I have a custom Qdrant Vector Store with 3 vectors defined on each document.
none of them are named “embeddings”.
The current Qdrant Vector Store Node allows me to specify the collection name and optionally a filter in the Get Many and Retrieve Documents options. but what is missing is the ability to specify which vector field to use when searching
1 option has a prompt field for a sentence of set of words i think, the other option has no input field.
Another thing is what if i already have the embedding that i want to use to search against a named field. I cant do that using that tool.
I had to fall back to using the Http Request and do all of the work.
Just wanted to chime in here to say an alternative to the HTTP request node approach would be to use the Langchain Code node. This would give you access to the Qdrant SDK which would allow you to use your named vectors and also the Scroll API to solve your search by field problem.
Downside is this only works in community edition (self-hosted) version of n8n and not cloud.
Thank you for the suggested alternative code, this is indeed another way to tackle the problem. I must admit I not familar with the langchain coding approach, however its not a problem to go and learn it. the more options the better.