Skip to content

feat: Add Chat Completion API support to SharedPrefixDataGenerator#287

Open
bongwoobak wants to merge 6 commits intokubernetes-sigs:mainfrom
moreh-dev:shared-prefix-chat
Open

feat: Add Chat Completion API support to SharedPrefixDataGenerator#287
bongwoobak wants to merge 6 commits intokubernetes-sigs:mainfrom
moreh-dev:shared-prefix-chat

Conversation

@bongwoobak
Copy link

@bongwoobak bongwoobak commented Nov 19, 2025

Summary

Extended SharedPrefixDataGenerator to support Chat Completion API alongside the existing Completion API.

Changes

  • Add Chat Completion API support to SharedPrefixDataGenerator
  • Shared prefix --> system message, question --> user message
  • Backward compatible with existing Completion API

Modified Files

  • inference_perf/datagen/shared_prefix_datagen.py

@k8s-ci-robot
Copy link
Contributor

k8s-ci-robot commented Nov 19, 2025

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: bongwoobak
Once this PR has been reviewed and has the lgtm label, please assign arangogutierrez for approval. For more information see the Code Review Process.

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

Details Needs approval from an approver in each of these files:

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

@linux-foundation-easycla
Copy link

linux-foundation-easycla bot commented Nov 19, 2025 •
edited
Loading

CLA Signed

The committers listed above are authorized under a signed CLA.

@k8s-ci-robot
Copy link
Contributor

k8s-ci-robot commented Nov 19, 2025

Welcome @bongwoobak!

It looks like this is your first PR to kubernetes-sigs/inference-perf 🎉. 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/inference-perf 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 k8s-ci-robot added cncf-cla: no Indicates the PR's author has not signed the CNCF CLA. size/M Denotes a PR that changes 30-99 lines, ignoring generated files. cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. and removed cncf-cla: no Indicates the PR's author has not signed the CNCF CLA. labels Nov 19, 2025
@SachinVarghese
Copy link
Contributor

SachinVarghese commented Nov 20, 2025

/assign @Bslabe123
Make sure compatibility with multi-turn

@k8s-ci-robot k8s-ci-robot added the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Nov 25, 2025
@bongwoobak
add chat completion in shared prefix datagen
1fe4026
@bongwoobak
feat: Add Chat API + Multi-turn support
0f0012a 
- Implement UserSessionChatAPIData for Chat API with multi-turn chat
- Update SharedPrefixDataGenerator to support Chat API + Multi-turn combination
- Fix prompt selection to use different questions for each turn
- Initialize user session context as list for Chat API compatibility
@k8s-ci-robot k8s-ci-robot added size/L Denotes a PR that changes 100-499 lines, ignoring generated files. and removed needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. size/M Denotes a PR that changes 30-99 lines, ignoring generated files. labels Dec 11, 2025
@bongwoobak
Copy link
Author

bongwoobak commented Dec 11, 2025

Sorry for the delay. I've updated the code to support multi-turn chat properly.

@jjk-g
Copy link
Collaborator

jjk-g commented Dec 18, 2025

should fix #303

Bslabe123
Bslabe123 reviewed Jan 22, 2026
View reviewed changes
inference_perf/apis/user_session.py
class LocalUserSession:
user_session_id: str
context: str
user_session_id: str
Copy link
Contributor

@Bslabe123 Bslabe123 Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Duplicate user_session_id?

Copy link
Author

@bongwoobak bongwoobak Jan 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry for the mistake. Fixed it.

Bslabe123
Bslabe123 reviewed Jan 22, 2026
View reviewed changes
inference_perf/apis/user_session.py Outdated
user_session_id: str
context: str
user_session_id: str
context: Any
Copy link
Contributor

@Bslabe123 Bslabe123 Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it possible to use a more precise type for context?

Copy link
Author

@bongwoobak bongwoobak Jan 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done. Changed from Any to list[ChatMessage] | str

@k8s-ci-robot k8s-ci-robot added size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. and removed size/L Denotes a PR that changes 100-499 lines, ignoring generated files. labels Jan 29, 2026
@bongwoobak
Copy link
Author

bongwoobak commented Jan 29, 2026

Found a bug where group_id was incorrectly calculated after shuffle in multi-turn mode. Fixed the logic.

Changes

Bug Fixes:

  • shared_prefix_datagen.py: Fixed multi-turn group_id tracking bug after shuffle
  • user_session.py: Fixed contexts → context naming

Added:

  • group_id to LocalUserSession
  • questions_by_group, shared_prefixes_by_group for correct question selection
  • 17 new tests in test_shared_prefix_multiturn.py

@jjk-g
Copy link
Collaborator

jjk-g commented Jan 29, 2026

#331 seems to also address shuffle.

Let's revert this PR to the chat completion API support, and follow up the multi-turn shuffle PR with the tests you added in the later commit.

Thanks!

@huaxig
Copy link
Contributor

huaxig commented Jan 29, 2026

#332 - add unit test for shared prefix generator

@k8s-ci-robot k8s-ci-robot added size/M Denotes a PR that changes 30-99 lines, ignoring generated files. and removed size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. labels Feb 4, 2026
@bongwoobak
test: Add unit tests for SharedPrefixDataGenerator
fe6196f 
Also revert shuffle synchronization to match origin/main - only shuffle
prompts list. Shuffle synchronization for prompt_pairs and user_sessions
will be handled separately in PR kubernetes-sigs#331.
@k8s-ci-robot k8s-ci-robot added size/L Denotes a PR that changes 100-499 lines, ignoring generated files. and removed size/M Denotes a PR that changes 30-99 lines, ignoring generated files. labels Feb 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@wangchen615 wangchen615 Awaiting requested review from wangchen615

1 more reviewer

@Bslabe123 Bslabe123 Bslabe123 left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

@Bslabe123 Bslabe123

Labels

cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. size/L Denotes a PR that changes 100-499 lines, ignoring generated files.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

6 participants

@bongwoobak @k8s-ci-robot @SachinVarghese @jjk-g @huaxig @Bslabe123