WeaviateEx.API.VectorConfig (WeaviateEx v0.7.4)
View SourceVector configuration builders for Phase 2.4.
Provides builder functions for:
- 25+ vectorizer configurations
- 3 index types (HNSW, FLAT, DYNAMIC)
- 3 quantization methods (PQ, BQ, SQ)
- Complete collection configurations
Usage
config = VectorConfig.new("Article")
|> VectorConfig.with_vectorizer(:text2vec_openai, model: "text-embedding-ada-002")
|> VectorConfig.with_hnsw_index(ef: 100, max_connections: 64)
|> VectorConfig.with_product_quantization(enabled: true)
|> VectorConfig.with_properties([
%{name: "title", dataType: ["text"]}
])Summary
Functions
Configure Binary Quantization (BQ)
Configure a custom/unlisted vectorizer module.
List all distance metrics
Configure DYNAMIC index
Configure FLAT index
Configure HNSW index.
Configure img2vec-neural vectorizer for images.
Configure multi2multivec-jinaai vectorizer.
Configure multi2vec-aws for multimodal embeddings.
Configure multi2vec-bind vectorizer
Configure multi2vec-clip vectorizer
Configure multi2vec-cohere for multimodal embeddings.
Configure multi2vec-google (Palm) for multimodal embeddings.
Configure multi2vec-jinaai for multimodal embeddings.
Configure multi2vec-nvidia for multimodal embeddings.
Configure multi2vec-voyageai for multimodal embeddings.
Create new collection configuration
Configure no vectorizer (custom vectors)
Configure Product Quantization (PQ)
Configure ref2vec-centroid vectorizer.
Configure reranker-cohere module.
Configure reranker-contextualai module.
Configure reranker-jinaai module.
Configure reranker-nvidia module.
Configure reranker-transformers module (local model).
Configure reranker-voyageai module.
Configure Rotational Quantization (RQ).
Alias for rotational_quantization/1.
Configure Scalar Quantization (SQ).
Alias for scalar_quantization/1 with enabled defaulting to true.
List all supported vectorizers
Configure text2colbert-jinaai vectorizer (multi-vector).
Configure text2vec-aws vectorizer (deprecated, use text2vec_aws_bedrock or text2vec_aws_sagemaker).
Configure text2vec-aws with AWS Bedrock service.
Configure text2vec-aws with AWS SageMaker service.
Configure text2vec-azure-openai vectorizer.
Configure text2vec-cohere vectorizer.
Configure text2vec-contextionary vectorizer
Configure text2vec-databricks vectorizer.
Configure text2vec-google with Google AI Studio (Gemini).
Configure text2vec-google with Google Vertex AI.
Configure text2vec-gpt4all vectorizer
Configure text2vec-huggingface vectorizer
Configure text2vec-jinaai vectorizer.
Configure text2vec-mistral vectorizer.
Configure text2vec-model2vec vectorizer.
Configure text2vec-morph vectorizer.
Configure text2vec-nvidia vectorizer.
Configure text2vec-ollama vectorizer.
Configure text2vec-openai vectorizer
Configure text2vec-palm vectorizer (deprecated, use text2vec_google_vertex or text2vec_google_gemini).
Configure text2vec-transformers vectorizer
Configure text2vec-voyageai vectorizer.
Configure text2vec-weaviate vectorizer (Weaviate-hosted embeddings).
Check if distance metric is valid
Check if vectorizer is valid
Add Binary Quantization to configuration
Add a custom vectorizer to the collection configuration.
Add DYNAMIC index to configuration
Add FLAT index to configuration
Add HNSW index to configuration
Add multi-tenancy configuration
Add named vectors configuration
Add Product Quantization to configuration
Add properties to configuration
Add replication configuration.
Add reranker configuration to a collection.
Add Rotational Quantization to configuration
Add Scalar Quantization to configuration
Add sharding configuration
Add vectorizer to configuration
Types
Functions
Configure Binary Quantization (BQ)
Configure a custom/unlisted vectorizer module.
Use this for vectorizers not explicitly supported or for user-provided modules.
Options
All options are passed directly to the module configuration with automatic snake_case to camelCase conversion.
Examples
# Custom vectorizer with arbitrary configuration
VectorConfig.custom("my-custom-vectorizer",
model: "custom-model",
api_endpoint: "https://custom-api.example.com",
dimensions: 768
)
# Use in collection creation
VectorConfig.new("Article")
|> VectorConfig.with_custom_vectorizer("text2vec-custom", model: "my-model")List all distance metrics
Configure DYNAMIC index
Configure FLAT index
Configure HNSW index.
