feat(ruvllm): add TwinKV composable KV redundancy repair - #946
Draft
ruvnet wants to merge 6 commits into
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implements the bounded TwinKV repair primitive tracked in #945.
Changes include a policy agnostic repair pass under ruvLLM optimization, exact cache budget preservation, protected sink and recent regions, local similarity exclusion, deterministic orphan and donor ranking, malformed input rejection, unit tests, ADR 330, and a full reproduction protocol.
The implementation deliberately computes only against the retained set for O(n K d) repair specific work rather than constructing the full pairwise matrix used by the paper implementation.
This PR is draft because model level benchmark reproduction is not complete. Do not merge until matched baseline versus candidate tests demonstrate a workload specific gain, short context controls remain stable, and repair latency does not erase the cache benefit.