owen-elliott

4 ms·

1. 1.▲
   by owen-elliott 2y ago

   The composability of RRF is definitely one of its most appealing characteristics. It doesn't matter what algorithm or vendor you have, you can just fuse with ranks alone. I've seen it shine when fusing lexical and vector search re
2. 2.▲
   by owen-elliott 3y ago

   You could use Marqo, it is a vector search engine that includes the text chunking, inference for calculating embeddings, vector storage, and vector search. You can pick from a heap of open-source models or bring your own fine-tuned ones. It
3. 3.▲
   by owen-elliott 3y ago

   In this example the context is the images provided via URLs. For example, the search for backpacks is contextualised by a picture of a forest which updates the results to be hiking/outdoors backpacks.