search

milvus-io / milvus

A cloud-native vector database, storage for next generation AI applications
https://milvus.io
Apache License 2.0
30.09k stars 2.89k forks source link

[Feature]: Support multi vector（array of embeddings） #37090

Open xiaofan-luan opened 2 hours ago

xiaofan-luan commented 2 hours ago

Is there an existing issue for this?

Is your feature request related to a problem? Please describe.

Multi vector or vector list is widely used in the following use case.

  1. when we chunk doc into multi chunks, each chunks can be a vector in the list. And abstract/time can be another vector. so the data model looks like image

There is usually two operations on this array

  1. doing ann search on each array
  2. doing maxsim on array

Describe the solution you'd like.

No response

Describe an alternate solution.

No response

Anything else? (Additional Context)

No response

xiaofan-luan commented 2 hours ago

Is there an existing issue for this?

  • [x] I have searched the existing issues

Is your feature request related to a problem? Please describe.

Multi vector or vector list is widely used in the following use case.

  1. when we chunk doc into multi chunks, each chunks can be a vector in the list. And abstract/time can be another vector. so the data model looks like
image

There is usually two operations on this array

  1. doing ann search on each array
  2. doing maxsim on array

with those two abilities, colbert can be also supported

Describe the solution you'd like.

No response

Describe an alternate solution.

No response

Anything else? (Additional Context)

No response