From 7231185552dc3cada93dbde346b52792ef10a994 Mon Sep 17 00:00:00 2001
From: Garrett Maring <garrett.maring@gmail.com>
Date: Thu, 1 Feb 2024 16:57:17 -0500
Subject: [PATCH] get full node content in KeywordExtractor (#10398)
---
llama_index/extractors/metadata_extractors.py | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/llama_index/extractors/metadata_extractors.py b/llama_index/extractors/metadata_extractors.py
index 6dca49f0c9..bb43f6c7fe 100644
--- a/llama_index/extractors/metadata_extractors.py
+++ b/llama_index/extractors/metadata_extractors.py
@@ -192,13 +192,14 @@ class KeywordExtractor(BaseExtractor):
return {}
# TODO: figure out a good way to allow users to customize keyword template
+ context_str = node.get_content(metadata_mode=self.metadata_mode)
keywords = await self.llm.apredict(
PromptTemplate(
template=f"""\
{{context_str}}. Give {self.keywords} unique keywords for this \
document. Format as comma separated. Keywords: """
),
- context_str=cast(TextNode, node).text,
+ context_str=context_str,
)
return {"excerpt_keywords": keywords.strip()}
--
GitLab