免费在 GitHub 公开

data-juicer

Data-Juicer repo router for local recipes, Ray recovery, and service/MCP workflows.

中文简介正在准备中。当前显示原文说明。

VectorSpaceLab/AREX-Skill

查看安装方法

包含的文件（28）

SKILL.md2.0 KB
references/api-reference.md1.2 KB
references/cli-reference.md956 B
references/install-and-scope.md2.1 KB
references/optional-extras.md1.5 KB
references/package-overview.md1.2 KB
references/repo-provenance.md2.2 KB
references/repo-routing-metadata.json486 B
references/troubleshooting.md1.4 KB
references/workflow-map.md809 B
scripts/check_environment.py2.2 KB
sub-skills/ray-and-recovery/references/checkpointing.md1.2 KB
sub-skills/ray-and-recovery/references/job-management.md713 B
sub-skills/ray-and-recovery/references/troubleshooting.md918 B
sub-skills/ray-and-recovery/references/workflows.md1.0 KB
sub-skills/ray-and-recovery/scripts/ray_smoke.py953 B
sub-skills/ray-and-recovery/SKILL.md1.7 KB
sub-skills/recipes-and-ops/references/cli-reference.md826 B
sub-skills/recipes-and-ops/references/configuration.md1.4 KB
sub-skills/recipes-and-ops/references/troubleshooting.md1.1 KB
sub-skills/recipes-and-ops/references/workflows.md1.3 KB
sub-skills/recipes-and-ops/SKILL.md1.7 KB
sub-skills/service-mcp/references/api-service.md1.4 KB
sub-skills/service-mcp/references/mcp.md772 B
sub-skills/service-mcp/references/operator-discovery.md944 B
sub-skills/service-mcp/references/troubleshooting.md1.0 KB
sub-skills/service-mcp/scripts/start_api_service.py6.0 KB
sub-skills/service-mcp/SKILL.md1.7 KB

SKILL.md（原文）

安装前可以查看交给 AI 代理的指令内容。

Data-Juicer Repo Skill

Use this skill as the top-level router for the Data-Juicer repository. Keep workflow depth, long examples, and troubleshooting details in the sub-skills and shared references.

Start with the route

User need	Read
Local recipe processing, dataset config, export, operator choice, or simple analysis	`sub-skills/recipes-and-ops/SKILL.md`
Ray execution, partitioning, checkpointing, job monitoring, or recovery	`sub-skills/ray-and-recovery/SKILL.md`
FastAPI service, MCP server, operator search, or request encoding	`sub-skills/service-mcp/SKILL.md`

Quick environment check

Install guidance lives in references/install-and-scope.md; start with py-data-juicer and add tools or distributed only when the selected route needs them.

Use the bundled smoke check when you only need a safe install/import sanity check:

python scripts/check_environment.py

Useful CLI help checks:

dj-process --help
dj-analyze --help
dj-install --help
dj-mcp --help

Shared references

references/repo-provenance.md
references/repo-routing-metadata.json
references/install-and-scope.md
references/package-overview.md
references/cli-reference.md
references/api-reference.md
references/optional-extras.md
references/troubleshooting.md
references/workflow-map.md

Scope boundaries

Do not mix local recipe guidance with Ray recovery details.
Do not mix HTTP/MCP request transport with local dataset semantics.
Do not assume optional extras are installed unless the workflow needs them.
Do not depend on the source checkout remaining present once the skill is created.

When in doubt

Start from references/workflow-map.md, then open the matching sub-skill.

评价

还没有评价。欢迎分享你的使用感受。

无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

同一仓库的技能

3d-resnets-pytorch

免费

Routes 3D ResNets PyTorch video action-recognition workflows across training, inference, and data preparation.

中文简介正在准备中。当前显示原文说明。