v0.3.2.post1

What's Changed

• Fix(#374)(master): Pass eviction ration flag to the master service by @maobaolong in #405
• [TransferEngine]: Add a new configuration option of log path by @SCDESPERTATE in #399
• [TransferEngine] Avoid query segment desc too often by @alogfans in #404
• chore: bump version to 0.3.2.post1 in pyproject.toml by @ShangmingCai in #409

New Contributors

• @SCDESPERTATE made their first contribution in #399

Full Changelog: v0.3.2...v0.3.2.post1

v0.3.2

Highlights

• TE supports fault tolerency
• Store: supports eviction and lease

What's Changed

• chore(deps): bump golang.org/x/net from 0.36.0 to 0.38.0 in /mooncake-p2p-store/src/p2pstore by @dependabot in #376
• [Build]: exclude cuda so files in auditwheel. by @doujiang24 in #379
• [TransferEngine] Remove unused local variable by @jellor in #382
• [Doc] Add document for integrating Mooncake Store to LMCache V1 by @XucSh in #385
• [Store] Add features: lease and eviction by @ykwd in #374
• fix bug in async apis in python, the batch_id's dtype is int64_t, not int by @niqi-lyu in #387
• [Bugfix] Fix PID retrieval when destroying the vLLM thread in the vllm benchmark demo by @0x777a6c in #389
• [TE FIX] Updating the outstanding work request counting when closing a QP by @alogfans in #390
• [TransferEngine] Fix hang problem due to previous failed connection by @alogfans in #392
• Remove duplicated code between transferSync[Read|Write] and transferSync by @alogfans in #394
• [FIX] avoid locking for the same spinlock multiple times by @alogfans in #395
• [Build] do not hard code release build type. by @doujiang24 in #393
• [TransferEngine] add lock for handle_map_ to avoid segfault. by @doujiang24 in #396
• [TransferEngine] Update software-based timeout mechanism by @alogfans in #398
• [TransferEngine] Revert to disable slice timeout by @alogfans in #401
• chore: bump version to 0.3.2 in pyproject.toml by @ShangmingCai in #402

New Contributors

• @jellor made their first contribution in #382
• @niqi-lyu made their first contribution in #387
• @0x777a6c made their first contribution in #389

Full Changelog: v0.3.1...v0.3.2

v0.3.1

Highlights

• Performance: Optimized local data transfer via memcpy, Enhanced buffer allocation logic and path selection strategy.
• CI/CD: more build scripts, Docker support for master server and CI testing.
• Observability: Improved error logging, metrics and null pointer checks
• Create mooncake website
• Bug fixes: GCC10 build fixes, Dependency and RDMA transport fixes.

What's Changed

• docs: add lmcache integration documentation by @xiaguan in #336
• [Store] Improve: Add RemoveAll rpc for remove all keys by @maobaolong in #327
• [fix]fix compile error for gcc10 by @qicosmos in #339
• feat(Mooncake Integration): Support pure client without store mode by @maobaolong in #341
• ARM build_wheel.sh by @johnnynunez in #344
• [DOC] Add news about NIXL supports Mooncake as a backend by @alogfans in #348
• Fix news render in README by @alogfans in #350
• chore: enhance error logging for tcp transport by @xiaguan in #349
• add arm compatibility by @johnnynunez in #343
• gitmodules: use full path instead of relative path. by @doujiang24 in #345
• feat(transfer-engine): enhance logging for RPC and topology discovery by @xiaguan in #352
• chore(ci): expand python version testing matrix to include 3.8-3.13 by @xiaguan in #353
• feat(Mooncake Integration): Supply a MooncakeConfig into whl file by @maobaolong in #338
• feat(docs): build documentation website for Mooncake using Sphinx by @Risc-lt in #354
• [Build] feature: start the Mooncake master server through Docker by @Chasing1020 in #320
• fix wrong command in XpYd by @gujingit in #357
• fix(master): Fix negative storage size metrics after removeAll by @maobaolong in #355
• [CI] add dockerfile CI test by @stmatengss in #362
• [FIX] avoid sending request after setting inactive by @alogfans in #361
• [Transfer Engine] Check the buffer size before register by @XucSh in #364
• fix: incorrect urls and update page deployment by @Risc-lt in #363
• [Build] Call find_package() before using external deps by @tchaikov in #359
• [BugFix] Buffer Allocation Always Tries on the Same Allocator by @ykwd in #365
• [TransferEngine] Add sanity check on nullptrs by @alogfans in #366
• fix(store): the metadata leak after umount segment by @maobaolong in #337
• store: optimize local data transfer with memcpy fast path by @xiaguan in #340
• [DOC] update blog url by @stmatengss in #369
• build(transport): add glog and pthread to rdma target link libraries by @xiaguan in #370
• Add LRU in MasterService, complexity O(1) by @zhaoyongke in #287
• [Store] feat: Add a MooncakeStoreService to serve store and rest api by @maobaolong in #328
• [hotfix] Allow compile flag USE_LRU_MASTER to enable/disable the LRU feature by @ykwd in #372
• [FIX] update path reselection stragegy to cover all possible available devices by @alogfans in #373
• feat(py): integrate python http metadata server by @xiaguan in #367
• chore: bump version to 0.3.1 in pyproject.toml by @xiaguan in #377

