Struct MockEmbeddingModel 

pub struct MockEmbeddingModel;

Deterministic 8-dimensional mock embedding model.

Uses the SHA-256 of the input text as a pseudo-random source for 8 f32 components, then L2-normalizes the result. Never use for semantic search — the outputs are semantically meaningless. Suitable for pipeline correctness tests (RFC-008 §24 tests 1–10).

Trait Implementations

impl EmbeddingModel for MockEmbeddingModel

fn name(&self) -> &str

Stable name stored in models.model_name.

fn version(&self) -> &str

Version string stored in models.model_version.

fn dimension(&self) -> u32

Output dimension — must match stored embeddings (RFC-008 §11).

fn embed_batch(&self, texts: &[&str]) -> OrbokResult<Vec<Vec<f32>>>

Embed a batch of normalized texts. Returns one vector per input, each L2-normalized.

fn embed_batch_with_stats( &self, texts: &[&str], ) -> OrbokResult<(Vec<Vec<f32>>, Option<EmbeddingBatchStats>)>

embed_batch, plus this call’s batch statistics (RFC-048 Task 011) — None when a backend doesn’t track them. The default implementation delegates to embed_batch and reports no stats, so every existing backend and call site is unaffected; a backend that can report real figures overrides this instead, as a byproduct of work it already performs, adding no cost to plain embed_batch.