Feature catalog

An outline of all the features offered by LanceDB, across its open source, cloud, and enterprise offerings.

Storage

LanceDB provides flexible storage backends that support both cloud object storage and local high-performance storage for different deployment scenarios.

Feature	Description	OSS	Cloud	Enterprise
Object, File, Block Storage	Support for AWS, GCS, Azure and S3-compatible vendors.	✅	✅	✅
Local SSD/NVMe Storage	Support for storage on customer’s custom servers.	✅		✅

Tables

LanceDB’s table abstraction provides ACID-compliant data management with schema evolution, versioning, and consistency guarantees for vector and scalar data.

Feature	Description	OSS	Cloud	Enterprise
Tables - CRUD Operations	Basic API to create, read, update, drop tables.	✅	✅	✅
Tables - Data Evolution	Alter column schema, datatype, backfill + merge data	✅	✅	✅
Tables - Versioning	Append, overwrite, check versions + tag them.	✅	✅	✅
Tables - Consistency	Synchronize database with underlying storage.	✅	✅	✅

Ingestion

LanceDB’s ingestion pipeline handles both vector embedding generation and data loading with support for multiple formats and efficient batch operations.

Feature	Description	OSS	Cloud	Enterprise
Embedding - Text Data	Generate vector embeddings from text data using various embedding models.	✅	✅	✅
Embedding - Multimodal Data	Generate embeddings from images, audio, and other multimodal content.	✅	✅	✅
Embedding - CPU & GPU Device Configuration	Configure CPU or GPU acceleration for embedding generation performance.	✅	✅	✅
Embedding - Environment Variables	Manage API keys and configuration for embedding model access.	✅	✅	✅
Data Ingestion - Default	Formerly called Adding Data to a Table.	✅	✅	✅
Data Ingestion - Formats	Pandas, Polars, Pyarrow, Pydantic	✅	✅	✅
Data Ingestion - Upsert	Update existing records or insert new ones based on key.	✅	✅	✅
Data Ingestion - Merge Insert	Combine data from multiple sources into a single table.	✅	✅	✅

Indexing

LanceDB’s indexing system provides multiple vector and scalar index types with automated optimization for fast similarity search and retrieval operations.

Feature	Description	OSS	Cloud	Enterprise
Vector Index - IVF_FLAT	Minimal index that looks at IVF partitions, instead of brute forcing.	✅	✅	✅
Vector Index - IVF_PQ	Default vector index using Euclidean distance.	✅	✅	✅
Vector Index - IVF_SQ	IVF index built using scalar quantized vectors.	✅	✅	✅
Vector Index - IVF_HNSW_SQ	HNSW built on IVF’s partitions + vectors that are scalar quantized.	✅	✅	✅
Vector Index - Binary	IVF_FLAT with Hamming distance for binary vectors.	✅	✅	✅
Scalar Index	BTREE, BITMAP, LABEL_LIST	✅	✅	✅
Automated Indexing	Indexing happens in the background no config.		✅	✅
Bypass Automated Indexing	When you want to search over all available vectors.	✅	✅	✅
Reindexing - Manual	User needs to specify that they want to reindex.	✅	✅	✅
Reindexing - Automated	Reindexing happens in the background no config		✅	✅
GPU Indexing - Manual	User needs to specify which indexing device to use.	✅	✅	✅
GPU Indexing - Automated	Indexing device is automatically set for user.		✅	✅
Full Text Search Index	Inverted index	✅	✅	✅

Search

LanceDB’s search capabilities combine vector similarity search, full-text search, and hybrid approaches to provide comprehensive retrieval functionality across different data types.

Feature	Description	OSS	Cloud	Enterprise
Vector Search - No Index	Goes through all the available vectors.	✅	✅	✅
Vector Search - ANN Index	Retrieves top K similar vectors.	✅	✅	✅
Vector Search - Multivectors	Late interaction vector search.	✅	✅	✅
Vector Search - Distance Range	Search for vectors within a specific distance threshold.	✅	✅	✅
Vector Search - Binary Vectors	Search using binary vector representations for efficiency.	✅	✅	✅
Vector Search - Filtering	Apply scalar filters during vector search operations.	✅	✅	✅
Vector Search - Batch API	Process multiple search queries in a single request.	✅	✅	✅
Vector Search - Async Indexing	Fallback brute force for fast performance.		✅	✅
Full Text Search - FTS Index	Inverted Index	✅	✅	✅
Full Text Search - Tokenizer	Ngram and other common methods of splitting text data.	✅	✅	✅
Full Text Search - Scalar Index	BTREE, BITMAP, LABEL_LIST for non-vector data.	✅	✅	✅
Full Text Search - Fuzzy Search	Searching when there is a typo on the query.	✅	✅	✅
Full Text Search - Prefix Matching	Search for text that starts with specific characters.	✅	✅	✅
Full Text Search - Score Boosting	Increase relevance scores for specific terms or fields.	✅	✅	✅
Full Text Search - Boolean Logic	Use AND, OR, NOT operators in text search queries.	✅	✅	✅
Full Text Search - Array Fields	Search within array or list data types.	✅	✅	✅
Hybrid Search - FTS Index	Combine vector and full-text search in single query.	✅	✅	✅
Hybrid Search - Reranking	Reorder search results using additional ranking models.	✅	✅	✅
SQL Queries	Execute standard SQL queries on LanceDB tables.	✅	✅	✅
Query Optimization	Explain query plan, analyze query plan, optimization config settings.	✅	✅	✅

Filtering

LanceDB’s filtering system provides flexible query capabilities that can be applied independently or in combination with vector and full-text search operations.

Feature	Description	OSS	Cloud	Enterprise
Filtering - no Vector Search	Apply filters without vector search operations.	✅	✅	✅
Filtering - Vector Search	Apply filters during vector search operations.	✅	✅	✅
Filtering - Full Text Search	Apply filters during full-text search operations.	✅	✅	✅