New Contributors

• @qicosmos made their first contribution in #339
• @johnnynunez made their first contribution in #344
• @Risc-lt made their first contribution in #354
• @XucSh made their first contribution in #364
• @tchaikov made their first contribution in #359
• @ykwd made their first contribution in #365

Full Changelog: v0.3.0...v0.3.1

v0.3.0

Highlights

• Mooncake Store: Develop a new Distributed Object Store for XpYd PD disaggregation
• Support vLLM V0, see: vllm-project/vllm#12957
• Support vLLM V1 based on LMCache, see: https://blog.lmcache.ai/2025-04-22-tencent/
• Support SGLang, see: https://lmsys.org/blog/2025-05-05-large-scale-ep/
• Transfer Engine: (1) Handshaking-based connection; (2) Automatically generate Topology Matrix; (3) Asynchronous transfer mode;
• Provide better installation methods (pip install mooncake-transfer-engine).
• ....

What's Changed

• fix: when detect async event, stop the outstanding requests & prevent cq burn out leading severe error by @alogfans in #84
• [TransferEngine] fix: add auto_discover in transfer_engine_c by @chestnut-Q in #87
• [TransferEngine] feature: registerLocalMemory support the "*" location. by @doujiang24 in #86
• [Build] feature: add dockerfile to enable use Mooncake in Docker by @stmatengss in #92
• [Transfer Engine] add USE_ETCD option to retire the etcd dependency. by @doujiang24 in #94
• [TransferEngine]chore: replace slices vector with slice counter by @RuixiangMa in #93
• [Build] feature: add dev container. by @doujiang24 in #96
• [TransferEngine] fix: wrong numa number in example. by @doujiang24 in #99
• [TransferEngine] fix: use ibv_get_device_list to get the IB devices. by @doujiang24 in #102
• fix: memcpy direction wrong for read operations by @alogfans in #101
• [TransferEngine] Build: build/install shared object for transfer engine. by @doujiang24 in #100
• Fix P2P store bugs and improvement by @alogfans in #80
• [Bugfix] Add shared and unique lock for local_memory_regions_ (#107) by @power-more in #108
• [TransferEngine] Correct the Typo by @power-more in #109
• [TransferEngine] Fix typos by @fengquyoumo in #110
• [TransferEngine] BenchMark: Introduce report gflags for enhanced configurability by @fengquyoumo in #111
• [TransferEngine] feature: introduce USE_NVMEOF to enable NVMe-oF separately. by @doujiang24 in #106
• [Doc] add FAST25 paper and traces by @chestnut-Q in #116
• [TransferEngine] Suppress gtest type cast compile warning and fix some trivial bugs by @yuan-luo in #115
• [Doc] update README.md by @chestnut-Q in #128
• [Bugfix] fix the gid choice of IB device: only choose IBV_GID_TYPE_ROCE_V2 now by @fengquyoumo in #113
• [MooncakeStore] Introduce MooncakeStore: A new Distributed Object Store for XpYd PD disaggregation by @alogfans in #126
• [FIX] Fix the content of README.md by @alogfans in #130
• [DOC] add description about Docker-based deployment by @alogfans in #133
• Fix compilation error with clang toolchain by @xiaguan in #135
• [MooncakeStore] Add getSize method to DistributedObjectStore by @xiaguan in #137
• [DOC] remove useless metadata_type flags in example.md by @cxz66666 in #138
• [FIX] golang vulnerabilities in P2P store by @alogfans in #140
• [TransferEngine] Support Status return value by @yuan-luo in #125
• [Doc] Fix typo in mooncake-store-preview by @simpx in #142
• [MooncakeStore] Enhance performance in DistributedObjectStore::get by @xiaguan in #141
• [Doc] Add vllm integration v1 doc by @ShangmingCai in #129
• [Optimize] Update EndpointStore interface to use const references for… by @xiaguan in #152
• fix compilation errors by @xhyf77 in #145
