Struct BertEncoder 

pub struct BertEncoder { /* private fields */ }

BERT-family encoder implementation of VectorEncoder.

Constructed by main.rs::load_pipeline with detected backends and the model’s HF tokenizer. The model_repo and hidden_dim are recorded at construction time for cache keying and diagnostics; both are known from the model selection in the CLI.

Implementations

impl BertEncoder

pub fn new( backends: Vec<Box<dyn EmbedBackend>>, tokenizer: Tokenizer, model_repo: String, hidden_dim: usize, ) -> Self

Build a BertEncoder from already-loaded backends and tokenizer.

model_repo is the HuggingFace repo string for cache keying and identity(). hidden_dim is the embedding output dimension (384 for BGE-small-en-v1.5; 768 for ModernBERT-embed-base).

pub fn backends(&self) -> &[Box<dyn EmbedBackend>]

Borrow the underlying backends.

Useful for query-time embedding (e.g., interactive mode re-uses the loaded backend to embed a query string).

pub fn tokenizer(&self) -> &Tokenizer

Borrow the underlying tokenizer.

Same query-time use case as backends.

Trait Implementations

impl VectorEncoder for BertEncoder

fn embed_root( &self, root: &Path, cfg: &SearchConfig, profiler: &Profiler, ) -> Result<(Vec<CodeChunk>, Vec<Vec<f32>>)>

Walk root, chunk every supported file, and embed every chunk.

fn hidden_dim(&self) -> usize

Hidden dimension of the emitted embeddings.

fn identity(&self) -> &str

Stable identifier used as the cache-manifest key.