Module embed 

Search configuration, results, and file I/O helpers.

The transformer streaming pipeline (embed_all, embed_all_batch, embed_all_streaming, embed_distributed) was removed when the transformer engines came out. Embedding is now dispatched exclusively through VectorEncoder::embed_root.

Surviving items:

• SearchConfig — pipeline tuning parameters (walk filters, batch size, corpus).
• Corpus — which chunks participate (code / docs / all).
• RerankPolicy — whether the cross-encoder reranker fires (auto / always / never).
• PROSE_EXTENSIONS — canonical prose file extensions.
• SearchResult — chunk + similarity score pair.
• apply_structural_boost — PageRank boost post-processing for MCP.

4.0 API change

Scope is replaced by two orthogonal axes:

• Corpus controls which chunks are searched (previously coupled to rerank policy in Scope).
• RerankPolicy controls whether the cross-encoder reranker fires (previously implicit).

A backward-compatibility type alias Scope = Corpus is provided so that callers outside the MCP layer (examples, CLI) continue to compile; prefer Corpus in new code.

Structs

SearchConfig

Runtime configuration for the search pipeline.

SearchResult

A search result pairing a code chunk with its similarity score.

Enums

Corpus

Corpus axis for a search invocation: which chunks participate.

RerankPolicy

Reranker firing policy for a search invocation.

Constants

DEFAULT_BATCH_SIZE

Default batch size for embedding inference.

PROSE_EXTENSIONS

Canonical prose file extensions for Scope::Docs. Kept in sync with crate::encoder::ripvec::ranking::is_prose_path.

Functions

apply_structural_boost

Normalize similarity scores to [0,1] and apply a PageRank structural boost.

Type Aliases

Scope

Backward-compatibility alias: Scope is now Corpus.