• [MooncakeStore] fix: treat object_not_found/already_exists as info in glog by @stmatengss in #156
• [Fix]: Handle SIGINT & SIGTERM for graceful shutdown by @xhyf77 in #153
• [Doc] typo error in README by @Unprincess17 in #158
• [FEATURE] catch exceptions during allocator creation by @xiaguan in #157
• [FEATURE] Update MooncakeStroe CI workflow by @xiaguan in #146
• [DOC] remove MULAN license to ease vllm deployment by @stmatengss in #163
• [Refactor] Replace rpc logic with MasterClient in Client class by @xiaguan in #164
• [CI] Add spell check process to CI by @xiaguan in #165
• [Chore] Disable garbage collection in Mooncake master by @xiaguan in #167
• [TransferEngine] fix typos for cancel status in transfer engine by @alogfans in #169
• Extend VLLMAdaptor API by @xhyf77 in #173
• [MooncakeAdaptor] Init sglang_adaptor to support SGLang using transfer engine by @stmatengss in #181
• [HOTFIX] add env MC_DISABLE_METACACHE to force pull metadata from etcd by @alogfans in #175
• [TransferEngine] Fix compilation bugs of nvmeof transport by @alogfans in #174
• Package mooncake_vllm_adaptor and MooncakeDistributedStore into a wheel by @xhyf77 in #177
• [Fix] Fix issues with run path and etcd dependency integration by @xhyf77 in #182
• [WorkFlow] Add a release.yaml to publish the .whl file by @xhyf77 in #186
• [Chore] Change logging level from INFO to VLOG for object existence by @xiaguan in #185
• [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc by @xiaguan in #179
• [Doc] Update vllm MooncakeStore integration doc by @ShangmingCai in #191
• [CI] Refine release workflow to enable publishing to PyPI by @xiaguan in #189
• [Enhance] Add an RPATH that takes precedence over LD_LIBRARY_PATH and other paths in the wheel package by @xhyf77 in #190
• [MooncakeStore] fix: python env path in sglang by @stmatengss in #196
• [CI] Update etcd-cpp CMake policy version in dependencies script by @xiaguan in #199
• [Feature] use randomly-selected port in TCP-based OOB communication by @alogfans in #202
• [FIX] prevent allocating the same port by using real random values by @alogfans in #206
• [TransferEngine] set use http protocol as default by @stmatengss in #208
• [CI] Update GitHub Actions permissions for build-and-release job by @xiaguan in #209
• Enable ccache in Mooncake compile by @yuan-luo in #203
• [TransferEngine] fix: avoid leaking Slice memory in RDMA transport by @eryugey in #210
• [Refactor] Remove the dependency of etcd-cpp-api-v3 by @alogfans in #188
• [MooncakeIntegration] Refactor py mod arch by @stmatengss in #212
• feat(ci): enhance build process with new import structure test by @xiaguan in #214
• chore(scripts): set rpath for shared objects in build script by @xiaguan in #216
• feat(mooncakestore): add mooncake_master CLI entry point and tests by @xiaguan in #218
• [TransferEngine] Use RDMA transport to transfer data in local process rapidly by @alogfans in #220
• [Doc] Refactor dependencies.sh and for improved installation process by @xiaguan in #195
• [FIX] reduce performance overhead from allocating slices by @alogfans in #223
• [FIX] bind failure from legacy RPC_PORT_BINDING by @alogfans in #226
• [Doc] Add sglang integration doc and update vllm integration doc by @ShangmingCai in #227
• Update go version to 1.23.8 in dependencies.sh and other enhancements by @eryugey in #231
• chore(ci): integrate auditwheel ...

v0.3.0-beta

Highlights

• Develop MooncakeStore: A new Distributed Object Store for XpYd PD disaggregation
• Support vLLM V0, see: vllm-project/vllm#12957
• Support SGLang, see: sgl-project/sglang#4880 (WIP)
• Provide better installation methods.
• ....

