Skip to main content
LanceDB mounts a table as a read-only filesystem: group-by columns become nested folders, each row is a card plus an optional blob file, and semantic search is the search command.

​
Connection

​
Local / embedded

Point uri at a directory. LanceDB stores the table there, no server needed. 
# .env.development
LANCEDB_URI=/data/fashion.lancedb

​
Object storage

uri also accepts s3://, gs://, and az:// prefixes for tables backed by object storage.

​
Cloud and Enterprise

db:// URIs connect to LanceDB Cloud (api_key + region) or Enterprise (api_key + host_override): 
# .env.development
LANCEDB_URI=db://my-database
LANCEDB_API_KEY=sk_...
LANCEDB_REGION=us-east-1
# Enterprise only
LANCEDB_HOST_OVERRIDE=https://my-database.us-east-1.api.lancedb.com
Search is the search "<query>" <path> command. It returns ranked rows as their canonical <id>.md file paths plus a score, so results compose with cat, wc, and pipes (grep/rg stay lexical). It requires a table built with an embedding function on a source field, so tbl.search(text) auto-embeds the query.

​
Limits

The resource is read-only and bounds how much an agent can pull:
  • search_limit (10): default top-k returned by search.
  • max_rows (1,000): hard ceiling on rows listed per folder.
See the Python and TypeScript resource pages for full config.