界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
效率与协作 · SKILL.md

backends

backends

【效率与协作】Select, register, and troubleshoot Fugue execution backends and plugin packages.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: backends description: "Select, register, and troubleshoot Fugue execution backends and plugin packages." metadata: disco-role: operating disable-model-invocation: true license: Apache 2.0

backends

Use this sub-skill for execution-engine and backend-package questions.

Covers

  • make_execution_engine(...), engine_context(...), and engine resolution rules
  • register_execution_engine(...), register_sql_engine(...), and the default-engine registration helpers
  • backend plugin packages and aliases for Native, DuckDB, Dask, Spark, Ray, Ibis, and Polars
  • backend import side effects, object inference, and config-specific runtime failures

Excludes

  • Workflow DAG mechanics and dataframe transformations, which belong in ../workflow/
  • FugueSQL syntax, YIELD, and PRINT, which belong in ../sql/
  • Notebook extension setup and %%fsql, which belong in ../notebook/

Read these files

  • references/backend-reference.md for the backend matrix, alias map, and resolution order
  • references/troubleshooting.md for optional backend failures and service-specific issues
  • scripts/smoke_backends.py for a safe import-and-registration smoke check

Typical user prompts

  • "Why does Fugue not recognize my Spark/Dask/Ray object?"
  • "How do I select a DuckDB or Spark execution engine?"
  • "How do I register a custom execution engine with Fugue?"
  • "Which Fugue extra do I need for notebook, DuckDB, or Polars support?"

If the task is mainly about writing the dataframe workflow or FugueSQL text that happens to run on a backend, open the workflow or SQL sub-skill first and only come here for engine selection details.

我的收藏