Struct ContextCompiler 

pub struct ContextCompiler { /* private fields */ }

The deterministic context compiler. Build one with a policy, optionally inject an estimator and a pricing table, then compile.

use velesdb_memory::context::{
    CompilePolicy, CompileRequest, ContextCompiler, ContextFragment,
};

let compiler = ContextCompiler::new(CompilePolicy::default());
let out = compiler
    .compile(&CompileRequest {
        query: "deploy status".to_owned(),
        fragments: vec![ContextFragment {
            id: None,
            content: "The deploy pipeline is green.".to_owned(),
            path: None,
            kind: None,
            priority: None,
            metadata: None,
            media: None,
        }],
        project: None,
        target_model: None,
        token_budget: 1_000,
        memory_scope: None,
        policy: None,
    })
    .expect("a generous budget compiles");
assert!(out.content.contains("deploy pipeline"));

Implementations

impl ContextCompiler

pub fn new(policy: CompilePolicy) -> Self

A compiler over policy, with the default char-ratio estimator and no pricing (insights then report tokens only).

pub fn with_estimator(self, estimator: DynTokenEstimator) -> Self

Replace the token estimator (e.g. a model-exact tokenizer).

pub fn with_pricing(self, pricing: PricingTable) -> Self

Inject a versioned pricing table so insights also report estimated cost savings for the request’s target model.

pub fn compile( &self, request: &CompileRequest, ) -> Result<CompiledContext, MemoryError>

Compile request into a budgeted, fully-audited context.

Errors

MemoryError::ContextOverLimit when the request exceeds a crate::limits cap (fragment count or single-fragment size), MemoryError::MetadataTooLarge when a fragment’s metadata exceeds crate::limits::MAX_METADATA_BYTES, and MemoryError::ContextBudget when the token budget minus the policy’s response reserve leaves no room for any context.