Prefill CP previously replicated NSA/MLA persistent KV on every CP rank, so CP8 consumed eight copies of KV memory while exposing only one rank of logical cache capacity. This change splits logical KV locs from per-rank physical storage, shards MLA latent KV and NSA index K/scale by deterministic page ownership, and keeps existing NSA attention kernels working through a full-view runtime materialization layer. Mooncake PD transfer now sends each prefill CP rank's owned physical pages with explicit logical page positions so non-CP decode can reconstruct full-layout KV. The implementation is guarded by an explicit server flag and startup checks, and the design documentation records the implemented scope, debug environment, and Phase 3 boundary. Constraint: Phase 2 must preserve existing NSA attention/index kernels via runtime full-view materialization Constraint: Decode side remains non-CP and receives full KV through Mooncake Rejected: Shard-aware NSA attention in this change | belongs to Phase 3 because it requires distributed topk/softmax/output contracts Rejected: Request-contiguous CP ownership | unstable under chunked prefill and tied to attention split mode Confidence: medium Scope-risk: broad Directive: Do not enable round-robin CP shared KV without wiring runtime materialization/PD transfer contracts for that split mode Directive: Keep SGLANG_DEBUG_CP_SHARED_KV disabled for perf measurements; it intentionally enables CUDA-syncing diagnostics Tested: Remote py_compile for shared-KV touched Python files in g0034 container Tested: Remote pytest selected cp_shared/shared_kv/nsa suite: 37 passed, 34 deselected Not-tested: Full GLM5 multi-node throughput/regression run after final doc update Not-tested: Phase 3 shard-aware runtime, round-robin CP mode, and non-Mooncake PD backends
SGLang Documentation
This is the documentation website for the SGLang project (https://github.com/sgl-project/sglang).
We recommend new contributors start from writing documentation, which helps you quickly understand SGLang codebase.
Most documentation files are located under the docs/ folder.
Docs Workflow
Install Dependency
Linux:
apt-get update && apt-get install -y pandoc parallel retry
pip install -r requirements.txt
macOS:
brew install pandoc parallel retry
pip install -r requirements.txt
Update Documentation
Update your Jupyter notebooks in the appropriate subdirectories under docs/. If you add new files, remember to update index.rst (or relevant .rst files) accordingly.
pre-commit run --all-filesmanually runs all configured checks, applying fixes if possible. If it fails the first time, re-run it to ensure lint errors are fully resolved. Make sure your code passes all checks before creating a Pull Request.
# 1) Compile all Jupyter notebooks
make compile # This step can take a long time (10+ mins). You can consider skipping this step if you can make sure your added files are correct.
make html
# 2) Compile and Preview documentation locally with auto-build
# This will automatically rebuild docs when files change
# Open your browser at the displayed port to view the docs
bash serve.sh
# 2a) Alternative ways to serve documentation
# Directly use make serve
make serve
# With custom port
PORT=8080 make serve
# 3) Clean notebook outputs
# nbstripout removes notebook outputs so your PR stays clean
pip install nbstripout
find . -name '*.ipynb' -exec nbstripout {} \;
# 4) Pre-commit checks and create a PR
# After these checks pass, push your changes and open a PR on your branch
pre-commit run --all-files
Documentation Style Guidelines
- For common functionalities, we prefer Jupyter Notebooks over Markdown so that all examples can be executed and validated by our docs CI pipeline. For complex features (e.g., distributed serving), Markdown is preferred.
- Keep in mind the documentation execution time when writing interactive Jupyter notebooks. Each interactive notebook will be run and compiled against every commit to ensure they are runnable, so it is important to apply some tips to reduce the documentation compilation time:
- Use small models (e.g.,
qwen/qwen2.5-0.5b-instruct) for most cases to reduce server launch time. - Reuse the launched server as much as possible to reduce server launch time.
- Use small models (e.g.,
- Do not use absolute links (e.g.,
https://docs.sglang.io/get_started/install.html). Always prefer relative links (e.g.,../get_started/install.md). - Follow the existing examples to learn how to launch a server, send a query and other common styles.
Documentation Build, Deployment, and CI
The SGLang documentation pipeline is based on Sphinx and supports rendering Jupyter notebooks (.ipynb) into HTML/Markdown for web display. Detailed logits can be found in the Makefile.
Notebook Execution (make compile)
The make compile target is responsible for executing notebooks before rendering:
- Finds all
.ipynbfiles underdocs/(excluding_build/) - Executes notebooks in parallel using GNU Parallel, with a relatively small
--mem-fraction-static - Wraps execution with
retryto reduce flaky failures - Executes notebooks via
jupyter nbconvert --execute --inplace - Records execution timing in
logs/timing.log
This step ensures notebooks contain up-to-date outputs with each commit in the main branch before rendering.
Web Rendering (make html)
After compilation, Sphinx builds the website:
- Reads Markdown, reStructuredText, and Jupyter notebooks
- Renders them into HTML pages
- Outputs the website into:
docs/_build/html/
This directory is the source for online documentation hosting.
Markdown Export (make markdown)
To support downstream consumers, we add a new Makefile target:
make markdown
This target:
- Does not modify
make compile - Scans all
.ipynbfiles (excluding_build/) - Converts notebooks directly to Markdown using
jupyter nbconvert --to markdown - Writes Markdown artifacts into the existing build directory:
docs/_build/html/markdown/<relative-path>.md
Example:
docs/advanced_features/lora.ipynb
→ docs/_build/html/markdown/advanced_features/lora.md
CI Execution
In our CI, the documentation pipeline first gets all the executed results and renders HTML and Markdown by:
make compile # execute notebooks (ensure outputs are up to date)
make html # build website as usual
make markdown # export markdown artifacts into _build/html/markdown
Then, the compiled results are forced pushed to sgl-project.io for rendering. In other words, sgl-project.io is push-only. All the changes of SGLang docs should be made directly in SGLang main repo, then push to the sgl-project.io.