L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Données et recherche · SKILL.md

contexts-and-configuration

contexts-and-configuration

Choose and configure Great Expectations Data Contexts, metadata stores, credentials, Data Docs, analytics, and import sanity checks.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: contexts-and-configuration description: "Choose and configure Great Expectations Data Contexts, metadata stores, credentials, Data Docs, analytics, and import sanity checks." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Contexts and Configuration

Use this sub-skill when a task needs a Great Expectations (GX Core) Data Context, project configuration, safe credential substitution, metadata-store/Data Docs settings, analytics toggles, or a quick import/context sanity check.

Start Here

  1. Choose the context deliberately: gx.get_context(mode="ephemeral") for temporary in-memory work, gx.get_context(mode="file", project_root_dir=...) or context_root_dir=... for persisted projects, and avoid Cloud mode in this GX Core line because Cloud construction raises a shutdown error.
  2. Keep context setup separate from downstream work: route data connections to ../datasources-and-assets/SKILL.md, expectation suites to ../expectations-and-suites/SKILL.md, validation runs to ../validations-and-results/SKILL.md, and checkpoint/Data Docs action orchestration to ../checkpoints-actions-and-data-docs/SKILL.md.
  3. For file-backed projects, edit context.variables.config, call context.variables.save(), and re-initialize the context so stores, Data Docs sites, analytics, and substitutions take effect.
  4. Never hardcode credentials in skill outputs or committed project config; use environment variables, config_variables.yml, runtime_environment, or supported secret-manager references.

Reference Map

  • Context types and stores: read when choosing gx.get_context arguments, diagnosing file-vs-ephemeral persistence, or understanding context managers and stores.
  • Configuration and credentials: read when configuring DataContextConfig, config variables, environment substitutions, store paths, Data Docs sites, or analytics toggles.
  • Troubleshooting: read when imports fail, the wrong context type appears, config variables do not substitute, Cloud mode is triggered, or store/Data Docs paths behave unexpectedly.
  • Inspect context script: run for a safe local smoke check of GX import/version, gx.get_context, context mode/type, managers, store names, and optional file-context scaffolding.

Safe Inspection

From this sub-skill directory, run the bundled helper with no arguments for an ephemeral context sanity check:

python scripts/inspect_context.py

From the root great-expectations/ skill directory, use the full sub-skill path:

python sub-skills/contexts-and-configuration/scripts/inspect_context.py

Run it against a disposable or intended file project root when you need to prove file-mode scaffolding/loading works:

python sub-skills/contexts-and-configuration/scripts/inspect_context.py --mode file --project-root-dir ./gx-project

The helper prints a compact JSON summary and does not print absolute project paths or credentials.

Skills enregistrées