Has anyone faced the problem when using

At a glance

The community member is facing an issue with the SentenceWindowNodeParser for document indexing. When they set the window_size to 6 (the default is 3), the LLM responses using that index are very long and unrelated to general questions like "Hey, how are you today?" or "Who is Elon Musk". Instead, the responses contain the content and window of the top nodes.

In the comments, another community member suggests that the window size might be too big, causing the LLM to ignore the instructions in the prompt. However, there is no explicitly marked answer to the issue.

bbenzen

Has anyone faced the problem when using the SentenceWindowNodeParser for document indexing that when setting a window_size of like 6 (default is 3), the LLM responses using that index is very long and unrelated to general questions like "Hey, how are you today?" or "Who is Elon Musk" and responds with the content & window of the top nodes?

Plain Text

  node_parser = SentenceWindowNodeParser.from_defaults(
                window_size=3,
                window_metadata_key="window",
                original_text_metadata_key="original_text",
            )

1 comment

LLogan M

Probably the window is too big, causing the LLM it ignore the instructions in the prompt? Just a guess anyways 🤔

Add a reply

Find answers from the community

Has anyone faced the problem when using