Implementation:CrewAIInc CrewAI LanceDB Adapter

Knowledge Sources	CrewAI
Domains	Vector_Search, RAG, Embeddings
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete adapter for LanceDB vector database integration provided by CrewAI.

Description

The LanceDBAdapter class implements the Adapter interface from CrewAI's RAG tool framework, providing seamless integration with LanceDB, a fast embedded vector database. It connects to a LanceDB instance using a configurable URI (file path or connection string), opens a specified table, and exposes methods for semantic search (query) and data ingestion (add). The adapter uses a configurable embedding function that defaults to OpenAI's text-embedding-ada-002 model to convert text queries into vector representations for similarity search. Results are retrieved via LanceDB's vector search with configurable top_k limit and returned as newline-joined text strings.

Usage

Import and instantiate LanceDBAdapter when you need to connect CrewAI's RAG tool pipeline to a LanceDB vector database for semantic retrieval. This is particularly suited for local or embedded deployments where a lightweight vector store is preferred over cloud-hosted alternatives.

Code Reference

Source Location

• Repository: CrewAI
• File: lib/crewai-tools/src/crewai_tools/adapters/lancedb_adapter.py
• Lines: 1-59

Signature

class LanceDBAdapter(Adapter):
    uri: str | Path
    table_name: str
    embedding_function: Callable = Field(default_factory=_default_embedding_function)
    top_k: int = 3
    vector_column_name: str = "vector"
    text_column_name: str = "text"

Import

from crewai_tools.adapters.lancedb_adapter import LanceDBAdapter

I/O Contract

Inputs

Name	Type	Required	Description
uri	Path	Yes	LanceDB connection URI (file path or connection string)
table_name	str	Yes	Name of the LanceDB table to open
embedding_function	Callable	No	Function to generate embeddings; defaults to OpenAI text-embedding-ada-002
top_k	int	No	Number of top results to return from vector search (default: 3)
vector_column_name	str	No	Name of the vector column in the table (default: "vector")
text_column_name	str	No	Name of the text column in the table (default: "text")

Outputs

Name	Type	Description
query() return	str	Newline-joined text values from the top_k most similar results
add() return	None	Adds data directly to the underlying LanceDB table (pass-through)

Usage Examples

Basic Usage

from crewai_tools.adapters.lancedb_adapter import LanceDBAdapter

# Connect to a local LanceDB instance
adapter = LanceDBAdapter(
    uri="/path/to/lancedb",
    table_name="my_documents",
    top_k=5,
)

# Query for semantically similar content
results = adapter.query("What is the company refund policy?")
print(results)

Custom Embedding Function

from crewai_tools.adapters.lancedb_adapter import LanceDBAdapter

def custom_embeddings(texts):
    # Custom embedding logic
    return [[0.1, 0.2, ...] for _ in texts]

adapter = LanceDBAdapter(
    uri="/path/to/lancedb",
    table_name="my_documents",
    embedding_function=custom_embeddings,
    vector_column_name="embeddings",
    text_column_name="content",
)

Related Pages

• Principle:CrewAIInc_CrewAI_Built_In_Tool_Selection