Attributes Reference

Complete reference for all managed attributes.

Overview

Attribute	Target	Purpose
`WeaviateCollection`	Class	Collection name, sharding, replication, multi-tenancy
`InvertedIndex`	Class	Inverted index settings (timestamps, null state)
`Generative<T>`	Class	RAG configuration (15+ providers)
`Reranker<T>`	Class	Reranking configuration (6 providers)
`QueryProjection<T>`	Class	Query projection type for a collection entity
`Property`	Property	Data type, name, description
`Index`	Property	Filtering, searching, range filters
`Tokenization`	Property	Text tokenization strategy
`Vector<T>`	Property	Named vector with vectorizer
`VectorIndex<T>`	Property	Vector index configuration (HNSW, Flat, Dynamic)
`Quantizer*`	Property	Vector compression (BQ, PQ, SQ, RQ)
`Encoding`	Property	Multi-vector (ColBERT) configuration
`Reference`	Property	Cross-reference to another collection
`NestedType`	Property	Nested object type override
`MetadataProperty`	Property	Inject query metadata (score, distance)
`WeaviateUUID`	Property	Marks a property as the entity's UUID
`MapFrom`	Property	Maps projection property from a different source name
`Vector` (projection)	Property	Includes a named vector in projection results
`Reference` (projection)	Property	Includes a reference in projection results
`QueryAggregate<T>`	Class	Marks a type as an aggregation projection for a collection
`Metrics`	Property	Specifies which aggregate metrics to compute (Number, Integer, Text, Boolean, Date)

Class-Level Attributes

WeaviateCollection

Defines a Weaviate collection. Required on model classes.

[WeaviateCollection("Articles", Description = "Blog articles")]
public class Article { }

Properties

Property	Type	Default	Description
`Name`	`string?`	Class name	Collection name in Weaviate
`Description`	`string?`	null	Collection description
Multi-Tenancy
`MultiTenancyEnabled`	`bool?`	false	Enable multi-tenancy (immutable)
`AutoTenantCreation`	`bool?`	null	Auto-create tenants on reference
`AutoTenantActivation`	`bool?`	null	Auto-activate created tenants
Sharding
`ShardingDesiredCount`	`int`	-1 (default)	Number of shards (immutable)
`ShardingVirtualPerPhysical`	`int`	-1 (default)	Virtual shards per physical
`ShardingDesiredVirtualCount`	`int`	-1 (default)	Total virtual shards
`ShardingKey`	`string?`	"_id"	Property to shard on
Replication
`ReplicationFactor`	`int`	-1 (default)	Number of replicas (immutable)
`ReplicationAsyncEnabled`	`bool`	false	Enable async replication
Lifecycle
`CollectionConfigMethod`	`string?`	null	Static method name for lifecycle hooks
`ConfigMethodClass`	`Type?`	null	Class containing the config method

CollectionConfigMethod Signature

The CollectionConfigMethod must point to a static method that accepts an OnCollectionConfig parameter:

[WeaviateCollection("Products", CollectionConfigMethod = nameof(OnConfig))]
public class Product
{
    [Property]
    public string Name { get; set; } = "";

    public static void OnConfig(OnCollectionConfig config)
    {
        config.OnCreate(createParams =>
        {
            // Modify collection creation params before they're sent to Weaviate
            return createParams;
        });
    }
}

Examples

// Basic
[WeaviateCollection("Products")]
public class Product { }

// Multi-tenant
[WeaviateCollection("Orders",
    MultiTenancyEnabled = true,
    AutoTenantCreation = true,
    AutoTenantActivation = true)]
public class Order { }

// Sharded and replicated
[WeaviateCollection("Events",
    ShardingDesiredCount = 3,
    ReplicationFactor = 2)]
public class Event { }

InvertedIndex

Configures inverted index settings at collection level.

[WeaviateCollection("Articles")]
[InvertedIndex(IndexTimestamps = true)]
public class Article { }

Properties

Property	Type	Default	Description
`IndexTimestamps`	`bool`	false	Index creation/update timestamps
`IndexNullState`	`bool`	false	Index null values for filtering
`IndexPropertyLength`	`bool`	false	Index property lengths
`CleanupIntervalSeconds`	`int`	-1 (default)	Cleanup interval

Generative<T>

Configures RAG (Retrieval Augmented Generation) module.