What's Changed

• fix: when detect async event, stop the outstanding requests & prevent cq burn out leading severe error by @alogfans in #84
• [TransferEngine] fix: add auto_discover in transfer_engine_c by @chestnut-Q in #87
• [TransferEngine] feature: registerLocalMemory support the "*" location. by @doujiang24 in #86
• [Build] feature: add dockerfile to enable use Mooncake in Docker by @stmatengss in #92
• [Transfer Engine] add USE_ETCD option to retire the etcd dependency. by @doujiang24 in #94
• [TransferEngine]chore: replace slices vector with slice counter by @RuixiangMa in #93
• [Build] feature: add dev container. by @doujiang24 in #96
• [TransferEngine] fix: wrong numa number in example. by @doujiang24 in #99
• [TransferEngine] fix: use ibv_get_device_list to get the IB devices. by @doujiang24 in #102
• fix: memcpy direction wrong for read operations by @alogfans in #101
• [TransferEngine] Build: build/install shared object for transfer engine. by @doujiang24 in #100
• Fix P2P store bugs and improvement by @alogfans in #80
• [Bugfix] Add shared and unique lock for local_memory_regions_ (#107) by @power-more in #108
• [TransferEngine] Correct the Typo by @power-more in #109
• [TransferEngine] Fix typos by @fengquyoumo in #110
• [TransferEngine] BenchMark: Introduce report gflags for enhanced configurability by @fengquyoumo in #111
• [TransferEngine] feature: introduce USE_NVMEOF to enable NVMe-oF separately. by @doujiang24 in #106
• [Doc] add FAST25 paper and traces by @chestnut-Q in #116
• [TransferEngine] Suppress gtest type cast compile warning and fix some trivial bugs by @yuan-luo in #115
• [Doc] update README.md by @chestnut-Q in #128
• [Bugfix] fix the gid choice of IB device: only choose IBV_GID_TYPE_ROCE_V2 now by @fengquyoumo in #113
• [MooncakeStore] Introduce MooncakeStore: A new Distributed Object Store for XpYd PD disaggregation by @alogfans in #126
• [FIX] Fix the content of README.md by @alogfans in #130
• [DOC] add description about Docker-based deployment by @alogfans in #133
• Fix compilation error with clang toolchain by @xiaguan in #135
• [MooncakeStore] Add getSize method to DistributedObjectStore by @xiaguan in #137
• [DOC] remove useless metadata_type flags in example.md by @cxz66666 in #138
• [FIX] golang vulnerabilities in P2P store by @alogfans in #140
• [TransferEngine] Support Status return value by @yuan-luo in #125
• [Doc] Fix typo in mooncake-store-preview by @simpx in #142
• [MooncakeStore] Enhance performance in DistributedObjectStore::get by @xiaguan in #141
• [Doc] Add vllm integration v1 doc by @ShangmingCai in #129
• [Optimize] Update EndpointStore interface to use const references for… by @xiaguan in #152
• fix compilation errors by @xhyf77 in #145
• [MooncakeStore] fix: treat object_not_found/already_exists as info in glog by @stmatengss in #156
• [Fix]: Handle SIGINT & SIGTERM for graceful shutdown by @xhyf77 in #153
• [Doc] typo error in README by @Unprincess17 in #158
• [FEATURE] catch exceptions during allocator creation by @xiaguan in #157
• [FEATURE] Update MooncakeStroe CI workflow by @xiaguan in #146
• [DOC] remove MULAN license to ease vllm deployment by @stmatengss in #163
• [Refactor] Replace rpc logic with MasterClient in Client class by @xiaguan in #164
• [CI] Add spell check process to CI by @xiaguan in #165
• [Chore] Disable garbage collection in Mooncake master by @xiaguan in #167
• [TransferEngine] fix typos for cancel status in transfer engine by @alogfans in #169
• Extend VLLMAdaptor API by @xhyf77 in #173
• [MooncakeAdaptor] Init sglang_adaptor to support SGLang using transfer engine by @stmatengss in #181
• [HOTFIX] add env MC_DISABLE_METACACHE to force pull metadata from etcd by @alogfans in #175
• [TransferEngine] Fix compilation bugs of nvmeof transport by @alogfans in #174
• Package mooncake_vllm_adaptor and MooncakeDistributedStore into a wheel by @xhyf77 in #177
• [Fix] Fix issues with run path and etcd dependency integration by @xhyf77 in #182
• [WorkFlow] Add a release.yaml to publish the .whl file by @xhyf77 in #186
• [Chore] Change logging level from INFO to VLOG for object existence by @xiaguan in #185
• [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc by @xiaguan in #179
• [Doc] Update vllm MooncakeStore integration doc by @ShangmingCai in #191
• [CI] Refine release workflow to enable publishing to PyPI by @xiaguan in #189
• [Enhance] Add an RPATH that takes precedence over LD_LIBRARY_PATH and other paths in the wheel package by @xhyf77 in #190
• [MooncakeStore] fix: python env path in sglang by @stmatengss in #196
• [CI] Update etcd-cpp CMake policy version in dependencies script by @xiaguan in #199
• [Feature] use randomly-selected port in TCP-based OOB communication by @alogfans in #202
• [FIX] prevent allocating the same port by using real random values by @alogfans in #206
• [TransferEngine] set use http protocol as default by @stmatengss in #208
• [CI] Update GitHub Actions permissions for build-and-release job by @xiaguan in #209
• Enable ccache in Mooncake compile by @yuan-luo in #203
• [TransferEngine] fix: avoid leaking Slice memory in RDMA transport by @eryugey in #210
• [Refactor] Remove the dependency of etcd-cpp-api-v3 by @alogfans in #188
• [MooncakeIntegration] Refactor py mod arch by @stmatengss in #212
• feat(ci): enhance build process with new import structure test by @xiaguan in #214
• chore(scripts): set rpath for shared objects in build script by @xiaguan in #216

