Struct ValidationPipeline 

pub struct ValidationPipeline;

The validation pipeline: parses, validates, and deserializes LLM output.

Supports two modes:

• Static (typed): validate_and_parse::<T>() for compile-time Rust types
• Dynamic: validate_dynamic() for runtime-defined schemas from the DSL

The dynamic path validates serde_json::Value against a JSON Schema without requiring a Rust type, which is essential for user-defined output shapes composed at runtime via the DSL.

Implementations

impl ValidationPipeline

pub fn validate_and_parse<T: DeserializeOwned>( raw_text: &str, schema: Option<&Validator>, ) -> Result<T, SchemaValidationError>

Run the full validation pipeline with static typing: strip fences → parse JSON → validate → deserialize into T.

Use this when you have a compile-time Rust type for the output.

pub fn validate_dynamic( raw_text: &str, schema: Option<&Validator>, ) -> Result<Value, SchemaValidationError>

Run the validation pipeline for dynamic schemas: strip fences → parse JSON → validate against schema → return Value.

Use this when output shapes are defined at runtime (e.g., from the DSL). The returned serde_json::Value is guaranteed to conform to the schema.

pub fn parse_json(raw_text: &str) -> Result<Value, SchemaValidationError>

Parse raw text as JSON without schema validation.

pub fn validate_schema( value: &Value, validator: &Validator, ) -> Result<(), SchemaValidationError>

Validate a JSON value against a pre-compiled schema.

pub fn compile_schema( schema: &Value, ) -> Result<Validator, SchemaValidationError>

Create a validator from a raw JSON Schema value.

This is the primary way to create validators for dynamic schemas defined at runtime (e.g., from DSL configurations or TOML pipelines).