Semantic Validity
The SDK also offers function to validate a Loli specification semantically. What does this mean?
Semantic validation is a step performed after validating the schema. It asserts that the content itself makes sense.
Let's look at a very simple Loli specification. Although the following specification has a valid schema as per the schema definition, but it has one semantic issue. The two feature flags have the exact same name ai-pilot.
That does not make sense, and would cause issues during the evaluation.
{
"schemaVersion": 1,
"featureFlags": [
{
"id": "dfg4d756fg4",
"name": "ai-pilot",
"description": "",
/* ... */
},
{
"id": "y3ck69o1",
"name": "ai-pilot",
"description": "",
/* ... */
}
],
/* ... */
}The SDK offers functions to detect any semantic issues. Especially the Evaluation Client emphasizes schema and semantic validation to only operate on 100% correct Loli specifications.
INFO
The overall goal is to know that a certain specification is valid so that the evaluation logic can safely operate and make certain assumptions based on the validation constraints.
Semantic Issues
The SDK exposes the enum SemanticIssueType that lists all semantic issue types. When the SDK finds semantic issues, they are exposed as instances of SemanticIssue.
Here, all semantic issues that the SDK checks for with their corresponding docs: