Hi Paimon Devs,

I'd like to start a discussion about the Vector data type in Paimon[1].

Vector data is commonly used in AI workloads to encode multimodal inputs.
The most basic format is a fixed-length, densely packed float array, a
representation adopted by nearly all dedicated vector DBs. Meanwhile,
typical usage patterns have emerged around the storage and retrieval of
data in vector scenarios, making them diverge substantially from
general-purpose storage models.

Given this context, I believe that defining a new Vector data type, along
with a point‑lookup optimized storage model, will help Paimon expand its
support in the vector data field and serve AI workflows more effectively.

What do you think about this proposal?

[1]
https://cwiki.apache.org/confluence/display/PAIMON/PIP-40%3A+Introduce+a+new+Vector+data+type

Best,
Han Liu

Reply via email to