Struct BuiltinEngine 

pub struct BuiltinEngine {
    pub memories: Vec<Memory>,
    pub weights: Weights,
    /* private fields */
}

The default, zero-setup engine: scores an in-memory set with the explainable scorer. (A SQLite-backed variant just loads memories from the DB first.)

With an Embedder attached, similarity becomes semantic (cosine over embeddings); without one, it falls back to lexical term overlap.

Fields

memories: Vec<Memory>

weights: Weights

Implementations

impl BuiltinEngine

pub fn new(memories: Vec<Memory>) -> Self

pub fn with_weights(self, weights: Weights) -> Self

pub fn with_embedder(self, embedder: Arc<dyn Embedder>) -> Result<Self>

Attach an embedder and precompute embeddings for every memory that lacks one. Returns an error if embedding fails (e.g. Ollama not running).

Trait Implementations

impl MemoryEngine for BuiltinEngine

fn name(&self) -> &str

fn retrieve(&self, query: &Query, k: usize) -> Vec<ScoredMemory>

Return the top-k memories for the query, each with its score + reasons.

fn explain(&self, id: i64, query: &Query) -> Option<ScoredMemory>

Full explanation for a single memory id (the memory explain <id> view).