Python Client API

First, install the oso-cloud package on PyPI:

Before going through this guide, make sure you follow the Oso Cloud Quickstart to get your Oso Cloud API Key properly set in your environment.

Instantiating an Oso Cloud client

The Oso Cloud client provides an Oso class that takes your Oso Cloud URL and API key:

Specifying an Oso Fallback host

If you have deployed Oso Fallback nodes to your infrastructure, you may specify the host when instantiating the Oso Cloud client.

Passing application entities into the client

Under the hood, Oso Cloud represents an entity in your application as a combination of a type and an ID, which together uniquely identify the entity. The Python client represents these entities with the Value class, which has both type and id fields. For example:

You will pass objects like these into nearly every function call you make to the Python client.

Primitive value conversion

Additionally, for fields with primitive values you may pass an instance of the corresponding Python type directly instead of a full Value object:

Centralized Authorization Data API

Oso Cloud clients provide an API to manage authorization data stored directly in Oso Cloud.

Add fact: oso.insert((name, *args))

Adds a fact named name with the provided arguments. Example:

oso.insert((
  "has_role",
  Value("User", "bob"),
  "owner",
  Value("Organization", "acme")
))

Delete fact: oso.delete((name, *args))

Deletes a fact. Does not throw an error if the fact is not found. Example:

When deleting facts, you can use None as a wildcard to delete many facts at once.

You can also use oso_cloud.ValueOfType to delete facts with an argument of a particular type.

Transactionally delete and add facts: oso.batch()

Allows deleting and inserting many facts in one atomic transaction. Deletions and insertions are run in the order they appear in the context. Example:

As in calls to delete, you may use wildcards (None) and type-constrained wildcards (ValueOfType) in tx.delete calls.

List facts: oso.get((name, *args))

Lists facts that are stored in Oso Cloud. Can be used to check the existence of a particular fact, or used to fetch all facts that have a particular argument:

You can also use oso_cloud.ValueOfType to get all facts with an argument of a certain type:

Check API

Context facts

The Check API lets you provide context facts with each request. When Oso Cloud performs a check, it considers the request's context facts in addition to any other centralized authorization data. Context facts are only used in the API call in which they're provided-- they do not persist across requests.

For more details, see Context Facts.

Check a permission: oso.authorize(actor, action, resource)

Determines whether or not an action is allowed, based on a combination of authorization data and policy logic. Example:

You may provide a list of context facts as an optional fourth argument to this method. Example:

List authorized resources: oso.list(actor, action, resource_type)

Fetches a list of resources on which an actor can perform a particular action. Example:

You may provide a list of context facts as an optional fourth argument to this method. Example:

List authorized actions: oso.actions(actor, resource)

Fetches a list of actions which an actor can perform on a particular resource. Example:

You may provide a list of context facts as an optional third argument to this method. Example:

Query for any rule: oso.build_query((predicate, *args))

Query Oso Cloud for any predicate and any combination of concrete and wildcard arguments. Unlike oso.get, which only lists facts you've added, you can use oso.build_query to list derived information about any rule in your policy. Example:

Query Builder API

The oso.build_query() API is a builder-style API where you chain methods to construct a query and then execute it.

oso.build_query((predicate, *args))

The oso.build_query() function takes the name of the rule you want to query and a list of arguments. The arguments can be concrete values (e.g., "read" or Value("User", "bob")) or type-constrained variables constructed via the typed_var function:

Note: once you've finished building up your query, you must call evaluate to run it and get the results.

QueryBuilder.and_((predicate, *args))

This function adds another condition that must be true of the query results.

For example:

Note: once you've finished building up your query, you must call evaluate to run it and get the results.

QueryBuilder.in_(variable, values)

This function requires a given typed_var query variable to be included in a given set of values. You can only call in_ once per variable per query. Calling in_ a second time with the same variable on the same query builder will throw an error.

For example:

Note: once you've finished building up your query, you must call evaluate to run it and get the results.

QueryBuilder.with_context_facts(context_facts)

This function adds the given context facts to the query. For example:

For more information on context facts, see this section.

Note: once you've finished building up your query, you must call evaluate to run it and get the results.

QueryBuilder.evaluate()

This function evaluates the built query, fetching the results from Oso.

The return type of this function varies based on the arguments you pass in.

• If you pass no arguments, this function returns a boolean. For example:

  
  

  allowed = (

  oso

  .build_query(("allow", actor, action, resource))

  .evaluate()

  )

  # => true if the given actor can perform the given action on the given resource

  
  

• If you pass a single typed_var query variable, this function returns a list of values for that variable. For example:

  
  

  action = typed_var("String")

  actions = (

  oso

  .build_query(("allow", actor, action, resource))

  .evaluate(action)

  )

  # => all the actions the actor can perform on the given resource- eg. ["read", "write"]

  
  

• If you pass a tuple of typed_var query variables, this function returns a list of tuples of values for those variables. For example:

  
  

  action = typed_var("String")

  repository = typed_var("Repository")

  pairs = (

  oso

  .build_query("allow", actor, action, repository)

  .evaluate((action, repository))

  )

  # => an array of pairs of allowed actions and repo IDs-

  # eg. [("read", "acme"), ("read", "anvil"), ("write", "anvil")]

  
  

• If you pass a dict mapping one typed_var query variable (call it K) to another (call it V), returns a dict grouping unique values of K to unique values of V for each value of K. For example:

  
  

  action = typed_var("String")

  repository = typed_var("Repository")

  repo_actions = (

  oso

  .build_query(("allow", actor, action, repository))

  .evaluate({repository: action})

  )

  # => a dict of repo IDs to allowed actions-

  # eg. { "acme": ["read"], "anvil": ["read", "write"]}

  
  

Some queries have unconstrained results. For instance, maybe users with the admin role can read all Repository entities in your application. In this case, rather than returning an array containing the ID of every repository, evaluate will return an array containing the string "*". For example:

Query Builder examples

Field-level access control

Checking a global permission

Fetching authorized actions for a collection of resources

Filtering out unauthorized resources from a collection

Filtering an authorize() query based on a relation

Learn more about how to query Oso Cloud.

