Parallax Files

Highlights

• Support for Kimi-K2, DeepSeek, MiniMax, GLM and gpt-oss-safeguard
• Optimize communication methods and add relay server support for remote connections
• Add parallax chat command to run a chat page server
• Adjust the readme structure and improve the user guide and contributing guide

What's Changed

• fix(frontend): chat page sidebar subtitle and add node dialog by @xz-gradient in https://github.com/GradientHQ/parallax/pull/44
• fix(node): Fixed the issue of slow reporting of node leave by @sl-gn in https://github.com/GradientHQ/parallax/pull/46
• fix(frontend): setup input, node offline by @xz-gradient in https://github.com/GradientHQ/parallax/pull/47
• feat(model): Add Model kimi k2(DeepSeekV3) by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/10
• docs(readme): update readme for scheduling algorithm by @Youhe-Jiang in https://github.com/GradientHQ/parallax/pull/43
• chore(lattica): use lattica 1.0.0 by @gufengc in https://github.com/GradientHQ/parallax/pull/50
• fix(batch scheduler): add mlx model name map & add gpt-oss monkey patch by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/45
• fix: ui issues by @yuyin-zhang in https://github.com/GradientHQ/parallax/pull/49
• feat(frontend): chat message output fit gpt format by @xz-gradient in https://github.com/GradientHQ/parallax/pull/51
• feat(readme): update readme for product by @ramenyu in https://github.com/GradientHQ/parallax/pull/52
• feat(readme): update readme images by @ramenyu in https://github.com/GradientHQ/parallax/pull/53
• feat(backend): add ascii animation for parallax run/parallax join by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/48
• feat(Relay): Adding relay server by @gufengc in https://github.com/GradientHQ/parallax/pull/19
• fix(frontend): UI improvements for chat, sidebar, select, and number input by @yuyin-zhang in https://github.com/GradientHQ/parallax/pull/55
• fix(model): mlx gpt-oss model name by @gufengc in https://github.com/GradientHQ/parallax/pull/59
• chore(readme): Update README.md by @Youhe-Jiang in https://github.com/GradientHQ/parallax/pull/58
• feat(backend): Backend support connnect relay server & Disable caching for index.html by @sl-gn in https://github.com/GradientHQ/parallax/pull/56
• feat(readme): update readme images and cluster setup tutorial by @ramenyu in https://github.com/GradientHQ/parallax/pull/57
• fix(node): Close http server process in finally and fix the problem t… by @sl-gn in https://github.com/GradientHQ/parallax/pull/61
• fix(readme): update supported models info and add is-local-network param when start backend by @sl-gn in https://github.com/GradientHQ/parallax/pull/62
• fix(scheduler): Layer allocator parameter size calculation by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/60
• feat(scheduler): sending chat complete request through p2p by @gufengc in https://github.com/GradientHQ/parallax/pull/63
• fix(node): change ascii animation display time by @sl-gn in https://github.com/GradientHQ/parallax/pull/64
• docs(readme): add uninstalling in readme by @sl-gn in https://github.com/GradientHQ/parallax/pull/65
• fix(http): fix localhost bypass proxy by @gufengc in https://github.com/GradientHQ/parallax/pull/67
• fix(LICENSE): Create LICENSE by @sl-gn in https://github.com/GradientHQ/parallax/pull/66
• fix(description): update project description and cli help info and re… by @sl-gn in https://github.com/GradientHQ/parallax/pull/68
• fix(scheduler): Fix not read model config from cache by @gufengc in https://github.com/GradientHQ/parallax/pull/70
• fix(relay): scheduler add default initial_peers in remote mode by @sl-gn in https://github.com/GradientHQ/parallax/pull/69
• chore(lattica): upgrade lattica to 1.0.4 by @gufengc in https://github.com/GradientHQ/parallax/pull/71
• feat(model): Add qwen3-235B-int4 by @gufengc in https://github.com/GradientHQ/parallax/pull/72
• feat(readme): Guide macOS to use python venv by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/73
• fix(frontend): chat model determine incorrect by @xz-gradient in https://github.com/GradientHQ/parallax/pull/75