[WeaviateCollection("Articles")]
[Generative<GenerativeConfig.OpenAI>(Model = "gpt-4")]
public class Article { }

Supported Providers

Type	Provider
`GenerativeConfig.OpenAI`	OpenAI
`GenerativeConfig.AzureOpenAI`	Azure OpenAI
`GenerativeConfig.Anthropic`	Anthropic Claude
`GenerativeConfig.Cohere`	Cohere
`GenerativeConfig.AWS`	AWS Bedrock
`GenerativeConfig.Google`	Google AI
`GenerativeConfig.Mistral`	Mistral AI
`GenerativeConfig.Ollama`	Ollama (local)
`GenerativeConfig.Nvidia`	NVIDIA
`GenerativeConfig.Databricks`	Databricks
`GenerativeConfig.Anyscale`	Anyscale
`GenerativeConfig.FriendlyAI`	FriendliAI
`GenerativeConfig.OctoAI`	OctoAI
`GenerativeConfig.XAI`	xAI

Common Properties

Property	Type	Default	Description
`Model`	`string?`	null	Model name
`BaseURL`	`string?`	null	Custom API endpoint
`MaxTokens`	`int`	-1 (default)	Max tokens to generate
`Temperature`	`double`	-1 (default)	Randomness (0-1)
`TopP`	`double`	-1 (default)	Nucleus sampling
`TopK`	`int`	-1 (default)	Token sampling

Provider-Specific Properties

Property	Providers	Description
`ResourceName`	AzureOpenAI	Azure resource name
`DeploymentId`	AzureOpenAI	Azure deployment ID
`ProjectId`	Google	Google Cloud project
`Region`	AWS	AWS region
`Service`	AWS	AWS service (bedrock)

Reranker<T>

Configures reranking module for result re-ordering.

[WeaviateCollection("Articles")]
[Reranker<RerankerConfig.Cohere>(Model = "rerank-english-v3.0")]
public class Article { }

Supported Providers

Type	Provider
`RerankerConfig.Cohere`	Cohere
`RerankerConfig.VoyageAI`	Voyage AI
`RerankerConfig.JinaAI`	Jina AI
`RerankerConfig.Nvidia`	NVIDIA
`RerankerConfig.ContextualAI`	Contextual AI
`RerankerConfig.Transformers`	Local transformers

Properties

Property	Type	Default	Description
`Model`	`string?`	null	Model name
`BaseURL`	`string?`	null	Custom API endpoint
`TopN`	`int`	-1 (default)	Return top N results
`Instruction`	`string?`	null	Query instruction (some providers)

QueryProject<T>

Marks a type as a query projection for a specific collection entity type. Projections allow selecting a subset of properties with optional renaming, metadata injection, and vector inclusion.

[QueryProjection<Article>]
public class ArticleSummary
{
    public string Title { get; set; } = "";

    [MapFrom(nameof(Article.WordCount))]
    public int Words { get; set; }

    [MetadataProperty]
    public double? Score { get; set; }
}

Use projections with Project<TProjection>():

var summaries = await context.Products.Project<ArticleSummary>()
    .NearText("search term")
    .Execute();

Property-Level Attributes

Property

Defines a Weaviate property. Required for all stored properties.

[Property(DataType.Text)]
public string Title { get; set; }

// Type inference (recommended)
[Property]
public string Title { get; set; }

Properties

Property	Type	Default	Description
`DataType`	`DataType`	`Unknown` (inferred)	Weaviate data type
`Name`	`string?`	Property name (camelCase)	Custom property name
`Description`	`string?`	null	Property description

Type Inference

C# Type	Inferred DataType
`string`	`Text`
`int`, `long`	`Int`
`float`, `double`, `decimal`	`Number`
`bool`	`Bool`
`DateTime`, `DateTimeOffset`	`Date`
`Guid`	`Uuid`
`byte[]`	`Blob`
`List<string>`, `string[]`	`TextArray`
`List<int>`, `int[]`	`IntArray`
Class type	`Object`
`List<Class>`	`ObjectArray`

Index

Configures indexing for filtering and searching.

