This repository contains vllm-hust documentation, sync notes, and fork-specific operation guides.
- Upstream sync notes and decision logs.
- Fork-only deployment and hardware compatibility guides.
- Workspace-level developer playbooks.
- Keep runtime code changes in
vllm-hustminimal and upstream-merge-safe. - Move process and documentation changes here whenever possible.
- Track compatibility matrices and migration notes explicitly.
presentations/— reusable introduction decks, overview slides, and externally reusable presentation assets.meetings/— internal meeting materials, decks, talk tracks, and action lists.sync-notes/— upstream sync notes and merge records.operations/— deployment, runtime, and hardware operation guides.architecture/— source code deconstruction, module maps, and internal design notes.
operations/extension-author-guide.md— 从分类、manifest、实现和打包,到安装、启停、卸载、测试、发布及回滚的完整扩展作者指南。operations/bidkv-packaging-and-release-guide.md— 以 BidKV 为例的插件打包、构建、发布到 PyPI 与安装启用的完整实操指南(uv 构建/发布与命令示例)。operations/optimization-repository-playbook.md— a reusable playbook for optimization repositories that pin runtime dependencies with submodules, manage feature branches, preserve experiment provenance, and decide when upstream PRs are appropriate.architecture/vllm-hust-source-deconstruction.md— a layered source walkthrough of thevllm-hustfork, including request flow, engine structure, execution path, and fork-specific extension surfaces.architecture/vllm-hust-engine-execution-chain.md— a deeper walkthrough of the request-to-engine-to-worker execution path.architecture/vllm-hust-platform-plugin-chain.md— a focused breakdown of platform detection, plugin activation, and fork-safe hardware extension points.architecture/vllm-hust-multimodal-agi4s-capabilities.md— a focused map of multimodal, reasoning, tool-calling, and AGI4S-facing capability modules.architecture/vllm-hust-source-call-flow.md— detailed source call-flow diagrams for CLI serve startup and OpenAI request handling.