Find answers from the community

Updated 4 months ago

Content

Hello!
I'm trying to use llamaindex with an existing qdrant collection, is it possible to do a mapping or to have a custom schema so that llama index understand it? I'm not a fan of the _node_content with json data, I prefer to have a well structured DB

3 comments

LLogan M

The node_content field is just there to reconstruct the text node, since there are many options on it

If you wanted to work with any db, you'd have to overwrite the function that's parsing query results.

https://github.com/run-llama/llama_index/blob/729d5f2e76cca29284157c6fd8b55fb9953739dd/llama-index-integrations/vector_stores/llama-index-vector-stores-qdrant/llama_index/vector_stores/qdrant/base.py#L990

LLogan M

I'm not sure what you mean by "well structured" -- the metadata of the nodes is still pulled out into fields so that you can filter

BBafS

Thank you! I'll check there. By "well structured" I meant that we don't have nested json in _node_content, instead we have key/values directly in the DB.
For example instead of having jsonDecode('_node_content').text we would just have the 'text' key in the DB.

Add a reply