New Contributors

• @RuixiangMa made their first contribution in #93
• @power-more made their first contribution in #108
• @yuan-luo made their first contribution in #115
• @xiaguan made their first contribution in #135
• @cxz66666 made their first contribution in #138
• @simpx made their first contribution in #142
• @xhyf77 made their first contribution in #145
• @Unprincess17 made their first contribution in #158
• @eryugey made their first contribution in #210

Full Changelog: v0.2.0...v0.3.0-beta

v0.3.0-alpha

Highlights

• Develop MooncakeStore: A new Distributed Object Store for XpYd PD disaggregation
• Support vLLM V0, see: vllm-project/vllm#12957
• Support SGLang, see: sgl-project/sglang#4880 (WIP)
• Provide better installation methods.
• ....

What's Changed

• fix: when detect async event, stop the outstanding requests & prevent cq burn out leading severe error by @alogfans in #84
• [TransferEngine] fix: add auto_discover in transfer_engine_c by @chestnut-Q in #87
• [TransferEngine] feature: registerLocalMemory support the "*" location. by @doujiang24 in #86
• [Build] feature: add dockerfile to enable use Mooncake in Docker by @stmatengss in #92
• [Transfer Engine] add USE_ETCD option to retire the etcd dependency. by @doujiang24 in #94
• [TransferEngine]chore: replace slices vector with slice counter by @RuixiangMa in #93
• [Build] feature: add dev container. by @doujiang24 in #96
• [TransferEngine] fix: wrong numa number in example. by @doujiang24 in #99
• [TransferEngine] fix: use ibv_get_device_list to get the IB devices. by @doujiang24 in #102
• fix: memcpy direction wrong for read operations by @alogfans in #101
• [TransferEngine] Build: build/install shared object for transfer engine. by @doujiang24 in #100
• Fix P2P store bugs and improvement by @alogfans in #80
• [Bugfix] Add shared and unique lock for local_memory_regions_ (#107) by @power-more in #108
• [TransferEngine] Correct the Typo by @power-more in #109
• [TransferEngine] Fix typos by @fengquyoumo in #110
• [TransferEngine] BenchMark: Introduce report gflags for enhanced configurability by @fengquyoumo in #111
• [TransferEngine] feature: introduce USE_NVMEOF to enable NVMe-oF separately. by @doujiang24 in #106
• [Doc] add FAST25 paper and traces by @chestnut-Q in #116
• [TransferEngine] Suppress gtest type cast compile warning and fix some trivial bugs by @yuan-luo in #115
• [Doc] update README.md by @chestnut-Q in #128
• [Bugfix] fix the gid choice of IB device: only choose IBV_GID_TYPE_ROCE_V2 now by @fengquyoumo in #113
• [MooncakeStore] Introduce MooncakeStore: A new Distributed Object Store for XpYd PD disaggregation by @alogfans in #126
• [FIX] Fix the content of README.md by @alogfans in #130
• [DOC] add description about Docker-based deployment by @alogfans in #133
• Fix compilation error with clang toolchain by @xiaguan in #135
• [MooncakeStore] Add getSize method to DistributedObjectStore by @xiaguan in #137
• [DOC] remove useless metadata_type flags in example.md by @cxz66666 in #138
