Struct CreateCollectionRequest 

pub struct CreateCollectionRequest {

    pub name: String,
    pub dimension: Option<usize>,
    pub metric: String,
    pub storage_mode: String,
    pub collection_type: String,
    pub hnsw_m: Option<usize>,
    pub hnsw_ef_construction: Option<usize>,
    pub hnsw_alpha: Option<f32>,
    pub hnsw_max_elements: Option<usize>,
    pub pq_rescore_oversampling: Option<u32>,
    pub deferred_indexing: Option<Value>,
    pub async_index_builder: Option<Value>,
    pub graph_schema: Option<Value>,
}

Request to create a new collection.

Fields

name: String

Collection name.

dimension: Option<usize>

Vector dimension (required for vector collections, ignored for metadata_only).

metric: String

Distance metric (cosine, euclidean, dot, hamming, jaccard).

storage_mode: String

Storage mode (full, sq8, binary). Defaults to full.

collection_type: String

Collection type: “vector” (default) or “metadata_only”.

hnsw_m: Option<usize>

HNSW M parameter (number of bi-directional links per node).

hnsw_ef_construction: Option<usize>

HNSW ef_construction parameter (candidate list size during build).

hnsw_alpha: Option<f32>

VAMANA alpha for neighbor diversification (default derived from dimension).

Values > 1.0 bias select_neighbors toward diversity over proximity, producing a more navigable graph with better recall. The engine default is 1.2 (VAMANA paper recommendation). Set to 1.0 for strict nearest-neighbor selection.

hnsw_max_elements: Option<usize>

Initial HNSW capacity (grows automatically if exceeded).

Pre-sizing matters for bulk imports: the engine avoids repeated realloc-and-copy cycles when the final population is known upfront. Use this when migrating a large existing corpus into a fresh collection.

pq_rescore_oversampling: Option<u32>

PQ rescore oversampling factor (default 4).

The search pipeline fetches max(k * factor, k + 32) candidates from HNSW and rescores them with full-precision ADC. Only meaningful for quantised storage modes (SQ8, PQ). Some(0) is treated as “disabled”.

deferred_indexing: Option<Value>

Deferred indexing configuration (US-366).

When present, inserts are buffered in memory and batch-merged into the HNSW index when the buffer reaches merge_threshold. Accepted as a free-form JSON object matching velesdb_core::collection::streaming::DeferredIndexerConfig.

async_index_builder: Option<Value>

Async index builder configuration (Issue #488 — Bulk Insert V2).

When present, enables the AsyncIndexBuilder for deferred HNSW insertion during bulk import. Accepted as a free-form JSON object matching velesdb_core::collection::streaming::AsyncIndexBuilderConfig.

graph_schema: Option<Value>

Graph schema (only for collection_type = "graph").

Accepted as a free-form JSON object matching velesdb_core::GraphSchema. GraphSchema::schemaless() is used when the field is absent.

Trait Implementations

impl Debug for CreateCollectionRequest

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for CreateCollectionRequest

fn deserialize<__D>( __deserializer: __D, ) -> Result<CreateCollectionRequest, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl ToSchema for CreateCollectionRequest

fn name() -> Cow<'static, str>

Return name of the schema.

fn schemas(schemas: &mut Vec<(String, RefOr<Schema>)>)

Implement reference utoipa::openapi::schema::Schemas for this type.