Love @langchain 🦜⛓️? Love DocArray? Langchain just launched their new integrations hub and DocArray is right there in the mix - twice! The best part: You can even give us a ❤️ integrations.langchain.com/v…

⚡️Deserialize DocVec directly to numpy, torch, or tf When you deserialize a DocVec from Protobuf, you can now choose if its tensors should be loaded to numpy, pytorch, or tensorflow:

💡 Easily check if a Document is already in a DocumentIndex Just use the `in` keyword to see if a Document has already been indexed.

🔍 Nested search for InMemoryExactNNIndex with subindexes You can now use find_subindex() to search through Documents that are in a DocList inside of another Document!

The Tensor coercion feature is a must-have and allows you to seamlessly work with different tensor backends (numpy, @PyTorch , @TensorFlow ) at the same time, thanks to @pydantic validation.

Last but not least, save that precious memory by using quantization, allowing vector search even on puny low-level machines

Want to move from your old vector database to Qdrant? Just swap out a line or two of code

Next, Qdrant can incorporate filtering directly into the vector search phase. No need for that pre- or post-filtering!

First up, with DocArray + Qdrant you can experiment locally first, without needing to run a server. You can move to on-prem or cloud when you're ready

DocArray's @qdrant_engine support keeps getting better and better. Let's dive into the benefits in this thread!