vizro
Use Vizro to build, validate, debug, extend, and maintain dashboards across vizro-core, Vizro-MCP, deprecated Vizro-AI, experimental chat, and Vizro Dash Components.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- vizro
- license
- Apache 2.0
- metadata
- { "disco-role": "operating", "source-repo": "mckinsey/vizro", "source-commit": "99634b8e837d371f0d25c53692278b39236594e6" }
- description
- Use Vizro to build, validate, debug, extend, and maintain dashboards across vizro-core, Vizro-MCP, deprecated Vizro-AI, experimental chat, and Vizro Dash Components.
- disable-model-invocation
- true
02 资源文件 8
- 参考文档references/environment.md4.5 KB
- 参考文档references/package-map.md4.9 KB
- 参考文档references/repo-provenance.md2.8 KB
- 参考文档references/repo-routing-metadata.json398 B
- 参考文档references/verification.md4.2 KB
- 脚本scripts/inspect_vizro_schema.py2.1 KB
- 脚本scripts/probe_vizro_environment.py4.6 KB
- 脚本scripts/validate_vizro_dashboard.py2.2 KB
03 SKILL.md 正文
Vizro
Use this skill when a task involves the Vizro monorepo or the installed Vizro packages: dashboard authoring, model/YAML validation, charts/tables/layouts/actions, MCP-assisted dashboard generation, the deprecated vizro-ai chart agent, experimental chat/popup components, or vizro-dash-components custom Dash components.
Start Here
- Read
references/repo-provenance.mdto confirm the source snapshot, evidence scope, and freshness limits. - Read
references/package-map.mdto identify the relevant package, public API surface, and known version-specific facts. - Read
references/environment.mdbefore installing packages, buildingvizro-dash-components, running tests, using browser-backed examples, or touching provider-backed LLM features. - Run
python scripts/probe_vizro_environment.pyin the target environment for a quick installed-package smoke check. - Use the route map below. Each sub-skill owns task-specific workflows and near references.
- When modifying this repository, also route to
sub-skills/repo-development-testing/SKILL.mdfor package-specific commands, backend-classified tests, and AGENTS constraints.
Route Map
| User task | Use sub-skill | Why |
|---|---|---|
Build/run a Vizro dashboard in Python or YAML; validate Dashboard/Page/layout basics |
sub-skills/core-dashboard-build/SKILL.md |
Owns Vizro().build(), Vizro().run(), dashboard/page models, Python/YAML authoring, layout-first debug patterns. |
| Add or debug cards, graphs, tables, filters, selectors, data sources, custom components, or actions | sub-skills/core-components-data-actions/SKILL.md |
Owns Vizro model composition, data_manager, action wiring, controls, table/grid behavior, and validation failures. |
| Choose or author charts/figures/tables with Plotly/Vizro conventions | sub-skills/charts-and-figures/SKILL.md |
Owns vizro.plotly.express, custom chart callables, visual-vocabulary guidance, and figure model pitfalls. |
| Use or change Vizro-MCP agent/server workflows | sub-skills/mcp-agent-workflows/SKILL.md |
Owns FastMCP server tools/resources, dashboard creation/validation helpers, schema extraction, and PyCafe workflows. |
Inspect or maintain the deprecated vizro-ai chart agent |
sub-skills/legacy-ai-chart-agent/SKILL.md |
Owns chart_agent, response models, safeguard behavior, and deprecation-safe routing. |
Use or modify vizro-experimental chat component or floating popup |
sub-skills/experimental-chat-popup/SKILL.md |
Owns Chat, add_chat_popup, lazy agent helpers, BYO callback mode, and security constraints. |
Use/maintain vizro-dash-components Cascader/Markdown components or generated wrappers |
sub-skills/dash-components-build/SKILL.md |
Owns TypeScript source-of-truth, npm/webpack/dash-generate-components pipeline, props, and browser-test constraints. |
| Change repository code, docs, tests, build config, or package metadata | sub-skills/repo-development-testing/SKILL.md |
Owns monorepo layout, Hatch commands, package-local test commands, changelog, and backend-classified verification. |
Public Package Facts
-
vizro-coreinstalls the public distributionvizro. Public Python examples typically use:import vizro.models as vm from vizro import Vizro dashboard = vm.Dashboard(pages=[vm.Page(title="Home", components=[vm.Card(text="Hello")])]) app = Vizro().build(dashboard) # returns the Vizro wrapper, not the raw Dash app app.run()The underlying Dash app is
app.dashafter build. -
vizro.plotly.expressprovides Vizro-oriented wrappers around Plotly Express; use it when authoring standard Vizro charts unless the task requires custom Plotly code. -
vizro-mcpis the preferred maintained agent-assistance path for new AI/dashboard workflows. -
vizro-aiis deprecated in this checkout. Maintain existing chart-agent behavior when requested, but do not recommend it for greenfield workflows when Vizro-MCP/e2e-flow is a better fit. -
vizro-experimentalAPIs may change or graduate intovizro-core; state that caveat in user-facing solutions. -
vizro-dash-componentssource of truth is TypeScript/React undersrc/ts. Python files and JS bundles undervizro_dash_components/are generated; do not hand-edit them.
Safety Defaults
- Do not run live LLM provider calls unless the user explicitly supplies/authorizes credentials and network use. Prefer import, schema, or BYO callback checks.
- Do not require browser-backed
dash_duo/Selenium/e2e tests on hosts without Chrome/Chromium. Treat them as optional backend-specific gates and document the missing backend. - Do not run full monorepo test suites first. Start with package-local focused tests for the touched behavior; then broaden if needed.
- Do not edit generated
vizro_dash_components/*.pyor JS bundles directly. Edit TypeScript invizro-dash-components/src/tsand regenerate. - For repo edits, obey package-local
AGENTS.md/CLAUDE.md: Hatch commands run from the specific package directory;vizro-e2e-flowis a Claude plugin and its guide says no Hatch commands.
Bundled Helpers
scripts/probe_vizro_environment.pychecks imports and small CPU-only smoke behavior across Vizro packages.scripts/inspect_vizro_schema.pysummarizes public Vizro model fields/schemas for dashboard-writing tasks.scripts/validate_vizro_dashboard.pyvalidates a small Python or YAML dashboard path when the installed package and PyYAML are available.
Common First Moves
- User wants a dashboard: route to
core-dashboard-build, then tocharts-and-figuresandcore-components-data-actionsas components/charts/actions/data appear. - User wants AI-generated or agent-validated dashboards: route to
mcp-agent-workflows; only uselegacy-ai-chart-agentif the task explicitly mentionsvizro-aior existing chart-agent code. - User wants chat in a dashboard: route to
experimental-chat-popup; prefer BYOgenerate_responsefor credential-free examples. - User touches custom Dash components: route to
dash-components-buildandrepo-development-testing. - User touches repository code: load the relevant feature subskill plus
repo-development-testing, then run focused tests with the package-local working directory.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).