaconda/config.toml
• domain: Domain to use for requests, defaults to your default site
• ssl_verify: Enable SSL verification, defaults to True
• api_key: API key to use for requests, if unspecified, uses token set by anaconda login
• user_agent: Defaults to anaconda-auth/<package-version>
• api_version: Requested API version, defaults to latest available from the domain
• extra_headers: Dictionary or JSON string of extra headers to send in requests

When using the site="<name>" argument all other keyword args will override the configuration loaded from ~/.anaconda/config.toml.

To create a Client class specific to your package, subclass BaseClient and set an appropriate user-agent and API version for your needs.

from anaconda_auth.client import BaseClient
class Client(BaseClient):
    _user_agent = "anaconda-<package>/<version>"
    _api_version = "<api-version>"

CLI Error handlers

This plugin defines an error handler for the HTTPError exception when using .raise_for_status() on a response using BaseClient or subclasses of BaseClient. Errors are not caught automatically when using the BaseClient or subclasses outside of anaconda CLI subcommands.

Login required

For the following cases, the user is running the CLI command interactively and is asked if they wish to continue with interactive login. Once completed, the command will be re-tried.

• TokenNotFoundError: The subcommand requested to load the token from the keyring, but none were present.
• TokenExpiredError: The token was successfully loaded but has expired.
• AuthenticationMissing: Derived from requests.exceptions.HTTPError, the request was made without an API key or token to an endpoint that requires authentication.
• InvalidAuthentication: Derived from requests.exceptions.HTTPError, the request was made using an API key or token, but Anaconda determined that the API was invalid.

Here's an example demonstrating that the user has not previously run anaconda login but attempted a CLI command that at some point requires authentication. By typing y, the login action is triggered and their browser will open.

❯ anaconda auth api-key
TokenNotFoundError: Login is required to complete this action.
Continue with interactive login? [y/n]: y
<api-key>

If the user typed n or the command was not run on an interactive terminal, an error message is shown instructing the user how to log in or configure the API key.

❯ anaconda auth whoami
AuthenticationMissingError: Login is required to complete this action.
Continue with interactive login? [y/n]: n

To configure your credentials you can run
  anaconda login --at anaconda.com

or set your API key using the ANACONDA_AUTH_API_KEY env var

or set

[plugin.auth]
api_key = "<api-key>"

in ~/.anaconda/config.toml

To see a more detailed error message run the command again as
  anaconda --verbose auth whoami

HTTPError

In addition to the two special cases above, all HTTPError exceptions thrown during CLI subcommands will be handled to provide the error code and reason.

For example, a subcommand using BaseClient or a subclass of it may make a bad request.

@plugin.subcommand('do-something')
def do_something(inputs: Annotated[Any, typer.Argument()]):
    client = Client()
    res = client.post('api/something', data=inputs)
    res.raise_for_status()

For this example subcommand, the user may provide incorrect inputs that are passed to the endpoint. By using .raise_for_status(), the error is passed along to the CLI user and a short response is printed.

> anaconda plugin do-something 'input-data'
HTTPError: 422 Client Error: Unprocessable Entity for url: https://anaconda.com/api/something

To see a more detailed error message run the command again as
  anaconda --verbose plugin do-something

Panel OAuth Provider

In order to use the anaconda_auth auth plugin, you will need an OAuth client ID (key) and secret. The client must be configured as follows:

Set scopes: offline_access, openid, email, profile
Set redirect url to http://localhost:5006
Set grant type: Authorization Code
Set response types: ID Token, Token, Code
Set access token type: JWT
Set Authentication Method: HTTP Body

To run the app with the anaconda_auth auth provider, you will need to set several environment variables or command-line arguments. See the Panel OAuth documentation for more details.

PANEL_OAUTH_PROVIDER=anaconda_auth or --oauth-provider anaconda_auth
PANEL_OAUTH_KEY=<key>               or --oauth-key=<key>
PANEL_OAUTH_SECRET=<secret>         or --oauth-secret=<key>
PANEL_COOKIE_SECRET=<cookie-name>   or --cookie-secret=<value>
PANEL_OAUTH_REFRESH_TOKENS=1        or --oauth-refresh-tokens
PANEL_OAUTH_OPTIONAL=1              or --oauth-optional

panel serve <arguments> ...

If you do not specify the .env file, the production configuration should be the default. Please file an issue if you see any errors.

Setup for development

Ensure you have conda installed. Then run:

make setup

Run the unit tests

make test

Run the unit tests across isolated environments with tox

make tox