Caveats
Caveats are a feature within SpiceDB that allows for relationships to be defined conditionally: the relationship will only be considered present if the caveat expression evaluates to true.
Caveats allow for an elegant way to model dynamic policies and ABAC-style (Attribute Based Access Control) decisions while still providing scalability and performance guarantees.
Defining Caveats
Caveats are named expressions that are defined in schema alongside definitions for object types. A caveat definition includes a name, one or more well-typed parameters, and a CEL expression (opens in a new tab) returning a boolean value.
Here's schema snippet demonstrating what a simple caveat looks like:
caveat first_caveat(first_parameter int, second_parameter string) {
first_parameter == 42 && second_parameter == "hello world"
}
Parameter Types
The following table documents the CEL types available for values in caveat expressions:
Type | Description |
---|---|
any | any value is allowed; useful for types that vary |
int | 64-bit signed integer |
uint | 64-bit unsigned integer |
bool | boolean |
string | utf8-encoded string |
double | double-width floating point number |
bytes | sequence of uint8 |
duration | duration of time |
timestamp | specific moment in time (typically UTC) |
list<T> | generic sequence of values |
map<T> | generic mapping of strings to values |
ipaddress | spicedb-specific type for IP addresses |
Developers looking for the SpiceDB code that defines of these types can find them in the pkg/caveats/types module (opens in a new tab).
Some Examples
Basic comparison
caveat is_tuesday(today string) {
today == 'tuesday'
}
Attribute Matching
The example below defines a caveat that requires that any expected attributes found within the expected map are a subset of the attributes in the provided map:
caveat attributes_match(expected map<any>, provided map<any>) {
expected.isSubtreeOf(provided)
}
IP address checking
The example below defines a caveat that requires that a user’s IP address is within a specific CIDR range:
caveat ip_allowlist(user_ip ipaddress, cidr string) {
user_ip.in_cidr(cidr)
}
Allowing caveats on relations
To allow a caveat to be used when writing a relationship, the caveat must be specified on the relation within the schema via the with keyword:
definition resource {
relation viewer: user | user with ip_allowlist
}
In the above example, a relationship can be written for the viewer
relation to a user
without a caveat OR with the ip_allowlist
caveat.
To make the caveat required, the user |
can be removed.
Writing relationships with caveats and context
When writing a relationship for a relation, both the caveat and a portion of the “context” can be specified:
WriteRelationshipsRequest {
Updates: [
RelationshipUpdate{
Operation: CREATE
Relationship: {
Resource: …,
Relation: "viewer",
Subject: …,
OptionalCaveat: {
CaveatName: "ip_allowlist",
Context: structpb{ "cidr": "1.2.3.0/24" }
}
}
}
]
}
A few important notes:
- The Context of a caveat is defined both by the values written in the
Relationship
, as well as those provided in theCheckPermissionRequest
: if empty, then only the context specified on a CheckPermission request will be used. Otherwise, the values in theRelationship
take precedence over those in theCheckPermissionRequest
.- Context of a caveat provided in
Relationship
is stored alongside the relationship and is provided to the caveat expression at runtime. This allows for partial binding of data at write time.
- Context of a caveat provided in
- The Context is a
structpb
, which is defined by Google and represents JSON-like data: https://pkg.go.dev/google.golang.org/protobuf/types/known/structpb (opens in a new tab)- To send 64-bit integers, encode them as strings.
- A relationship cannot be duplicated, with or without a caveat, e.g. two relationships that differ only on their use of a caveat cannot both exist.
- When deleting a relationship, a caveat does not need to be specified; the matching relationship will be deleted if present.
Issuing Checks
When issuing a CheckPermission request (opens in a new tab), additional caveat context can be specified to represent the known context at the time of the check:
CheckPermissionRequest {
Resource: …,
Permission: …,
Subject: …,
Context: { "user_ip": "1.2.3.4" }
}
The check engine will automatically apply the context found on the relationships, as well as the context provided by the CheckPermission call, and return one of three states (opens in a new tab):
PERMISSIONSHIP_NO_PERMISSION
- subject does not have the permission on the resourcePERMISSIONSHIP_HAS_PERMISSION
- subject has permission on the resourcePERMISSIONSHIP_CONDITIONAL_PERMISSION
- required context is missing to determine permissionship
In the case of PERMISSIONSHIP_CONDITIONAL_PERMISSION
, SpiceDB will also return the missing context fields in the CheckPermissionResponse (opens in a new tab) so the caller knows what additional context to fill in if they wish to rerun the check and get a determined answer.
LookupResources and LookupSubjects
Similarly to CheckPermission, both LookupResources and LookupSubjects can be provided with additional context and will return one of the two permission states for each of the results found (either has permission or conditionally has permission).
Full Example
A full example of a schema with caveats can be found below, which allows users to view
a resource if they are directly a viewer
or they are aviewer
within the correct IP CIDR range:
Schema
definition user {}
caveat has_valid_ip(user_ip ipaddress, allowed_range string) {
user_ip.in_cidr(allowed_range)
}
definition resource {
relation viewer: user | user with has_valid_ip
permission view = viewer
}
Write Relationships
WriteRelationshipsRequest {
Updates: [
RelationshipUpdate{
Operation: CREATE
Relationship: {
Resource: {
ObjectType: "resource",
ObjectId: "someresource",
},
Relation: "viewer",
Subject: {
ObjectType: "user",
ObjectId: "sarah",
},
OptionalCaveat: {
CaveatName: "has_valid_ip",
Context: structpb{ "allowed_range": "10.20.30.0/24" }
}
}
}
]
}
Check Permission
CheckPermissionRequest {
Resource: {
ObjectType: "resource",
ObjectId: "someresource",
},
Permission: "view",
Subject: {
ObjectType: "user",
ObjectId: "sarah",
},
Context: { "user_ip": "10.20.30.42" }
}
Validation with Caveats
The Assertions and Expected Relations definitions for validation of schema support caveats as well.
Assertions
Caveated permissions can be checked in assertions by the addition of the assertCaveated
block:
assertTrue:
- "document:specificdocument#reader@user:specificuser"
assertCaveated:
- "document:specificdocument#reader@user:caveateduser"
assertFalse:
- "document:specificdocument#reader@user:anotheruser"
To assert that a permission does or does not exist when some context it specified, the with
keyword can be used to provide the context:
assertTrue:
- "document:specificdocument#reader@user:specificuser"
- 'document:specificdocument#reader@user:caveateduser with {"somecondition": true}'
assertCaveated:
- "document:specificdocument#reader@user:caveateduser"
assertFalse:
- "document:specificdocument#reader@user:anotheruser"
- 'document:specificdocument#reader@user:caveateduser with {"somecondition": false}'
Expected Relations
Expected relations notes if a subject is caveated via the inclusion of the [...]
string on the end of the subject:
document:specificdocument#view:
- "[user:specificuser] is <document:specificdocument#reader>"
- "[user:caveateduser[...]] might be <document:specificdocument#writer>"
Expected Relations does not evaluate caveats, even if the necessary context is fully specified on the relationship.
This means that a caveated subject that might actually return HAS_PERMISSION
will appear as subject[...]
in expected relations