Index Parameters¶

Index parameters control how vector and scalar indexes are built. All parameter classes inherit from Zvec::IndexParams.

Inheritance Hierarchy¶

IndexParams (abstract)
  ├── VectorIndexParams (abstract)
  │     ├── HnswIndexParams
  │     ├── FlatIndexParams
  │     └── IVFIndexParams
  └── InvertIndexParams

IndexParams (Base)¶

All index params share these methods:

Method	Returns	Description
`type`	`IndexType`	HNSW, IVF, FLAT, or INVERT
`vector_index_type?`	Boolean	True for vector index types
`to_s`	String	Human-readable description

VectorIndexParams (Base)¶

Shared by all vector index types:

Method	Returns	Description
`metric_type`	`MetricType`	Distance metric (COSINE, L2, IP)
`metric_type=`	—	Set distance metric
`quantize_type`	`QuantizeType`	Quantization (FP16, INT8, INT4, or UNDEFINED)
`quantize_type=`	—	Set quantization

HnswIndexParams¶

params = Zvec::HnswIndexParams.new(metric_type, m: 50, ef_construction: 500,
           quantize_type: Zvec::QuantizeType::UNDEFINED)

Parameter	Type	Default	Description
`metric_type`	`MetricType`	required	Distance metric
`m`	Integer	`50`	Max connections per node
`ef_construction`	Integer	`500`	Search width during index build
`quantize_type`	`QuantizeType`	`UNDEFINED`	Optional quantization

Instance methods:

Method	Returns	Description
`m`	Integer	Max connections
`m=`	—	Set max connections
`ef_construction`	Integer	Construction search width
`ef_construction=`	—	Set construction search width

FlatIndexParams¶

params = Zvec::FlatIndexParams.new(metric_type, quantize_type: Zvec::QuantizeType::UNDEFINED)

Parameter	Type	Default	Description
`metric_type`	`MetricType`	required	Distance metric
`quantize_type`	`QuantizeType`	`UNDEFINED`	Optional quantization

No additional instance methods beyond VectorIndexParams.

IVFIndexParams¶

params = Zvec::IVFIndexParams.new(metric_type, n_list: 1024, n_iters: 10,
           use_soar: false, quantize_type: Zvec::QuantizeType::UNDEFINED)

Parameter	Type	Default	Description
`metric_type`	`MetricType`	required	Distance metric
`n_list`	Integer	`1024`	Number of clusters
`n_iters`	Integer	`10`	K-means iterations
`use_soar`	Boolean	`false`	Enable SOAR optimization
`quantize_type`	`QuantizeType`	`UNDEFINED`	Optional quantization

Instance methods:

Method	Returns	Description
`n_list`	Integer	Number of clusters
`n_list=`	—	Set cluster count
`n_iters`	Integer	K-means iterations
`n_iters=`	—	Set iteration count
`use_soar?`	Boolean	SOAR optimization enabled
`use_soar=`	—	Set SOAR optimization

InvertIndexParams¶

For scalar field indexes used in filtered queries:

params = Zvec::InvertIndexParams.new(
           enable_range_optimization: true,
           enable_extended_wildcard: false)

Parameter	Type	Default	Description
`enable_range_optimization`	Boolean	`true`	Optimize range queries
`enable_extended_wildcard`	Boolean	`false`	Extended wildcard support

Instance methods:

Method	Returns	Description
`enable_range_optimization?`	Boolean	Range optimization enabled
`enable_range_optimization=`	—	Set range optimization
`enable_extended_wildcard?`	Boolean	Extended wildcards enabled
`enable_extended_wildcard=`	—	Set extended wildcards