• [FIX] golang vulnerabilities in P2P store by @alogfans in #140
• [TransferEngine] Support Status return value by @yuan-luo in #125
• [Doc] Fix typo in mooncake-store-preview by @simpx in #142
• [MooncakeStore] Enhance performance in DistributedObjectStore::get by @xiaguan in #141
• [Doc] Add vllm integration v1 doc by @ShangmingCai in #129
• [Optimize] Update EndpointStore interface to use const references for… by @xiaguan in #152
• fix compilation errors by @xhyf77 in #145
• [MooncakeStore] fix: treat object_not_found/already_exists as info in glog by @stmatengss in #156
• [Fix]: Handle SIGINT & SIGTERM for graceful shutdown by @xhyf77 in #153
• [Doc] typo error in README by @Unprincess17 in #158
• [FEATURE] catch exceptions during allocator creation by @xiaguan in #157
• [FEATURE] Update MooncakeStroe CI workflow by @xiaguan in #146
• [DOC] remove MULAN license to ease vllm deployment by @stmatengss in #163
• [Refactor] Replace rpc logic with MasterClient in Client class by @xiaguan in #164
• [CI] Add spell check process to CI by @xiaguan in #165
• [Chore] Disable garbage collection in Mooncake master by @xiaguan in #167
• [TransferEngine] fix typos for cancel status in transfer engine by @alogfans in #169
• Extend VLLMAdaptor API by @xhyf77 in #173
• [MooncakeAdaptor] Init sglang_adaptor to support SGLang using transfer engine by @stmatengss in #181
• [HOTFIX] add env MC_DISABLE_METACACHE to force pull metadata from etcd by @alogfans in #175
• [TransferEngine] Fix compilation bugs of nvmeof transport by @alogfans in #174
• Package mooncake_vllm_adaptor and MooncakeDistributedStore into a wheel by @xhyf77 in #177
• [Fix] Fix issues with run path and etcd dependency integration by @xhyf77 in #182
• [WorkFlow] Add a release.yaml to publish the .whl file by @xhyf77 in #186
• [Chore] Change logging level from INFO to VLOG for object existence by @xiaguan in #185
• [Refactor] [MooncakeStore] Migrate Master Service from gRPC to coro_rpc by @xiaguan in #179
• [Doc] Update vllm MooncakeStore integration doc by @ShangmingCai in #191
• [CI] Refine release workflow to enable publishing to PyPI by @xiaguan in #189
• [Enhance] Add an RPATH that takes precedence over LD_LIBRARY_PATH and other paths in the wheel package by @xhyf77 in #190
• [MooncakeStore] fix: python env path in sglang by @stmatengss in #196
• [CI] Update etcd-cpp CMake policy version in dependencies script by @xiaguan in #199
• [Feature] use randomly-selected port in TCP-based OOB communication by @alogfans in #202
• [FIX] prevent allocating the same port by using real random values by @alogfans in #206
• [TransferEngine] set use http protocol as default by @stmatengss in #208

New Contributors

• @RuixiangMa made their first contribution in #93
• @power-more made their first contribution in #108
• @yuan-luo made their first contribution in #115
• @cxz66666 made their first contribution in #138
• @simpx made their first contribution in #142
• @xhyf77 made their first contribution in #145
• @Unprincess17 made their first contribution in #158

