Critical Patterns

Table Creation (REQUIRED)

import lancedb

✅ ALWAYS: Define schema clearly

db = lancedb.connect("./my_db")

data = [ {"id": 1, "text": "Hello world", "vector": [0.1, 0.2, ...]}, {"id": 2, "text": "Goodbye world", "vector": [0.3, 0.4, ...]}, ]

table = db.create_table("my_table", data)

Vector Search (REQUIRED)

✅ Search by vector similarity

results = table.search([0.1, 0.2, ...]).limit(10).to_list()

✅ With filter

results = table.search(query_vector)
.where("category = 'tech'")
.limit(5)
.to_list()

Decision Tree

Need semantic search? → Use vector search Need exact match? → Use where clause Need hybrid search? → Combine vector + filter Need persistence? → Use file-based connection

Resources

• Best Practices: best-practices.md