AI_EMBED_TEXT_768

Snowflake

Menghasilkan vector embedding 768 dimensi dari teks untuk semantic search.

Tipe hasil: VECTOR(FLOAT, 768)Diperbarui: 6 Jan 2026

Syntax

SQL

SNOWFLAKE.CORTEX.EMBED_TEXT_768(model, text)

Parameter

modelVARCHARwajib

Model embedding: 'e5-base-v2', 'snowflake-arctic-embed-m'

textVARCHARwajib

Teks untuk di-embed

Contoh Penggunaan

Create Embeddings

SQL

1 SELECT
2   document_id,
3   SNOWFLAKE.CORTEX.EMBED_TEXT_768('e5-base-v2', document_text) as embedding
4 FROM documents;

Generate embeddings untuk semantic search.

Semantic Search

SQL

1 WITH query_embedding AS (
2   SELECT SNOWFLAKE.CORTEX.EMBED_TEXT_768('e5-base-v2', 'machine learning tutorial') as vec
3 )
4 SELECT
5   d.title,
6   VECTOR_COSINE_SIMILARITY(d.embedding, q.vec) as similarity
7 FROM documents d, query_embedding q
8 ORDER BY similarity DESC
9 LIMIT 10;

Semantic search menggunakan embeddings.

Kembali ke Snowflake Functions

1	SELECT
2	document_id,
3	SNOWFLAKE.CORTEX.EMBED_TEXT_768('e5-base-v2', document_text) as embedding
4	FROM documents;

1	WITH query_embedding AS (
2	SELECT SNOWFLAKE.CORTEX.EMBED_TEXT_768('e5-base-v2', 'machine learning tutorial') as vec
3	)
4	SELECT
5	d.title,
6	VECTOR_COSINE_SIMILARITY(d.embedding, q.vec) as similarity
7	FROM documents d, query_embedding q
8	ORDER BY similarity DESC
9	LIMIT 10;