[Property]
[Index(Filterable = true, Searchable = true)]
public string Title { get; set; }

Properties

Property	Type	Default	Description
`Filterable`	`bool`	false	Enable equality/contains filters
`Searchable`	`bool`	false	Enable BM25 text search
`RangeFilters`	`bool`	false	Enable >, <, >=, <= filters

Tokenization

Specifies text tokenization strategy.

[Property]
[Tokenization(PropertyTokenization.Word)]
public string Title { get; set; }

Values

Value	Description
`Word`	Standard word tokenization (default)
`Lowercase`	Lowercase before tokenization
`Whitespace`	Split on whitespace only
`Field`	Treat entire field as single token
`Trigram`	Character trigrams
`Gse`	Chinese segmentation
`Kagome`	Japanese tokenization

Vector<T>

Defines a named vector with vectorizer configuration.

[Vector<Vectorizer.Text2VecOpenAI>(
    Model = "text-embedding-ada-002",
    SourceProperties = [nameof(Title), nameof(Content)]
)]
public float[]? Embedding { get; set; }

Property Type Determines Vector Type

C# Type	Vector Type
`float[]`	Single vector
`float[,]`	Multi-vector (ColBERT)

Common Properties

Property	Type	Default	Description
`Name`	`string?`	Property name (camelCase)	Custom vector name
`SourceProperties`	`string[]?`	null	Properties to vectorize
`VectorizeCollectionName`	`bool`	false	Include collection name
`Model`	`string?`	null	Model name
`Dimensions`	`int?`	null	Embedding dimensions
`BaseURL`	`string?`	null	Custom API endpoint

Multi-Modal Properties

Property	Type	Description
`TextFields`	`string[]?`	Text fields for multi-modal
`ImageFields`	`string[]?`	Image fields for multi-modal
`VideoFields`	`string[]?`	Video fields for multi-modal

Ref2Vec Properties

Property	Type	Description
`ReferenceProperties`	`string[]?`	Reference properties for Ref2Vec

ConfigMethod

[Vector<Vectorizer.Text2VecOpenAI>(
    Model = "text-embedding-3-small",
    ConfigMethod = nameof(ConfigureVector)
)]
public float[]? Embedding { get; set; }

public static Vectorizer.Text2VecOpenAI ConfigureVector(
    string vectorName,
    Vectorizer.Text2VecOpenAI prebuilt)
{
    prebuilt.VectorizeCollectionName = false;
    return prebuilt;
}

VectorIndex<T>

Configures the vector index type.

[Vector<Vectorizer.Text2VecOpenAI>]
[VectorIndex<VectorIndexConfig.Hnsw>(
    EfConstruction = 128,
    MaxConnections = 32
)]
public float[]? Embedding { get; set; }

Index Types

Type	Description
`VectorIndexConfig.Hnsw`	HNSW index (default, recommended)
`VectorIndexConfig.Flat`	Brute-force exact search
`VectorIndexConfig.Dynamic`	Auto-switches from Flat to HNSW

HNSW Properties

Property	Type	Default	Description
`EfConstruction`	`int`	128	Build-time quality
`MaxConnections`	`int`	32	Graph connectivity
`Ef`	`int`	-1	Query-time quality
`DynamicEfMin`	`int`	100	Min dynamic EF
`DynamicEfMax`	`int`	500	Max dynamic EF
`DynamicEfFactor`	`int`	8	Dynamic EF multiplier

Distance Metrics

Value	Description
`Cosine`	Cosine similarity (default)
`Dot`	Dot product
`L2Squared`	Euclidean distance squared
`Hamming`	Hamming distance (binary)

Quantizers

Vector compression for reduced memory usage.

