From a9865d0d99e329b652e43ed0237ea0f12db3a7d9 Mon Sep 17 00:00:00 2001 From: Dan Tasse Date: Fri, 3 Apr 2026 10:03:50 -0400 Subject: [PATCH] docs: if you add_column, then you have to backfill instead of refresh --- docs/geneva/udfs/scalar-udtfs.mdx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/geneva/udfs/scalar-udtfs.mdx b/docs/geneva/udfs/scalar-udtfs.mdx index e826a65..d7cf2fe 100644 --- a/docs/geneva/udfs/scalar-udtfs.mdx +++ b/docs/geneva/udfs/scalar-udtfs.mdx @@ -140,7 +140,7 @@ def clip_embedding(clip_bytes: bytes) -> list[float]: clips.add_columns({"embedding": clip_embedding}) # Backfill computes embeddings for all existing clips -clips.refresh() +clips.backfill("embedding") ``` This is a powerful pattern: expand source rows with a scalar UDTF, then enrich the expanded rows with standard UDFs. @@ -217,7 +217,7 @@ def embed_text(chunk_text: str) -> list[float]: return embedding_model.encode(chunk_text) chunks.add_columns({"embedding": embed_text}) -chunks.refresh() # Backfills embeddings on all existing chunks +chunks.backfill("embedding") # Backfills embeddings on all existing chunks # Query — parent columns available alongside chunk columns chunks.search(None).select(["doc_id", "title", "chunk_text", "embedding"]).to_pandas()