Local Check API

The local check API lets you perform authorization using data that's distributed across Oso Cloud and your own database.

After creating your Local Authorization configuration, provide the path to the YAML file that specifies how to resolve facts in your database.

For more information, see Local Authorization.

List authorized resources with local data: oso.list_local(actor, action, resource_type, column)

Fetches a filter that can be applied to a database query to return just the resources on which an actor can perform an action. Example with SQLAlchemy:

You may use the SQLAlchemy query builder (opens in a new tab) to combine this authorization filter with other things such as ordering and pagination.

You may provide a list of context facts as an optional final argument to this method.

Check a permission with local data: oso.authorize_local(actor, action, resource)

Fetches a query that can be run against your database to determine whether an actor can perform an action on a resource. Example with SQLAlchemy:

The query will return a single boolean value. The user is authorized to perform the action if that boolean value is true.

You may provide a list of context facts as an optional final argument to this method.

List authorized actions with local data: oso.actions_local(actor, resource)

Fetches a query that can be run against your database to fetch the actions an actor can perform on a resource. Example with SQLAlchemy:

You may provide a list of context facts as an optional final argument to this method.

Query for any rule with local data: QueryBuilder

You can use the Query Builder to construct SQL queries that you can run against your database to answer arbitrary questions about authorization.

See the Query Builder API documentation for information on how to construct queries. Once you've built your query, instead of calling evaluate (which would evaluate your query exclusively against data stored in Oso Cloud), call either evaluate_local_select or evaluate_local_filter to construct a SQL query which you can run against your database.

Local Query API

QueryBuilder.evaluate_local_select(column_names_to_query_vars)

Fetches a complete SQL query representing the given Query Builder query. (See the Query Builder API documentation for information on how to construct queries.)

The argument to this function is a dict that maps the column names that will appear in the SELECT clause of the SQL query to the Query Builder variables whose values should be selected. For example, query.evaluate_local_select({"user_id": user_var}) will select all the authorized values of user_var (that is, all the values of user_var that satisfy the Query Builder query) into a column called "user_id".

If you pass in an empty dict or omit this argument entirely, the SQL query will return a single row selecting a boolean column called result. This column will be True when there's some combination of data in your database that can satisfy the given Query Builder query and False otherwise.

Note that each query variable can appear at most once in the column_names_to_query_vars mapping. For example, query.evaluate_local_select({"user_id": user_var, "another_user_id": user_var}) will raise a ValueError, because user_var appears twice in the mapping parameter.

Note that column names will be double-quoted and are thus case-sensitive.

Limitations

evaluate_local_select has the following limitations:

• Queries that would return a wildcard for one of the selected query variables are currently unsupported.

  Example:

  policy.polar

  
  

  has_permission(user: User, "read", _: Repo) if is_global_admin(user);

  
  
  
  

  # alice is a global admin, and can read *any* repo

  oso.insert(("is_global_admin", Value("User", "alice")))

  # UNSUPPORTED: Attempt to query for each authorized user / repo pair

  user_var = typed_var("User")

  repo_var = typed_var("Repo")

  sql = oso.build_query(

  ("allow", user_var, "read", repo_var)

  ).evaluate_local_select({"user_id": user_var, "repo_id": repo_var})

  # => raises `Exception`, because alice can read any repo, so this query would return

  # a wildcard for the repos alice can read

  
  

QueryBuilder.evaluate_local_filter(column_name, query_var)

Fetches a SQL fragment representing the given Query Builder query that you can embed in the WHERE clause of some other SQL query. Use this to filter out unauthorized results from your SQL query.

(See the Query Builder API documentation for information on how to construct queries.)

column_name is the name of the column you want to filter in your SQL query, and query_var is the query variable to filter against. For example, query.evaluate_local_filter("user_id", user_var) will return a SQL fragment of the form "user_id" IN (...), constraining the column "user_id" to the authorized values of user_var.

Note that the column name will be double-quoted and is thus case-sensitive.

Local Query examples

These examples all use SQLAlchemy (opens in a new tab) to execute the returned SQL queries, but you can use any database client you prefer.

Field-level access control

Checking a global permission

Fetching authorized actions for a paginated collection of resources

Filtering an authorize() query based on a relation

Filtering on users who can read a certain resource

Policy API

Update the active policy: oso.policy(policy)

Updates the policy in Oso Cloud. The string passed into this method should be written in Polar. Example:

This command will run any tests defined in your policy. If one or more of these tests fail, your policy will not be updated.

Get policy metadata: oso.get_policy_metadata()

Returns metadata about the currently active policy. Example:

See the Policy Metadata guide for more information on use cases.

Talk to an Oso engineer

If you'd like to learn more about using Oso Cloud in your app or have any questions about this guide, schedule a 1x1 with an Oso engineer. We're happy to help.

Get started with Oso Cloud →