QuantizerBQ (Binary Quantization)

[Vector<Vectorizer.Text2VecOpenAI>]
[QuantizerBQ(Cache = true, RescoreLimit = 200)]
public float[]? Embedding { get; set; }

Property	Type	Default	Description
`Cache`	`bool`	true	Cache quantized vectors
`RescoreLimit`	`int`	-1	Candidates for rescoring

QuantizerPQ (Product Quantization)

[Vector<Vectorizer.Text2VecOpenAI>]
[QuantizerPQ(Segments = 128, Centroids = 256)]
public float[]? Embedding { get; set; }

Property	Type	Default	Description
`Segments`	`int`	0	Number of segments
`Centroids`	`int`	256	Centroids per segment
`TrainingLimit`	`int`	100000	Training samples
`EncoderType`	`PQEncoderType`	Tile	Encoder type
`EncoderDistribution`	`PQEncoderDistribution`	LogNormal	Distribution

QuantizerSQ (Scalar Quantization)

[Vector<Vectorizer.Text2VecOpenAI>]
[QuantizerSQ(TrainingLimit = 50000)]
public float[]? Embedding { get; set; }

Property	Type	Default	Description
`TrainingLimit`	`int`	100000	Training samples
`RescoreLimit`	`int`	-1	Candidates for rescoring

QuantizerRQ (Residual Quantization)

[Vector<Vectorizer.Text2VecOpenAI>]
[QuantizerRQ(Bits = 8)]
public float[]? Embedding { get; set; }

Property	Type	Default	Description
`Bits`	`int`	8	Bits per dimension
`Cache`	`bool`	true	Cache quantized vectors
`RescoreLimit`	`int`	-1	Candidates for rescoring

Encoding

Configures multi-vector (ColBERT) encoding.

[Vector<Vectorizer.SelfProvided>]
[Encoding(KSim = 3, DProjections = 512)]
public float[,]? ColBERTEmbedding { get; set; }

Properties

Property	Type	Default	Description
`KSim`	`int`	-1	K for k-nearest token similarity
`DProjections`	`int`	-1	Projection dimensions
`Repetitions`	`int`	-1	Repetitions

Reference

Defines a cross-reference to another collection.

// Single reference — target collection inferred from property type (Category)
[Reference]
public Category? Category { get; set; }

// Optional name override — Weaviate property name becomes "primaryCategory" by convention,
// use this only when you need a different name on the wire
[Reference("primaryCat")]
public Category? PrimaryCategory { get; set; }

// ID-only reference — Guid? cannot carry type info, so Target= is required
[Reference(Target = typeof(Author))]
public Guid? AuthorId { get; set; }

// Multi-reference — target inferred from List<Article>
[Reference]
public List<Article>? RelatedArticles { get; set; }

Properties

Property	Type	Default	Description
constructor `name`	`string?`	null	Weaviate property name override (defaults to camelCase of C# property name)
`Target`	`Type?`	null	Explicit target type — required when property is `Guid?` (target cannot be inferred)
`Description`	`string?`	null	Reference description
`Loading`	`ReferenceLoadingStrategy`	`Explicit`	Loading strategy (see below)
`SourceProperty`	`string?`	null	Used in projections to map from a differently-named source property

The target collection name is inferred at schema time:

If Target is set → use that type's [WeaviateCollection] name
If property type is Guid? and Target is not set → throws (target is ambiguous)
If property type is Category? → extract Category, look up its [WeaviateCollection] name
If property type is List<Article> → extract Article, look up its [WeaviateCollection] name

ReferenceLoadingStrategy

Controls when references are loaded in queries:

Value	Description
`Explicit`	Default. References are only loaded when `.WithReferences()` is called on the query.
`Eager`	References are automatically included in every query without needing `.WithReferences()`.

// Explicit loading (default) — requires .WithReferences()
[Reference]
public Category? Category { get; set; }

// Eager loading — auto-included in every query
[Reference(Loading = ReferenceLoadingStrategy.Eager)]
public Category? Category { get; set; }

NestedType

Overrides the inferred nested object type. Usually not needed.

// Auto-inferred (recommended)
[Property(DataType.Object)]
public Author Author { get; set; }

// Explicit (for interfaces/base classes)
[Property(DataType.Object)]
[NestedType(typeof(ConcreteAuthor))]
public IAuthor Author { get; set; }

Properties

Property	Type	Description
`Type`	`Type`	Concrete nested type

MetadataProperty

Injects query metadata into the property.

public class Article
{
    [Property]
    public string Title { get; set; }

    [MetadataProperty(MetadataType.Score)]
    public float? SearchScore { get; set; }

    [MetadataProperty(MetadataType.Distance)]
    public float? Distance { get; set; }
}

After executing a query with .WithMetadata(), these properties are automatically populated.