• fix(scheduler): do not calculate batch size by @gufengc in https://github.com/GradientHQ/parallax/pull/77
• refactor(ascii-anime): refactor ascii anime to static display by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/78
• feat(frontend): node statusn fix, node list style and dash animate by @xz-gradient in https://github.com/GradientHQ/parallax/pull/90
• feat(frontend): update web page title by @xz-gradient in https://github.com/GradientHQ/parallax/pull/94
• fix(lattica): Fix rpc_stream disconnect by @gufengc in https://github.com/GradientHQ/parallax/pull/97
• perf(cli): pass through other argument to launch.py and fix press twice ctrl-c by @gufengc in https://github.com/GradientHQ/parallax/pull/74
• fix(log): set log level by --log-level by @gufengc in https://github.com/GradientHQ/parallax/pull/99
• fix(scheduler): do not check heartbeat of deactive node by @gufengc in https://github.com/GradientHQ/parallax/pull/105
• fix(model): modify model name kimi2deepseek by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/100
• fix(backend): silence tokenizer process fork warnings by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/107
• fix(executor): p2p recv_req's hidden_states dtype not equal to self.dtype by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/96
• feat(lattica): specify tcp&udp port by @gufengc in https://github.com/GradientHQ/parallax/pull/110
• fix(backend): Use the project root dir to find static file resources by @sl-gn in https://github.com/GradientHQ/parallax/pull/117
• fix(Scheduler): fix scheduler request id dismatch with work request id for same request by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/116
• docs(readme): update readme and add model in model list config by @sl-gn in https://github.com/GradientHQ/parallax/pull/118
• fix(build): update python version limit by @sl-gn in https://github.com/GradientHQ/parallax/pull/119
• fix(lattica): fix deadlock issue by @gufengc in https://github.com/GradientHQ/parallax/pull/121
• docs: update readme by @ramenyu in https://github.com/GradientHQ/parallax/pull/128
• feat(node): node local chat page by @sl-gn in https://github.com/GradientHQ/parallax/pull/129
• docs(readme): add deepseek in support models by @sl-gn in https://github.com/GradientHQ/parallax/pull/130
• chore(relay): add more relay server by @gufengc in https://github.com/GradientHQ/parallax/pull/131
• feat(frontend): add more model logos by @xz-gradient in https://github.com/GradientHQ/parallax/pull/132
• docs(readme): add node local chat page in readme and change some images by @sl-gn in https://github.com/GradientHQ/parallax/pull/136
• fix(frontend): output crash (message offset out) with table by @xz-gradient in https://github.com/GradientHQ/parallax/pull/137
• feat(cli): add release version check and upload install info by @sl-gn in https://github.com/GradientHQ/parallax/pull/138
• feat(node): skip node chat server when scheduler_addr is not specified by @sl-gn in https://github.com/GradientHQ/parallax/pull/139
• fix(node): disable mdns if initial peers is specified by @gufengc in https://github.com/GradientHQ/parallax/pull/141
• docs(readme): update windows cli url by @sl-gn in https://github.com/GradientHQ/parallax/pull/142
• chore(lattica): upgrade lattica to 1.0.8 by @gufengc in https://github.com/GradientHQ/parallax/pull/149
• fix(http_server): return token usage data by @gufengc in https://github.com/GradientHQ/parallax/pull/150
• fix(model): fix qwen3 next bug & update sgalng to 0.5.4.post1 by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/133
• fix(backend): fix localhost can not be visited by other pc by @sl-gn in https://github.com/GradientHQ/parallax/pull/154
• docs(readme): add license and issue count in readme by @sl-gn in https://github.com/GradientHQ/parallax/pull/155
• chore(lattica): check nat type is symmetric by @gufengc in https://github.com/GradientHQ/parallax/pull/156
• docs(readme): update discord link by @sl-gn in https://github.com/GradientHQ/parallax/pull/162
• feat(chat): Separate the chat page server from the node by @sl-gn in https://github.com/GradientHQ/parallax/pull/163
