Module vector_index 

Vector index facade for cass.

cass uses the frankensearch FSVI vector index format and search primitives (via the frankensearch crate). The older CVVI format has been retired.

This module keeps cass-specific helpers (paths, role codes) in one place.

Structs

SearchParams

Configurable parameters for vector search parallelism.

SemanticDocId

Semantic doc_id fields encoded into FSVI records.

SemanticFilter

Semantic filter constraints expressed in numeric IDs for fast evaluation.

SemanticFilterMaps

Lookup maps for converting human filters (agent slug, workspace path, source id) into compact numeric IDs embedded into semantic doc_id strings.

VectorIndex

VectorIndexWriter

VectorSearchResult

Collapsed semantic search hit (best chunk per message).

Enums

Quantization

Vector element quantization stored in the FSVI slab.

Constants

ROLE_ASSISTANT

ROLE_SYSTEM

ROLE_TOOL

ROLE_USER

VECTOR_INDEX_DIR

Directory under the cass data dir where vector artifacts are stored.

Functions

dot_product_f16_scalar_bench

Scalar dot product benchmark helper for f16 stored vectors vs f32 query.

dot_product_f16_simd_bench

SIMD dot product benchmark helper for f16 stored vectors vs f32 query.

dot_product_scalar_bench

Scalar dot product benchmark helper.

dot_product_simd_bench

SIMD dot product benchmark helper (uses frankensearch’s portable SIMD).

parse_role_codes

Parse a list of role strings into a set of role codes.

parse_semantic_doc_id

Parse a cass semantic doc_id string.

role_code_from_str

Map a role string (from SQLite / connectors) to a compact u8 code.

vector_index_path

Path to the primary FSVI vector index for a given embedder.