Issues

• SGLang MHA/GQA small KV pools get zero backing pages and stall initialization

  ovg-project/kvcached#592 · Filter by author JasperChan929 opened on Oct 20, 2026

• PageAllocator background capacity queries use the wrong GPU

  ovg-project/kvcached#588 · Filter by author JasperChan929 opened on Oct 16, 2026

• System Design Discussion for Scheduling Policy + CPU Offloading

  ovg-project/kvcached#585 · Filter by author RixinLiu opened on Oct 16, 2026
  5

• Feature request: elastic backing for DSA caches in vLLM and SGLang (DeepSeek-V3.2, GLM-5)

  feature requested
  ovg-project/kvcached#560 · Filter by author shipiyouniao opened on Oct 13, 2026

• KVCached Weekly GPU CI Test Results

  ovg-project/kvcached#549 · Filter by author RixinLiu opened on Oct 8, 2026
  7

• SGLang: hybrid models crash instead of queueing when GPU memory runs short

  bug
  ovg-project/kvcached#547 · Filter by author RixinLiu opened on Oct 8, 2026
  1

• [Discussion] Support for Moore Threads MUSA and additional GPU backends

  ovg-project/kvcached#536 · Filter by author googs1025 opened on Oct 3, 2026
  3

• MLA models on vLLM 0.24: KVCacheManager post-init cannot reach worker IPC socket (Errno 2), alloc_page retry storm

  ovg-project/kvcached#471 · Filter by author rishabhsinha17 opened on Aug 27, 2026

• Design: safe cross-process admission for concurrent physical KV growth

  ovg-project/kvcached#467 · Filter by author shipiyouniao opened on Aug 27, 2026
  10

• PD Disaggregation Support Tracking Issue

  ovg-project/kvcached#442 · Filter by author RixinLiu opened on Aug 11, 2026

• kvcached support matrix

  help wanted
  ovg-project/kvcached#425 · Filter by author RixinLiu opened on Aug 5, 2026
  11

• KVCached Roadmap - Release v0.1.7

  ovg-project/kvcached#419 · Filter by author RixinLiu opened on Jul 30, 2026
  10

• page-size calculation incompatible with vLLM 0.24+ AttentionSpec.indexes_kv_by_block_stride

  bug
  ovg-project/kvcached#399 · Filter by author lianghao208 opened on Jul 21, 2026

• SGLang paged allocator free path should move indices to CPU before torch.unique

  ovg-project/kvcached#382 · Filter by author shipiyouniao opened on Jul 5, 2026
  3

• Discussion: stable extension points for production kvcached integrations

  feature requested
  ovg-project/kvcached#375 · Filter by author shipiyouniao opened on Jul 1, 2026
  19

• Feature request: support DeepSeek-V4 KV cache pooling in vLLM and SGLang

  feature requested
  ovg-project/kvcached#369 · Filter by author shipiyouniao opened on Jun 30, 2026
  6

• Add external TriAttention integration

  enhancement
  ovg-project/kvcached#355 · Filter by author iMmAseu opened on Jun 8, 2026

• Proposal: Integrate kvcached with Ollama

  ovg-project/kvcached#326 · Filter by author ztang2370 opened on May 8, 2026

• Encounter illegal memory access in FlashInfer backend under contiguous layout

  bug
  ovg-project/kvcached#321 · Filter by author lianghao208 opened on May 1, 2026
  4

• The kvcached environment variable set in the python script does not take effect

  ovg-project/kvcached#316 · Filter by author Dyouvrean opened on Apr 28, 2026
  12

• Does KVCached support PD disaggregation?

  question
  ovg-project/kvcached#302 · Filter by author wdx727 opened on Apr 14, 2026
  1

• Performance Optimization Track Issue

  ovg-project/kvcached#299 · Filter by author lianghao208 opened on Apr 8, 2026
  3

• Error when vllm-v0.16.0 uses --kv-offloading-size

  ovg-project/kvcached#267 · Filter by author cui36 opened on Mar 5, 2026
  1

• TimeoutError during startup with Pipeline Parallelism (PP) on Qwen3-8B

  bug
  ovg-project/kvcached#245 · Filter by author iMmAseu opened on Feb 21, 2026

• How can we reproduce the fact that using kvcached is 2-28 times faster than not using kvcached in TTFT?

  question
  ovg-project/kvcached#233 · Filter by author yugenlgy opened on Jan 12, 2026
  4