Find answers from the community

Updated 4 months ago

Please advise

At a glance

The post is a simple request for advice. The comments discuss how the system works, specifically regarding the generation of titles and the inclusion of metadata in the embedding model. Community members provide explanations and suggest building a custom solution if the default behavior is not desired. There is no explicitly marked answer in the provided information.

SSridhar

Please advise

5 comments

LLogan M

Thats how it works actually -- it takes the first X number of nodes (3 in this case) generates a title, then generates a common title for all nodes based on those

LLogan M

If you need something different, it's very easy to build your own 👍

SSridhar

on this context

document.excluded_embed_metadata_keys = ["creation_date", "last_modified_date", "file_size", "file_path", "file_extension"]

still these attributes are present in metadata.
how can i skip this

LLogan M

they will be included in the metadata yes.

This setting just means that they won't be included when you send nodes to the embedding model

LLogan M

Plain Text

from llama_index.schema import MetadataMode

print(document.get_content(metadata_mode=MetadataMode.EMBED))

That is what the embedding model will see

Add a reply