aibrix icon indicating copy to clipboard operation
aibrix copied to clipboard

[Feature/Integration] Add Integration Path with vLLM Semantic Router

Open sadath-12 opened this issue 3 months ago • 9 comments

🚀 Feature Description and Motivation

can semantic router ( https://github.com/vllm-project/semantic-router ) be integrated into aibrix's routing component ?

Use Case

This can give more intelligence to aibrix's routing

Proposed Solution

No response

sadath-12 avatar Sep 29 '25 13:09 sadath-12

I am very happy to see this issue track, and I am looking forward to this feature being integrated into aibrix. 😄

googs1025 avatar Sep 29 '25 15:09 googs1025

@googs1025 can I pick this one and see where it goes ?

sadath-12 avatar Sep 29 '25 16:09 sadath-12

Regarding routers, Aibrix currently uses Envoy Gateway for integration. The source code is here: https://github.com/vllm-project/aibrix/tree/main/pkg/plugins/gateway

If you have any ideas, please provide docs or share them in this issues. Other maintainers will actively provide answers and responses! cc @Jeffwan @varungup90

googs1025 avatar Sep 30 '25 06:09 googs1025

cool , let me take it from there @googs1025

sadath-12 avatar Oct 03 '25 12:10 sadath-12

/cc @Xunzhuo please also help evaluate it

Jeffwan avatar Oct 03 '25 16:10 Jeffwan

@Jeffwan Yes, I think this should be easy to integrate.

Xunzhuo avatar Oct 09 '25 03:10 Xunzhuo

Some work-in-progress integration:

  • Dynamo: https://vllm-semantic-router.com/docs/proposals/nvidia-dynamo-integration
  • PS: https://github.com/vllm-project/semantic-router/issues/295

If AIBrix is interested in, that would be cool, then I can help with the design and move this forward

Xunzhuo avatar Oct 09 '25 03:10 Xunzhuo

Better to unassign me from this now to unblock this issue , I got some other stuffs in middle and not sure how much time ill need to come back to this

sadath-12 avatar Oct 12 '25 09:10 sadath-12

@sadath-12 sounds good. We will ask @Xunzhuo 's help on this issue

Jeffwan avatar Oct 13 '25 03:10 Jeffwan