• feat(model): add minimax m2 inplace without updata mlx-lm by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/161
• feat(model): add glm4_moe support by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/165
• feat(frontend): add more model logos by @xz-gradient in https://github.com/GradientHQ/parallax/pull/164
• docs(readme): add minimax and glm in supported models by @sl-gn in https://github.com/GradientHQ/parallax/pull/167
• fix(readme): fix version by @gufengc in https://github.com/GradientHQ/parallax/pull/169
• docs(readme): local network permission for macos and --host params by @sl-gn in https://github.com/GradientHQ/parallax/pull/179
• feat(scheduer): support different bits model by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/171
• feat(backend): add param_hosting_ratio args by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/183
• fix(frontend): node skeleton style and dash, chat input problem by @xz-gradient in https://github.com/GradientHQ/parallax/pull/95
• fix(server): Added HardwareInfo for Apple M2 Ultra by @uebber in https://github.com/GradientHQ/parallax/pull/180
• feat(backend): display tps ttft input_tokens and output_tokens in logs by @sl-gn in https://github.com/GradientHQ/parallax/pull/178
• fix(backend): 2-phase continuous batching by @christ-tt in https://github.com/GradientHQ/parallax/pull/122
• docs(readme): add partners and news by @sl-gn in https://github.com/GradientHQ/parallax/pull/186
• feat(backend): refactor sglang monkey patch code by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/176
• feat(p2p): add layer allocation logging in heartbeat by @JasonOE in https://github.com/GradientHQ/parallax/pull/185
• feat(scheduler): Manually specify layers for convenient debug by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/189
• perf(scheduler): Don't route request if pipeline rtt is INF by @gufengc in https://github.com/GradientHQ/parallax/pull/184
• docs(contributing): add contributing guide by @sl-gn in https://github.com/GradientHQ/parallax/pull/193
• feat(backend): reload layers after layer change by @JasonOE in https://github.com/GradientHQ/parallax/pull/191
• feat(loader): Only read and download necessery weight files by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/192
• fix(loader): fix offline model loader bug by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/195
• fix(api): fix func args name by @yuhao-zh in https://github.com/GradientHQ/parallax/pull/199
• chore(model): add kimi k2 thinking and gpt-oss safeguard by @gufengc in https://github.com/GradientHQ/parallax/pull/201
• refactor(cli): prompt users to report information when connecting to … by @sl-gn in https://github.com/GradientHQ/parallax/pull/202
• feat(scheduler): scheduler change after node leave by @JasonOE in https://github.com/GradientHQ/parallax/pull/194
• fix(dependency): Add missing orjson dependency for ORJSONResponse by @IAMDAVID0920 in https://github.com/GradientHQ/parallax/pull/205
• feat(cli): change version info upload content by @sl-gn in https://github.com/GradientHQ/parallax/pull/211
• docs(readme): adjust the readme structure and add some note by @sl-gn in https://github.com/GradientHQ/parallax/pull/213
• feat(backend): Support GPU TP and enable TP for non-scheduler mode by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/173
• fix(backend): GPU seq_len_cpu mismatch & add dummy tree_cache methods by @TianyiZhao1437 in https://github.com/GradientHQ/parallax/pull/214
• feat(backend): add recommand vram and update version to 0.1.0 by @sl-gn in https://github.com/GradientHQ/parallax/pull/218

New Contributors

• @Youhe-Jiang made their first contribution in https://github.com/GradientHQ/parallax/pull/43
• @ramenyu made their first contribution in https://github.com/GradientHQ/parallax/pull/52
• @uebber made their first contribution in https://github.com/GradientHQ/parallax/pull/180
• @JasonOE made their first contribution in https://github.com/GradientHQ/parallax/pull/185
• @IAMDAVID0920 made their first contribution in https://github.com/GradientHQ/parallax/pull/205

Full Changelog: https://github.com/GradientHQ/parallax/compare/v0.0.1...v0.1.0