LMCache

Benchmark a hybrid-attention model across the LMCache performance ladder (vLLM no-hybrid-allocator → hybrid allocator + prefix caching → hybrid allocator + LMCache) and produce a decode-throughput / TTFT / cache-hit-rate comparison. Use when asked to…

Review a GitHub pull request against the LMCache coding standards

Create a GitHub pull request from the current branch to the upstream repo, using the repo's PR template

Check local changes against the LMCache coding standards and fix issues before creating a PR

Upgrade LMCache-Ascend to be compatible with a new upstream LMCache version. This skill takes the current LMCache-Ascend monkey patches, re-derives them by comparing against the old LMCache version, and applies them to the new LMCache version's structure. Use…