sgl-project/sglang

• First survey — no prior run to compare against. CAI 56.

• Code Health 62
• Architecture 94
• Maturity 77
• Readiness 56
• Security 56
• Event Sourcing 100
• Accessibility 51
• Performance 72

• 300 commits — 220 feature/other, 80 fixes

• python/sglang — 205 commits
• experimental/sgl-router — 27 commits
• test/registered — 22 commits
• .github/workflows — 9 commits
• scripts/ci — 5 commits
• docs/cookbook — 4 commits
• rust/sglang-server — 4 commits
• .github/CI_PERMISSIONS.json — 3 commits
• docs/docs — 3 commits
• docs/src — 3 commits
• .github/actions — 2 commits
• docker/rocm.Dockerfile — 2 commits
• python/pyproject.toml — 2 commits
• (root) — 1 commit
• .agents/skills — 1 commit
• docker/Dockerfile — 1 commit
• docs/cards — 1 commit
• rust/sglang-grpc — 1 commit
• rust/sglang-radix-tree — 1 commit
• rust/sglang-renderer — 1 commit

• fix: Fix Kimi-K3 MLA output gate dispatch on non-CUDA devices (#36406)
• fix: Fix LoRA usage-counter accounting across request lifecycle paths (#31808)
• fix: Fix MUSA detection under torch.compile fullgraph (#41065)
• fix: Fix pip install: exclude multimodal_gen/.claude symlink from package data (#41703)
• fix: [AMD] Fix AITER weight slicing for CP decode TP (#39950)
• fix: [AMD] Fix GLM-5.3 quark MoE MI35x test runner config (#41464)
• fix: [AMD] Fix Load and Inference of MLA models with Quark PTPC FP8 attention on ROCm (#28734)
• fix: [AMD] Fix MiniMax-M3 EAGLE3 verification on ROCm (#41497)
• fix: [AMD] fix kda decode flydsl import (#40546)
• fix: [AMD][CI] Fix VLM MMMU nightly max_tokens for CoT prompt (#41979)
• fix: [AMD][DI][CI] Fix four failures in the MI355X disaggregation nightly (#41025)
• fix: [AMD][V4.1][*/N] Fix shared-expert fusion accuracy and speed up MoE routing on ROCm (#42011)
• fix: [BugFix] Pass token-major Q/K tensors from Gemma-3 to RadixAttention (#37984)
• fix: [Bugfix] Fix DeepSeek V4 Pro TP24 vocabulary-padding failure on Hopper (#31801)
• fix: [Bugfix] Llama4 local attention: read page ids from the graph tables in CUDA-graph capture/replay (page_size > 1) (#38187)
• fix: [Docs] Keep the last CUDA 12 image tag fixed during install version bumps (#42201)
• fix: [Fix] Add name mapping in load_weights of nvidia/LocateAnything-3B (#41059)
• fix: [Fix] Allocate SWA KV pools inside the memory-saver region (#40979)
• fix: [Fix] EXAONE MoE under DP attention and DeepEP (#42305)
• fix: [Fix] EXAONE and ERNIE 4.5 VL MoE architecture, backend and PP issues (#42303)
• …and 280 more

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

sgl-project/sglang was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.