---
name: "lancedb-serverless-arrow-vectors"
description: "Execute serverless, disk-based vector search directly from S3 or local disk using LanceDB."
metadata:
  author: "LanceDB"
  version: "1.0.0"
  runtime_compatibility:
    - "codex"
    - "claude-code"
    - "cursor"
    - "windsurf"
triggers:
  - "query lancedb arrow table"
  - "create lancedb ivf pq index"
  - "serverless vector search lancedb"
permissions:
  fs_read: true
  fs_write: "workspace"
  shell_exec: false
  net_access: false
---

# LanceDB Serverless Vector Database on Apache Arrow

Embeds LanceDB in modern applications: zero-copy Apache Arrow integration, disk-based vector indexing with IVF-PQ, multimodal vector storage, and serverless querying without managing persistent servers.

## When to Invoke
Invoke this skill whenever you need to:
- query lancedb arrow table
- create lancedb ivf pq index
- serverless vector search lancedb

## Execution SOP & Rules
1. Initialize connection: db = lancedb.connect('s3://my-bucket/vectors') or local path.
2. Create table from Pandas/Arrow: tbl = db.create_table('embeddings', data=df).
3. Build index: tbl.create_index(metric='cosine', num_partitions=256, num_sub_vectors=96).
4. Search vectors: tbl.search(query_vec).limit(10).to_pandas().

## Verification Checklist
- Run local linting, syntax validation, and type-checks before reporting completion.
- Ensure all created or modified files adhere to workspace boundaries and avoid uncommitted secrets.
- Verify tests or build commands exit cleanly with return code 0.