Struct SchemaLearnerConfig 

pub struct SchemaLearnerConfig {
    pub max_schemas: usize,
    pub min_samples_for_validation: u32,
    pub max_nesting_depth: usize,
    pub max_fields_per_schema: usize,
    pub string_length_tolerance: f64,
    pub number_value_tolerance: f64,
    pub required_field_threshold: f64,
}

Configuration for the schema learner.

Security Considerations

The tolerance values (string_length_tolerance and number_value_tolerance) directly impact the security posture of schema validation. These multipliers determine how much deviation from learned baselines is permitted before a request is flagged as anomalous.

Tolerance Trade-offs

• Lower values (1.0-1.5): Stricter validation, higher security, but may cause false positives if legitimate traffic has natural variance.
• Higher values (2.0+): More permissive, fewer false positives, but allows attackers more room to inject oversized payloads or extreme values.

Recommendations

• Start with default tolerance (1.5) and monitor for false positives
• For high-security APIs: consider 1.2-1.3
• For APIs with high variance: consider 1.5-2.0
• Never set below 1.0 (would reject valid baseline data)

Example

use synapse_pingora::profiler::SchemaLearnerConfig;

// Stricter configuration for sensitive APIs
let config = SchemaLearnerConfig {
    string_length_tolerance: 1.3,  // 30% buffer above learned max
    number_value_tolerance: 1.25,  // 25% buffer above learned max
    ..Default::default()
};

// Validate config before use
config.validate().expect("Invalid configuration");

Fields

max_schemas: usize

Maximum number of endpoint schemas to track.

When this limit is reached, the least recently used (LRU) schema is evicted.

min_samples_for_validation: u32

Minimum samples required before validation is active.

Until an endpoint has been observed this many times, validation will not flag anomalies. This prevents false positives during the initial learning phase.

max_nesting_depth: usize

Maximum depth for nested object learning.

Prevents excessive memory usage from deeply nested JSON structures.

max_fields_per_schema: usize

Maximum fields per schema (memory protection).

Limits the number of fields tracked per endpoint to prevent memory exhaustion from APIs with dynamic or unbounded field sets.

string_length_tolerance: f64

String length tolerance multiplier for validation.

When validating string fields, the maximum allowed length is: learned_max_length * string_length_tolerance

Security Impact

• Lower values (1.0-1.3): Catches buffer overflow attempts more aggressively but may flag legitimate variance as anomalous.
• Higher values (1.5-2.0): More permissive, reducing false positives but allowing larger payloads that could exploit vulnerabilities.

Default: 1.5 (50% buffer above learned maximum)

Constraints

Must be >= 1.0. Values below 1.0 would reject strings that were seen in the baseline training data, causing immediate false positives.

number_value_tolerance: f64

Number value tolerance multiplier for validation.

When validating numeric fields:

• Maximum allowed: learned_max * number_value_tolerance
• Minimum allowed: learned_min / number_value_tolerance

Security Impact

• Lower values (1.0-1.3): Catches integer overflow attempts and extreme value injection more aggressively.
• Higher values (1.5-2.0): More permissive for APIs with high numeric variance.

Default: 1.5 (50% buffer on max values, 33% reduction on min values)

Constraints

Must be >= 1.0. Values below 1.0 would reject values that were seen in the baseline training data, causing immediate false positives.

required_field_threshold: f64

Required field threshold (fields seen in > threshold% of requests).

Fields that appear in more than this percentage of observed requests are considered “required” and their absence will trigger a MissingField violation.

Default: 0.9 (90% - fields must appear in 90% of samples to be required)

Implementations

impl SchemaLearnerConfig

pub fn validate(&self) -> Result<(), ConfigValidationError>

Validates the configuration, ensuring all values are within acceptable ranges.

Errors

Returns ConfigValidationError if:

• string_length_tolerance < 1.0
• number_value_tolerance < 1.0
• required_field_threshold is not in range [0.0, 1.0]

Example

use synapse_pingora::profiler::SchemaLearnerConfig;

let config = SchemaLearnerConfig {
    string_length_tolerance: 0.5, // Invalid!
    ..Default::default()
};

assert!(config.validate().is_err());

Trait Implementations

impl Clone for SchemaLearnerConfig

fn clone(&self) -> SchemaLearnerConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SchemaLearnerConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for SchemaLearnerConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SchemaLearnerConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SchemaLearnerConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.