Module embedding_onnx 

ONNX, sentence-transformer, and AST-aware embedding backends (K3c-G2).

These are alternative EmbeddingProvider implementations that complement the existing [LlmEmbeddingProvider] and MockEmbeddingProvider.

• OnnxEmbeddingProvider – local model inference (all-MiniLM-L6-v2, 384-d).
• SentenceTransformerProvider – documentation-optimised paragraph embedder.
• AstEmbeddingProvider – hybrid structural + semantic embedder for Rust code.

Structs

AstEmbeddingProvider

AST-aware embedding provider for Rust source code.

OnnxEmbeddingProvider

ONNX Runtime embedding provider.

RustCodeFeatures

Structural features extracted from Rust source code via regex parsing.

SentenceTransformerProvider

Sentence-transformer embedding provider for documentation.

WordPieceTokenizer

BERT-compatible WordPiece tokenizer.

Functions

extract_rust_features

Extract structural features from a Rust code snippet using simple regex-like parsing. Does not depend on tree-sitter so the kernel stays lightweight.

preprocess_markdown

Pre-process markdown text by stripping structural elements.

split_sentences

Simple sentence splitting on “. “ and newlines.