Options
:distance- Distance metric (:cosine, :dot, :l2_squared, :hamming, :manhattan):ef- Query time ef parameter (default: -1):ef_construction- Index build ef parameter (default: 128):max_connections- Maximum connections per node (default: 32):filter_strategy- Filter strategy (:sweeping or :acorn):quantizer- Quantization config (PQ, BQ, SQ, or RQ)
Configure img2vec-neural vectorizer for images.
Options
:image_fields- Image property names
Configure multi2multivec-jinaai vectorizer.
Options
:model- Model to use:base_url- Base URL for API (optional):image_fields- List of image property names (optional):text_fields- List of text property names (optional)
Configure multi2vec-aws for multimodal embeddings.
Options
:model- AWS model (required):region- AWS region (required):service- AWS service (bedrock or sagemaker):image_fields- Image property fields (optional):text_fields- Text property fields (optional)
Configure multi2vec-bind vectorizer
Configure multi2vec-clip vectorizer
Configure multi2vec-cohere for multimodal embeddings.
Options
:model- Cohere model (optional):image_fields- Image property fields (optional):text_fields- Text property fields (optional):truncate- Truncation mode (optional)
Configure multi2vec-google (Palm) for multimodal embeddings.
Options
:project_id- Google Cloud project ID (required):location- Model location (required):model- Model ID (optional):dimensions- Output dimensions (optional):image_fields- Image property fields (optional):text_fields- Text property fields (optional):video_fields- Video property fields (optional)
Configure multi2vec-jinaai for multimodal embeddings.
Options
:model- Jina model (optional):image_fields- Image property fields (optional):text_fields- Text property fields (optional):base_url- API base URL (optional)
Configure multi2vec-nvidia for multimodal embeddings.
Options
:model- NVIDIA model (optional):image_fields- Image property fields (optional):text_fields- Text property fields (optional):base_url- API base URL (optional)
Configure multi2vec-voyageai for multimodal embeddings.
Options
:model- VoyageAI model (optional):image_fields- Image property fields (optional):text_fields- Text property fields (optional):base_url- API base URL (optional)
Create new collection configuration
Configure no vectorizer (custom vectors)
Configure Product Quantization (PQ)
Configure ref2vec-centroid vectorizer.
Creates vectors from referenced objects using centroid calculation.
Options
:reference_properties- List of reference property names
Configure reranker-cohere module.
Options
:model- Model to use (optional):base_url- Base URL for API (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:cohere, model: "rerank-multilingual-v3.0")Configure reranker-contextualai module.
Contextual AI reranker with retrieval augmentation capabilities.
Options
:model- Model to use:base_url- Base URL for API (optional):context_source- Context source (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:contextualai, model: "contextual-rerank-v1")Configure reranker-jinaai module.
Options
:model- Model to use (e.g., "jina-reranker-v2-base-multilingual"):base_url- Base URL for API (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:jinaai, model: "jina-reranker-v2-base-multilingual")Configure reranker-nvidia module.
Options
:model- Model to use (e.g., "nvidia/nv-rerankqa-mistral-4b-v3"):base_url- Base URL for API (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:nvidia, model: "nvidia/nv-rerankqa-mistral-4b-v3")Configure reranker-transformers module (local model).
Runs locally using a transformer model.
Options
:model- Model to use (optional):inference_url- Inference URL (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:transformers)Configure reranker-voyageai module.
Options
:model- Model to use (e.g., "rerank-2", "rerank-lite-1"):base_url- Base URL for API (optional):truncate- Truncation mode (optional)
Example
VectorConfig.new("Article")
|> VectorConfig.with_reranker(:voyageai, model: "rerank-2")Configure Rotational Quantization (RQ).
RQ is an advanced quantization method that uses rotational transformations.
Options
:enabled- Enable RQ (default: true):cache- Enable cache (optional):bits- Number of bits for quantization (default: 8):rescore_limit- Number of candidates to rescore (optional):training_limit- Number of vectors to train on (optional)
Example
VectorConfig.hnsw_index(
quantizer: VectorConfig.rotational_quantization(bits: 8, cache: true)
)Alias for rotational_quantization/1.
Example
VectorConfig.hnsw_index(
quantizer: VectorConfig.rq(bits: 8, cache: true)
)Configure Scalar Quantization (SQ).
SQ provides memory reduction through scalar quantization of vectors.
Options
:enabled- Enable SQ (default: false):cache- Enable cache (optional):rescore_limit- Number of candidates to rescore (optional):training_limit- Number of vectors to train on (optional)
Example
VectorConfig.hnsw_index(
quantizer: VectorConfig.scalar_quantization(enabled: true, cache: true)
)Alias for scalar_quantization/1 with enabled defaulting to true.
Example
VectorConfig.hnsw_index(
quantizer: VectorConfig.sq(training_limit: 50_000)
)List all supported vectorizers
Configure text2colbert-jinaai vectorizer (multi-vector).