Values

Value	Type	Description
`Score`	`float?`	BM25/hybrid search score
`Distance`	`float?`	Vector distance
`Certainty`	`float?`	Vector certainty (1 - distance)
`CreationTime`	`DateTime?`	Object creation timestamp
`UpdateTime`	`DateTime?`	Last update timestamp
`ExplainScore`	`string?`	Score explanation

WeaviateUUID

Marks a property as the entity's Weaviate UUID. The property must be of type Guid or Guid?.

[WeaviateCollection("Books")]
public class Book
{
    [WeaviateUUID]
    public Guid BookId { get; set; }

    [Property]
    public string Title { get; set; } = "";
}

By default, the context looks for a property named UUID (matching Weaviate convention). Use [WeaviateUUID] to designate a differently-named property as the ID.

When inserting via WeaviateContext, the ID is automatically assigned back to the entity:

var book = new Book { Title = "My Book" };
await context.Insert(book);
Console.WriteLine(book.BookId); // Assigned Guid

MapFrom

Specifies the source property name for a projection property when it differs from the projection property name. Used on properties within [QueryProjection<T>] types.

[QueryProjection<Article>]
public class ArticleProjection
{
    [MapFrom(nameof(Article.WordCount))]
    public int Words { get; set; }
}

The constructor argument is the source property name in PascalCase. It will be converted to camelCase for Weaviate property lookup.

Vector (projection) {#vector-projection}

Marks a property in a query projection to receive a named vector from query results. The property must be of type float[].

[QueryProjection<Article>]
public class ArticleWithEmbedding
{
    public string Title { get; set; } = "";

    [Vector(VectorName = "embedding")]
    public float[]? Embedding { get; set; }
}

Vector (projection) Properties

Property	Type	Default	Description
`VectorName`	`string?`	null (uses property name as camelCase)	The name of the vector to include

Reference (projection) {#reference-projection}

Marks a property in a query projection to receive a cross-reference from query results. The property can be a single object, a List<T>, or a Guid? (ID-only).

[QueryProjection<Article>]
public class ArticleWithCategory
{
    public string Title { get; set; } = "";

    [Reference]
    public Category? Category { get; set; }

    // Rename source property
    [Reference(SourceProperty = "RelatedArticles")]
    public List<Article>? Related { get; set; }
}

Reference (projection) Properties

Property	Type	Default	Description
`SourceProperty`	`string?`	null (uses property name as camelCase)	The source reference property name

When used in a projection, reference includes are automatically configured on the query — no need to call .WithReferences() separately.

QueryAggregate

Marks a type as an aggregation projection for context-level Aggregate<TProjection>(). The generic argument T identifies the source collection type.

[QueryAggregate<Product>]
public class ProductStats
{
    [Metrics(Metric.Number.Mean, Metric.Number.Sum, Metric.Number.Count, Metric.Number.Min, Metric.Number.Max)]
    public Aggregate.Number Price { get; set; }

    [Metrics(Metric.Integer.Mean, Metric.Integer.Sum, Metric.Integer.Count)]
    public Aggregate.Integer Stock { get; set; }

    [Metrics(Metric.Text.Count, Metric.Text.TopOccurrences)]
    public Aggregate.Text Category { get; set; }

    [Metrics(Metric.Boolean.TotalTrue, Metric.Boolean.PercentageTrue)]
    public Aggregate.Boolean InStock { get; set; }
}

MetricsAttribute

Use [Metrics] to specify which aggregate metrics to compute. The attribute supports two patterns:

Pattern 1: Full Aggregate Type

Use with Aggregate.Number, Aggregate.Integer, etc. types to retrieve all metrics for a property:

[Metrics(Metric.Number.Mean, Metric.Number.Sum, Metric.Number.Count)]
public Aggregate.Number Price { get; set; }

Pattern 2: Single Metric Extraction

Extract a single metric to a scalar property by specifying the source property name:

[Metrics("price", Metric.Number.Mean)]
public double? AveragePrice { get; set; }

[Metrics("price", Metric.Number.Count)]
public long? PriceCount { get; set; }

[Metrics("stock", Metric.Integer.Sum)]
public long? TotalStock { get; set; }

[Metrics("createdAt", Metric.Date.Min)]
public DateTime? EarliestDate { get; set; }

The analyzer validates that the scalar type matches the metric (e.g., Mean → double?, Count → long?).

