Struct SchemaCleanr 

pub struct SchemaCleanr;

JSON Schema cleaner optimized for LLM tool calling.

Implementations

impl SchemaCleanr

pub fn clean_for_gemini(schema: Value) -> Value

Clean schema for Gemini compatibility (strictest).

This is the most aggressive cleaning strategy, removing all keywords that Gemini’s API rejects.

pub fn clean_for_anthropic(schema: Value) -> Value

Clean schema for Anthropic compatibility.

pub fn clean_for_openai(schema: Value) -> Value

Clean schema for OpenAI compatibility (most permissive).

pub fn clean_shared( schema: &Arc<Value>, strategy: CleaningStrategy, ) -> Arc<Value>

Zero-copy wrapper around Self::clean for Arc-shared tool schemas: returns the same Arc when the pre-scan proves cleaning is a no-op, deep-copying the tree only when a rewrite is actually needed.

pub fn needs_cleaning(schema: &Value, strategy: CleaningStrategy) -> bool

Conservative read-only pre-scan: true when Self::clean with strategy could change schema.

False positives are allowed (a flagged schema may clean to an equal value); false negatives are not — !needs_cleaning(s) must imply clean(s) == s. The triggers mirror every rewrite path in clean_object: strategy-specific keyword removal, plus the strategy-independent rewrites ($ref resolution, const → enum, anyOf/oneOf simplification and sibling-type skipping, and null-stripping in type arrays).

pub fn clean(schema: Value, strategy: CleaningStrategy) -> Value

Clean schema with specified strategy.

pub fn validate(schema: &Value) -> Result<()>

Validate that a schema is suitable for LLM tool calling.

Returns an error if the schema is invalid or missing required fields.