---

name: report-generation description: "Generate and render single-DataFrame Sweetviz reports safely." disable-model-invocation: true metadata: disco-role: operating license: MIT

Sweetviz report-generation router

Use this sub-skill when the task is to profile one pandas DataFrame with Sweetviz, optionally mark one target feature, then save or display the resulting DataframeReport.

Load order

1. Read references/workflows.md for single-DataFrame analyze() recipes, target/pairwise decisions, HTML/notebook rendering, and output validation.
2. Read references/api-reference.md when exact signatures, source shapes, layout/scale parameters, Comet hooks, or verbosity behavior matter.
3. Read references/troubleshooting.md when report construction, target validation, pairwise analysis, browser/notebook output, Comet, or font warnings fail.
4. Run scripts/sweetviz_smoke_report.py --help or the smoke examples in references/workflows.md when a deterministic local report-generation check is useful.

Scope

This sub-skill covers:

• sweetviz.analyze(source, target_feat=None, feat_cfg=None, pairwise_analysis='auto') for one DataFrame; source may be a DataFrame or [df, "Display name"] / (df, "Display name").
• DataframeReport.show_html(filepath='SWEETVIZ_REPORT.html', open_browser=True, layout='widescreen', scale=None).
• DataframeReport.show_notebook(w=None, h=None, scale=None, layout=None, filepath=None, file_layout=None, file_scale=None).
• Optional DataframeReport.log_comet(experiment) and Sweetviz auto-Comet logging only when the external comet_ml package and credentials/service are configured.
• Target checks, pairwise on/auto/off selection, browser suppression, layout/scale choices, and generated HTML validation.

Route elsewhere

• Two-dataset reports, train/test profiling, and intra-DataFrame split comparisons belong in ../dataset-comparison/.
• Detailed FeatureConfig, type inference, low-cardinality numeric forcing, config overrides, duplicate columns, mixed types, and input preflight belong in ../configuration-and-data-handling/.
• Install/import/package-data issues belong in the root troubleshooting reference ../../references/troubleshooting.md if that file is present.

Critical cautions

• Do not pass verbosity to public sweetviz.analyze() or sweetviz.compare(); those public functions do not accept it in the verified installed API. Use config defaults or DataframeReport(...) only when verbosity control is required.
• Targets must exist, must not be skipped, must have no missing values, and must be boolean or numeric after Sweetviz type detection/forcing.
• pairwise_analysis='auto' can print a threshold warning and return before building a usable full report on wide DataFrames. Choose pairwise_analysis='off' for fast deterministic reports or 'on' only when the user explicitly wants pairwise associations and accepts the cost.
• Use open_browser=False for scripts, CI, remote servers, and agent runs unless the user explicitly asks to launch a browser.