{"full_name":"sgl-project/sglang-omni","owner":"sgl-project","name":"sglang-omni","shard":"1","url":"https://github.com/sgl-project/sglang-omni","description":"High Performance Serving Runtime For Interactive Super Intelligence System","homepage":"https://sgl-project.github.io/sglang-omni/","topics":["asr","audio-generation","cuda","inference","model-serving","multimodal","openai-api","pytorch","sglang","tts","distributed-inference","music-generation","speech-recognition","streaming","text-to-speech"],"tag_pages":["asr","cuda","inference","model-serving","multimodal","openai-api","pytorch","speech-recognition","streaming","text-to-speech","tts"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/147780389?v=4","created_at":"2026-01-07T06:58:29Z","pushed_at":"2026-11-15T18:57:02Z","metrics":{"stars":1793,"forks":741,"watchers":9,"issues":665,"prs":3332,"releases":23,"issues_open":350,"issues_closed":315,"prs_open":812,"prs_closed":694,"prs_merged":1826,"discussions":1},"growth":{"day":{"stars":8,"forks":10,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[1756,1771,1771,1778,1785,1793],"forks":[720,726,726,731,731,741],"watchers":[9,9,9,9,9,9],"issues_open":[302,314,319,324,332,350],"prs_open":[811,836,842,788,794,812]},"boards":{},"same_owner":[{"full_name":"sgl-project/mini-sglang","name":"mini-sglang","shard":"1","boards":{}},{"full_name":"sgl-project/sglang","name":"sglang","shard":"1","boards":{"daily_forks":181}}],"issues":[{"number":2994,"title":"[Roadmap] Interactive Intelligence Runtime Optimization","comments":67,"created_at":"2026-10-20T16:42:26Z"},{"number":2941,"title":"[Roadmap] Speculative decoding for speech models","comments":48,"created_at":"2026-10-19T04:25:42Z"},{"number":1754,"title":"[Tracking] Qwen3-TTS Optimization Follow-Up Roadmap","comments":29,"created_at":"2026-08-26T18:51:29Z"},{"number":2888,"title":"[Roadmap] Apple Silicon (native MLX) 2026 Q4","comments":26,"created_at":"2026-10-17T02:33:37Z"},{"number":1967,"title":"[RFC] Apple Silicon support roadmap","comments":23,"created_at":"2026-09-05T06:11:08Z"},{"number":1652,"title":"[Roadmap] Optimization for CosyVoice3","comments":22,"created_at":"2026-08-21T15:49:07Z"},{"number":2839,"title":"[Roadmap] Cookbook Refactor Roadmap","comments":20,"created_at":"2026-10-15T06:17:47Z"},{"number":1650,"title":"[Roadmap] Runtime Profile Agent for SGLang Omni","comments":20,"created_at":"2026-08-21T08:18:16Z"}],"discussions":[{"number":2158,"title":"Performance of SGLang-Omni on Apple Silicon Mac","comments":15,"category":"Show and tell","created_at":"2026-09-14T01:54:03Z"}],"latest_release":{"name":"Voxt v2.0.3","tag":"voxt-v2.0.3","published_at":"2026-11-10T02:12:21Z","body":"### Requirements\r\n\r\n- A Mac with Apple silicon and macOS 15 or later.\r\n\r\n### Install\r\n\r\n- Open Voxt-2.0.3-arm64.dmg and drag Voxt to Applications. It replaces Voxt 1.x there.\r\n- This version does not update itself: install a newer one the same way.\r\n\r\n\r\n### Changed\r\n\r\n- **Plainer text across the app**: about 300 English strings stop chaining clauses with colons and semicolons and leave out details nobody needs, about 360 Chinese strings read as written in Chinese and name each thing one way, and about 50 help texts that only said what the page already shows are gone. In Chinese, Rewrite is now 改写 (转写 also meant a transcript), a transcript is 转录稿, speech models are 语音识别模型 and an app is always App.