Full Changelog: v0.2.0...v0.3.0

What's Changed

• fix: when detect async event, stop the outstanding requests & prevent cq burn out leading severe error by @alogfans in #84
• [TransferEngine] fix: add auto_discover in transfer_engine_c by @chestnut-Q in #87
• [TransferEngine] feature: registerLocalMemory support the "*" location. by @doujiang24 in #86
• [Build] feature: add dockerfile to enable use Mooncake in Docker by @stmatengss in #92
• [Transfer Engine] add USE_ETCD option to retire the etcd dependency. by @doujiang24 in #94
• [TransferEngine]chore: replace slices vector with slice counter by @RuixiangMa in #93
• [Build] feature: add dev container. by @doujiang24 in #96
• [TransferEngine] fix: wrong numa number in example. by @doujiang24 in #99
• [TransferEngine] fix: use ibv_get_device_list to get the IB devices. by @doujiang24 in #102
• fix: memcpy direction wrong for read operations by @alogfans in https...

v0.2.0

The first official release of Mooncake!

See our README for details.

What's Changed

• docs: update README.md by @eltociear in #9
• fix the p2pstore go module name, make it could be imported in other projects. by @doujiang24 in #7
• docs: add new vllm-integration guide. by @ShangmingCai in #11
• fix: remove atty and fix dependency path by @chestnut-Q in #19
• [DOC] Add contributing guidelines by @stmatengss in #22
• [Bugfix] cpp std::string by @wxsms in #18
• [Doc][Integration] Update commands, params description, and benchmark results for vllm integration v2. by @ShangmingCai in #25
• [Doc] Fix typos in contributing.md by @ShangmingCai in #26
• [TransferEngine][feature] support multiple metadata servers by @alogfans in #21
• [CI] add build test on ubuntu by @Ann-1024 in #28
• [Doc] Add metadata server backend explanation. by @ShangmingCai in #30
• [Bugfix] build error when USE_CUDA by @wxsms in #32
• [Bugfix] fix typos and fix rdma device selection when retry_count != 0 by @fengquyoumo in #33
• [Doc] Update the integration state of Mooncake Transfer Engine with vLLM. by @ShangmingCai in #40
• [Doc] Update Mooncake Icon by @stmatengss in #41
• [Doc] Update README to use v0.2 guide and benchmark results. by @ShangmingCai in #42
• [TransferEngine] Add topology discovery by @wx-csy in #46
• [TransferEngine][feature] add the http metadata server. by @doujiang24 in #37
• [TransferEngine] Refactor code to hide transport logics from user APIs by @alogfans in #51
• [Transfer Engine] fix metadata server connection string builder in vllm intergation by @alogfans in #56
• Update metadata_server config info for vllm integration doc. by @ShangmingCai in #57
• Bump golang.org/x/crypto to 0.31.0 & golang.org/x/net to v0.33.0 by @doujiang24 in #58
• [TransferEngine] fix: selectDevice() should return zero at all case by @alogfans in #60
• [p2pstore] fix memory leaking in cgo. by @doujiang24 in #61
• [TransferEngine] test: cmake enable testing. by @doujiang24 in #65
• [Doc] Re-enable cuda graph to improve inference performance. by @ShangmingCai in #67
• [TransferEngine] fix: improve error message by @doujiang24 in #63
• [TransferEngine] adjust transfer_engine_bench: 1.Introduce the gflag buffer_size for enhanced configurability. 2. Utilize uint64_t for block_size to prevent overflow. by @fengquyoumo in #72
• [TransferEngine] Add tests for Topology by @liusy58 in #74
• [TransferEngine] Correct the count of slice_quque_ in rdma/worker_pool, remove the repeated check for overlap by @fengquyoumo in #79
• fix invalid go release url by @gujingit in #83
• [TransferEngine] change: auto discover topology & install transport. by @doujiang24 in #73

New Contributors

• @eltociear made their first contribution in #9
• @doujiang24 made their first contribution in #7
• @ShangmingCai made their first contribution in #11
• @chestnut-Q made their first contribution in #19
• @stmatengss made their first contribution in #22
• @wxsms made their first contribution in #18
• @alogfans made their first contribution in #21
• @Ann-1024 made their first contribution in #28
• @fengquyoumo made their first contribution in #33
• @wx-csy made their first contribution in #46
• @liusy58 made their first contribution in #74
• @gujingit made their first contribution in #83

Full Changelog: https://github.com/kvcache-ai/Mooncake/commits/v0.2.0