Dear All,
Is there a way to vectorize string type by embedding in “AIP_Logic”?
Hi @mitsuki,
Great question!
AIP Logic deals with all the complexity of embedding strings yourself, and let’s you access semantic search capabilities of properties you have embedded in your ontology using the “Semantic search” capability.
Check out the docs here on how to set this up fully end to end:
https://www.palantir.com/docs/foundry/functions/using-palantir-provided-models-to-create-a-semantic-search-workflow/
JG
2 Likes
Currently, you are unable to directly creating embeddings from Strings in Logic directly. As @jgreensmith said, you would first need to add these embeddings into the Ontology as vector properties, which you can then use in Logic’s Semantic Search Board. The Docs he linked would be very helpful for this!
1 Like
@jgreensmith
@nickk
Thank you.
So is there a function that takes a string as an argument and returns a vector?
(The method for converting to a vector must be the same as PipelineBuilder’s “Text to embeddins”)
I want to add the text converted into a vector to Ontology so that I can reuse it in RAG.
Sorry for the delayed response! But currently, there does not exist a first class transform for performing string to vector embedding in Logic. The way you would do this is through Pipeline Builder right now or OMS. After you get the embeddings into the vector object property, you should be able to use to the Semantic Search board for your RAG workflow.
We are tracking this though. Would not expect to see this until sometime early next year.
2 Likes
Hi, nickk!
It looks like the solution is to install the model when creating a TypeScript function.
Thanks.
2 Likes