Options
:model- Model to use:dimensions- Output dimensions (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-aws vectorizer (deprecated, use text2vec_aws_bedrock or text2vec_aws_sagemaker).
Configure text2vec-aws with AWS Bedrock service.
Options
:model- The model to use (required, e.g., "amazon.titan-embed-text-v1"):region- AWS region (required):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Example
VectorConfig.text2vec_aws_bedrock(
model: "amazon.titan-embed-text-v1",
region: "us-east-1"
)Configure text2vec-aws with AWS SageMaker service.
Options
:endpoint- The SageMaker endpoint (required):region- AWS region (required):target_model- Target model (optional):target_variant- Target variant (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Example
VectorConfig.text2vec_aws_sagemaker(
endpoint: "my-endpoint",
region: "us-east-1"
)Configure text2vec-azure-openai vectorizer.
Options
:resource_name- Azure resource name (required):deployment_id- Azure deployment ID (required):base_url- Custom base URL (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-cohere vectorizer.
Options
:model- Model to use (optional):dimensions- Output dimensions (optional, new in Python client):truncate- Truncation mode (optional):base_url- Base URL for API (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-contextionary vectorizer
Configure text2vec-databricks vectorizer.
Options
:endpoint- Databricks serving endpoint (required):instruction- Instruction prefix (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-google with Google AI Studio (Gemini).
Options
:model- Model to use (optional):dimensions- Output dimensions (optional):title_property- Property to use as title (optional):task_type- Task type for embeddings (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Example
VectorConfig.text2vec_google_gemini(model: "text-embedding-004")Configure text2vec-google with Google Vertex AI.
Options
:project_id- Google Cloud project ID (required):api_endpoint- API endpoint (optional):model- Model to use (optional):dimensions- Output dimensions (optional):title_property- Property to use as title (optional):task_type- Task type for embeddings (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Example
VectorConfig.text2vec_google_vertex(
project_id: "my-project",
model: "textembedding-gecko@001"
)Configure text2vec-gpt4all vectorizer
Configure text2vec-huggingface vectorizer
Configure text2vec-jinaai vectorizer.
Options
:model- Jina model (e.g., "jina-embeddings-v3", "jina-embeddings-v4"):base_url- API base URL (optional):dimensions- Output dimensions (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-mistral vectorizer.
Options
:model- Mistral model name:base_url- Base URL for API (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-model2vec vectorizer.
Options
:inference_url- URL for inference service (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-morph vectorizer.
Options
:model- Model to use:base_url- Base URL for API (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-nvidia vectorizer.
Options
:model- NVIDIA model name:base_url- Base URL for API (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-ollama vectorizer.
Options
:model- Ollama model name:api_endpoint- Ollama API endpoint (default: http://localhost:11434):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-openai vectorizer
Configure text2vec-palm vectorizer (deprecated, use text2vec_google_vertex or text2vec_google_gemini).
Configure text2vec-transformers vectorizer
Configure text2vec-voyageai vectorizer.
Options
:model- Model to use (e.g., "voyage-3.5", "voyage-3-large", "voyage-context-3"):base_url- Base URL for API (optional):truncation- Whether to truncate (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Configure text2vec-weaviate vectorizer (Weaviate-hosted embeddings).
Options
:model- Model name:base_url- Base URL for API (optional):vectorize_collection_name- Whether to vectorize the collection name (default: true)
Check if distance metric is valid
Check if vectorizer is valid
Add Binary Quantization to configuration
Add a custom vectorizer to the collection configuration.
Examples
VectorConfig.new("Article")
|> VectorConfig.with_custom_vectorizer("text2vec-custom", model: "my-model")Add DYNAMIC index to configuration
Add FLAT index to configuration
Add HNSW index to configuration
Add multi-tenancy configuration
Add named vectors configuration
Add Product Quantization to configuration
Add properties to configuration
Add replication configuration.
Options
:factor- Number of replicas (default: 1):async_enabled- Enable async replication (v1.26.0+, optional):deletion_strategy- Conflict resolution strategy (optional):delete_on_conflict- Delete object on conflict:no_automated_resolution- No automated conflict resolution:time_based_resolution- Use timestamp for resolution
Add reranker configuration to a collection.
Rerankers
:cohere- Cohere reranker:transformers- Local transformers reranker:voyageai- VoyageAI reranker:jinaai- Jina AI reranker:nvidia- NVIDIA reranker:contextualai- Contextual AI reranker
Example
VectorConfig.new("Article")
|> VectorConfig.with_vectorizer(:text2vec_openai)
|> VectorConfig.with_reranker(:cohere, model: "rerank-multilingual-v3.0")Add Rotational Quantization to configuration
Add Scalar Quantization to configuration
Add sharding configuration
Add vectorizer to configuration