Naming Convention (WEAVIATE008): When using single-metric extraction, the analyzer recommends (warning) following the naming convention: PropertyName + MetricSuffix. For example, [Metrics("price", Metric.Number.Mean)] should use property name PriceMean (not AveragePrice). This is only a warning and can be ignored if you prefer more descriptive names.

// Recommended (follows convention):
[Metrics("price", Metric.Number.Mean)]
public double? PriceMean { get; set; }

[Metrics("price", Metric.Number.Sum)]
public double? PriceSum { get; set; }

[Metrics("stock", Metric.Integer.Count)]
public long? StockCount { get; set; }

// Also valid (analyzer warns but allows):
[Metrics("price", Metric.Number.Mean)]
public double? AveragePrice { get; set; }  // ⚠️ WEAVIATE008: Should be 'PriceMean'

Metrics Enum Types

Each aggregate type has its own enum of available metrics:

Metric.Number (for Aggregate.Number):

Mean, Sum, Count, Min, Max, Median, Mode

Metric.Integer (for Aggregate.Integer):

Mean, Sum, Count, Min, Max, Median, Mode

Metric.Text (for Aggregate.Text):

Count, TopOccurrences

Metric.Boolean (for Aggregate.Boolean):

Count, TotalTrue, TotalFalse, PercentageTrue, PercentageFalse

Metric.Date (for Aggregate.Date):

Count, Min, Max, Median, Mode

Type Matching

For full aggregate types, specify multiple metrics using params syntax (comma-separated values). The analyzer enforces that the metrics enum type matches the property's aggregate type.

For single metric extraction, the analyzer enforces the following type mappings:

Metric	Expected Scalar Type
`Number.Mean`, `Number.Sum`, `Number.Min`, `Number.Max`, `Number.Median`, `Number.Mode`	`double?`, `float?`, or `decimal?`
`Number.Count`	`long?` or `int?`
`Integer.Mean`	`double?` or `float?`
`Integer.Sum`, `Integer.Count`, `Integer.Min`, `Integer.Max`, `Integer.Median`, `Integer.Mode`	`long?` or `int?`
`Text.Count`	`long?` or `int?`
`Boolean.Count`, `Boolean.TotalTrue`, `Boolean.TotalFalse`	`long?` or `int?`
`Boolean.PercentageTrue`, `Boolean.PercentageFalse`	`double?` or `float?`
`Date.Count`	`long?` or `int?`
`Date.Min`, `Date.Max`, `Date.Median`, `Date.Mode`	`DateTime?` or `DateTimeOffset?`

The [QueryAggregate<T>] attribute tells the context which collection to aggregate against, so the collection doesn't need to be specified explicitly.

FilesExpand file tree

ATTRIBUTES.md

Latest commit

History

ATTRIBUTES.md

File metadata and controls

Attributes Reference

Overview

Class-Level Attributes

WeaviateCollection

Properties

CollectionConfigMethod Signature

Examples

InvertedIndex

Properties

Generative<T>

Supported Providers

Common Properties

Provider-Specific Properties

Reranker<T>

Supported Providers

Properties

QueryProject<T>

Property-Level Attributes

Property

Properties

Type Inference

Index

Properties

Tokenization

Values

Vector<T>

Property Type Determines Vector Type

Common Properties

Multi-Modal Properties

Ref2Vec Properties

ConfigMethod

VectorIndex<T>

Index Types

HNSW Properties

Distance Metrics

Quantizers

QuantizerBQ (Binary Quantization)

QuantizerPQ (Product Quantization)

QuantizerSQ (Scalar Quantization)

QuantizerRQ (Residual Quantization)

Encoding

Properties

Reference

Properties

ReferenceLoadingStrategy

NestedType

Properties

MetadataProperty

Values

WeaviateUUID

MapFrom

Vector (projection) {#vector-projection}

Vector (projection) Properties

Reference (projection) {#reference-projection}

Reference (projection) Properties

QueryAggregate

MetricsAttribute

Metrics Enum Types

Type Matching