\r\n- **Sound: Built-in, macOS or None** (Settings → Audio): one choice says where the recording's start and stop sounds come from. Built-in lists six pairs of a start and a stop sound, each to try with ▶ and pick with a click; macOS picks a system sound for each; None plays nothing. The default pair is now Pebble and Nib, softer than Tally and Wrap. The panel of sounds by group, tag and search is gone, and so is Use Voxt 1.x's sounds: macOS starts on Pop and Tink.\r\n- **Settings sits in the sidebar's System section**, under Logs, instead of a row of its own at the foot of the sidebar; the version number is on About. In the icon rail its icon is in that section too, with the dot while there are notifications.\r\n\r\n### 中文\r\n\r\n#### 变更\r\n\r\n- **中英文文案过了一遍**：约 300 条英文不再用冒号、分号把几句硬连在一起，也删掉了用不着的细节；约 360 条中文去掉了翻译腔，同一个东西只用一种叫法；约 50 条只是重复界面内容的说明删掉了。Rewrite 改叫“改写”（原来的“转写”也指转录稿），会议和文件的 transcript 统一叫“转录稿”，语音模型统一叫“语音识别模型”，app 统一写作“App”。\r\n- **音效：自带、系统或无**（设置 → 音频）：一个选项决定录音开始和结束的提示音从哪里来。“自带”列出 6 组成对的开始音和结束音，点 ▶ 试听，点一下选中；“系统”给开始和结束各选一个 macOS 声音；“无”就不响。默认换成更轻的 Pebble 和 Nib（原来是 Tally 和 Wrap）。按分组、标签和搜索找声音的面板去掉了，“使用 Voxt 1.x 的提示音”也去掉了：选“系统”默认就是 Pop 和 Tink。\r\n- **“设置”放进了侧边栏的“系统”分组**，在“日志”下面，不再单独占侧边栏底部一行；版本号在“诊断与关于”里。侧边栏收成图标时，“设置”图标也在这一组里，有通知时照样显示红点。\r\n\r\n### Build\r\n\r\n- Built from 4e03928295bba17c8196644f50d69e58facff6ea by the Voxt Release workflow. Signed with Developer ID and notarized by Apple. The SHA-256 checksum is attached.\r\n\r\n## What's Changed\r\n* [Voxt] Sign and notarize a nightly build of main by @XXXXRT666 in https://github.com/sgl-project/sglang-omni/pull/3637\r\n* [Docs] Shorten Get Started and align documentation paths with navigation by @AkazaAkane in https://github.com/sgl-project/sglang-omni/pull/3607\r\n* [Voxt] Import Voxt 1.x settings, add a speech context, build a signed beta of each merge, fix two layout bugs by @XXXXRT666 in https://github.com/sgl-project/sglang-omni/pull/3655\r\n* [Cleanup][Fun-CosyVoice3] Halve the DSpark exact verify's cost and kernel launches by @YifanLi3 in https://github.com/sgl-project/sglang-omni/pull/3470\r\n* [MLX] Native Fish S2-Pro server on the C++ runtime by @JiaxinD in https://github.com/sgl-project/sglang-omni/pull/3142\r\n* [Profiler] Capture CPU operations from scheduler threads by @DerekSun02 in https://github.com/sgl-project/sglang-omni/pull/2419\r\n* [Benchmark][Speculative] Add SeedTTS workloads to the draft sweep by @swsqy in https://github.com/sgl-project/sglang-omni/pull/3585\r\n* [Speculative][P5.5] Validate Qwen3-ASR realtime transcription with the DFlash draft by @HaolingPu in https://github.com/sgl-project/sglang-omni/pull/3572\r\n* [Qwen3-Omni] Forward the dialog configure to qwen_omni in the H100 session cfg by @0xjeffro in https://github.com/sgl-project/sglang-omni/pull/3662\r\n* [Fix] Reject negative ASR temperature and fall back to soundfile when torchcodec cannot decode by @HaolingPu in https://github.com/sgl-project/sglang-omni/pull/3634\r\n* [Benchmark] Retry in-session admission rejections in the duplex recorder (C2) by @ychu824 in https://github.com/sgl-project/sglang-omni/pull/3038\r\n* [Qwen3-ASR] Honor repetition_penalty on CUDA and Torch MPS by @minyi-tang in https://github.com/sgl-project/sglang-omni/pull/3415\r\n* [Voxt] 2.0.2: Voxt 1.x parity — shortcut presets, mouse buttons, translate fallback, in-app upda","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}