Add code for Envoy extension that supports body-to-header translation

[rramkumar1]

Ref: #321

[k8s-ci-robot]

Welcome @rramkumar1!

It looks like this is your first PR to kubernetes-sigs/gateway-api-inference-extension 🎉. Please refer to our pull request process documentation to help your PR have a smooth ride to approval.

You will be prompted by a bot to use commands during the review process. Do not be afraid to follow the prompts! It is okay to experiment. Here is the bot commands documentation.

You can also check if kubernetes-sigs/gateway-api-inference-extension has its own contribution guidelines.

You may want to refer to our testing guide if you run into trouble with your tests not passing.

If you are having difficulty getting your pull request seen, please follow the recommended escalation practices. Also, for tips and tricks in the contribution process you may want to read the Kubernetes contributor cheat sheet. We want to make sure your contribution gets all the attention it needs!

Thank you, and welcome to Kubernetes. 😃

[k8s-ci-robot]

Hi @rramkumar1. Thanks for your PR.

I'm waiting for a kubernetes-sigs member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[netlify]

✅ Deploy Preview for gateway-api-inference-extension ready!

Name	Link
🔨 Latest commit	5e9514a
🔍 Latest deploy log	https://app.netlify.com/sites/gateway-api-inference-extension/deploys/67c71e41864a740008d7b9e8
😎 Deploy Preview	https://deploy-preview-355--gateway-api-inference-extension.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[ahg-g]

/ok-to-test

[hzxuzhonghu]

I am not sure how it is related with this project. I mean you can deploy this not depending on gateway inference extension

[rramkumar1]

I am not sure how it is related with this project. I mean you can deploy this not depending on gateway inference extension

@hzxuzhonghu Sorry I forgot to link the related issue in the initial comment. Fixed that.

[hzxuzhonghu]

Thanks, now i get the inetention. But i am wondering, should we separate it from the ext-proc binart? IMO, it can be merged into the current binary. Correct me if miss some ctx

[rramkumar1]

Thanks, now i get the inetention. But i am wondering, should we separate it from the ext-proc binart? IMO, it can be merged into the current binary. Correct me if miss some ctx

I think it should be a separate binary because this particular extension is intended to execute before the routing decision while endpoint picker is intended to execute after routing decision.

[hzxuzhonghu]

hmm, the model name header will be used by the http route match later. What confused me most is the epp currenlty only support one model pool. It seems no model match needed to specify in the HTTPRoute

[robscott]

hmm, the model name header will be used by the http route match later. What confused me most is the epp currenlty only support one model pool. It seems no model match needed to specify in the HTTPRoute

@hzxuzhonghu for a bit more context, there are different points in time where we can attach an ext_proc extension. We want to attach this extension early in the process so it can add headers before routing logic is computed. That would allow someone to say that requests for the "foo" model should go to the "bar" InferencePool while requests for "baz" model should go to a different InferencePool.

Then when the request gets to an InferencePool, the Endpoint Picker extension can select the best endpoint to serve a request for that model within that InferencePool.

[hzxuzhonghu]

Ah, that makes sense. Would we plan to support multiple inferencepool within a epp process

[robscott]

Ah, that makes sense. Would we plan to support multiple inferencepool within a epp process

That's definitely been a point of discussion, I think it's inevitable that that will be a mode at some point, but it's not part of the initial releases. The rationale for this is:

1. We're currently relying on very frequent probing of every endpoint an EPP is supporting, this would not scale well to a large number of endpoints
2. We're relying on keeping some state in memory, this would also not scale to a large number of endpoints
3. Having a 1:1 mapping between InferencePool and EPP allows you to effectively traffic split between different EPPs or InferencePools

With all that said, I think these could all be temporary limitations.

[robscott]

Thanks @rramkumar1! Once we get the image pipeline set up and some corresponding docs, this will be a great part of the v0.2 release.

[ahg-g]

Can you also please add a README.md file; I added one for pkg/epp at #386

[rramkumar1]

What is the plan for tests? unit and integration.

I added a unit test for the code in request.go. Please let me know if any other code should have unit test coverage as some of it is boilerplatey.

For integration tests, should I file a follow-up issue to track? Those seem more involved.

[ahg-g]

What is the plan for tests? unit and integration.

I added a unit test for the code in request.go. Please let me know if any other code should have unit test coverage as some of it is boilerplatey.

For integration tests, should I file a follow-up issue to track? Those seem more involved.

Thanks for adding the unit tests, yes please, lets open a tracking issue for integration test coverage.

[rramkumar1]

What is the plan for tests? unit and integration.

I added a unit test for the code in request.go. Please let me know if any other code should have unit test coverage as some of it is boilerplatey.
For integration tests, should I file a follow-up issue to track? Those seem more involved.

Thanks for adding the unit tests, yes please, lets open a tracking issue for integration test coverage.

#443

[ahg-g]

Almost there! Can you please confirm that this passes manual testing.

As a followup, we need:

1. Manifests for configuring it with envoy gateway
2. User guide

[rramkumar1]

Almost there! Can you please confirm that this passes manual testing.

As a followup, we need:

1. Manifests for configuring it with envoy gateway
2. User guide

Yes, this passes manual testing!

Will file issues for those follow-ups.

[ahg-g]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ahg-g, rramkumar1

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [ahg-g]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment