Enums
ProviderName
Enumeration for supported vector database providers.
Values:
CHROMA = 'chroma'QDRANT = 'qdrant'WEAVIATE = 'weaviate'PINECONE = 'pinecone'MILVUS = 'milvus'FAISS = 'faiss'PG = "pgvector"
Mode
Enumeration for the operational mode of the provider.
Values:
CLOUD = 'cloud'LOCAL = 'local'EMBEDDED = 'embedded'IN_MEMORY = 'in_memory'
DistanceMetric
Enumeration for similarity calculation algorithms.
Values:
COSINE = 'Cosine'EUCLIDEAN = 'Euclidean'DOT_PRODUCT = 'DotProduct'
IndexType
Enumeration for the core Approximate Nearest Neighbor (ANN) index algorithm.
Values:
HNSW = 'HNSW'IVF_FLAT = 'IVF_FLAT'FLAT = 'FLAT'
WriteConsistency
Enumeration for write consistency in distributed databases.
Values:
STRONG = 'strong'EVENTUAL = 'eventual'
ConsistencyLevel
Values:
STRONG = "Strong"BOUNDED = "Bounded"SESSION = "Session"EVENTUALLY = "Eventually"
Classes
CoreConfig
Handles connection, identity, and the fundamental vector schema. Corresponds to Table 1: Core Configuration & Schema.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
db_path | Optional[str] | None | Database path for embedded mode |
provider_name | ProviderName | Required | The vector database provider to use |
mode | Mode | Required | The operational mode of the provider |
collection_name | str | "default_collection" | Name of the collection |
cloud | Optional[str] | None | Cloud provider identifier |
region | Optional[str] | None | Cloud region |
host | Optional[str] | None | Database host |
port | Optional[int] | None | Database port |
api_key | Optional[pydantic.SecretStr] | None | API key for authentication |
use_tls | bool | True | Whether to use TLS encryption |
vector_size | int | Required | Size of the dense vectors |
vector_size_sparse | Optional[int] | None | Size of the sparse vectors |
distance_metric | DistanceMetric | DistanceMetric.COSINE | Distance metric for similarity calculation |
recreate_if_exists | bool | False | Whether to recreate collection if it already exists |
HNSWTuningConfig
Fine-tunes the Hierarchical Navigable Small World (HNSW) index.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
index_type | Literal[IndexType.HNSW] | Required | Must be HNSW index type |
m | int | 16 | Number of bi-directional links created for every new element during construction |
ef_construction | int | 200 | Size of the dynamic candidate list for construction |
IVFTuningConfig
Fine-tunes Inverted File (IVF) based indexes.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
index_type | Literal[IndexType.IVF_FLAT] | Required | Must be IVF_FLAT index type |
nlist | int | 100 | Number of clusters for the index |
FlatTuningConfig
Configuration for a FLAT (brute-force) index. No tuning needed.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
index_type | Literal[IndexType.FLAT] | Required | Must be FLAT index type |
QuantizationConfig
Compresses vectors to reduce memory usage.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
quantization_type | Literal['Scalar', 'Product'] | Required | Type of quantization to apply |
bits | int | 8 | Number of bits for quantization |
PayloadIndexConfig
Defines a schema for a single payload index, allowing for optimized filtering on metadata fields.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
field_name | str | Required | Name of the field to index |
field_schema_type | Literal['text', 'keyword', 'integer', 'float', 'geo', 'boolean'] | Required | Schema type of the field |
params | Optional[Dict[str, Any]] | None | Additional parameters for the index |
enable_full_text_index | Optional[bool] | None | Whether to enable full-text search on this field |
IndexingConfig
Manages performance-critical aspects of index algorithm and memory usage. Corresponds to Table 2: Indexing, Storage & Performance Tuning.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
index_config | IndexTuningConfig | HNSWTuningConfig(index_type=IndexType.HNSW) | Index configuration |
quantization | Optional[QuantizationConfig] | None | Quantization settings |
payload_indexes | Optional[List[PayloadIndexConfig]] | None | List of payload field indexes |
create_dense_index | bool | True | Whether to create a dense vector index |
create_sparse_index | bool | False | Whether to create a sparse vector index |
SearchConfig
Defines the default parameters for all retrieval operations. Corresponds to Table 3: Search & Retrieval Operations. These can be overridden at query time.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
default_top_k | Optional[int] | None | Default number of results to return |
default_ef_search | Optional[int] | None | Default ef parameter for HNSW search |
default_nprobe | Optional[int] | None | Default nprobe parameter for IVF search |
default_hybrid_alpha | Optional[float] | None | Default alpha parameter for hybrid search |
default_fusion_method | Optional[Literal['rrf', 'weighted']] | None | Default fusion method for hybrid search |
default_similarity_threshold | Optional[float] | None | Default similarity threshold for results |
dense_search_enabled | Optional[bool] | None | Whether dense search is enabled |
full_text_search_enabled | Optional[bool] | None | Whether full-text search is enabled |
hybrid_search_enabled | Optional[bool] | None | Whether hybrid search is enabled |
filter | Optional[Dict[str, Any]] | None | Default filter to apply to all searches |
DataManagementConfig
Governs the behavior of data ingestion and lifecycle. Corresponds to Table 4: Data Management & Ingestion.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
batch_size | int | 128 | Number of records to process in each batch |
parallel_uploads | int | 4 | Number of parallel upload processes |
write_consistency | WriteConsistency | WriteConsistency.EVENTUAL | Write consistency level |
AdvancedConfig
Contains optional, enterprise-grade operational features. Corresponds to Table 5: Advanced & Operational Features.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
namespace | Optional[str] | None | Namespace for multi-tenancy |
num_shards | Optional[int] | None | Number of shards for the collection |
replication_factor | Optional[int] | None | Replication factor for the collection |
Config
The master configuration object for a VectorDBProvider.
This class is the single source of truth for a provider’s setup. It is composed of modular sub-configs, each handling a specific functional area. Upon instantiation, it performs a deep validation of all parameters and their interdependencies, ensuring a valid and complete configuration.
The object is immutable after creation to guarantee configuration stability.
Parameters:
| Parameter | Type | Default | Description |
|---|---|---|---|
core | CoreConfig | Required | Core configuration for connection and basic settings |
indexing | IndexingConfig | IndexingConfig() | Indexing configuration |
search | SearchConfig | SearchConfig() | Search configuration |
data_management | DataManagementConfig | DataManagementConfig() | Data management configuration |
advanced | AdvancedConfig | AdvancedConfig() | Advanced configuration |