[fitzee]

SUMMARY

Refine MCP dashboard-copy safety, list-scope validation, and analytical guidance (implementation and tests at 831598148220a36a37c0d7a4a3890a75bff6ed59):

• Deep-copy charts by default in duplicate_dashboard, including its compact tool description. Explicit duplicate_slices=false deliberately shares charts (test_compact_duplicate_description_keeps_default, test_duplicate_referencing_same_charts).
• Reject unknown list-scope arguments rather than silently ignoring them; explain direct user-editor membership and creator/editor union (test_list_dashboards_rejects_unknown_arguments, test_list_dashboards_self_lookup_and_aliases).
• Distinguish sampled output from input coverage. Ungrouped pivots retain ordinary completeness checks; row-collapsing pipelines conservatively report total_rows=null and truncated=true. Each layer uses its executed query limit, and longer completeness guidance appears once per dashboard (test_pivot_completeness_respects_groupby, test_multi_query_completeness_uses_each_executed_limit, test_dashboard_completeness_guidance_is_shared).
• Use neutral number-format examples and evidence-based currency guidance (test_number_format_examples_are_currency_neutral, test_guided_chart_prompt_does_not_assume_currency).

Compatibility changes are documented in UPDATING.md and the MCP guide. No frontend or migration changes.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

Not applicable: MCP contracts, guidance, and backend tests only.

TESTING INSTRUCTIONS

Validation at 831598148220a36a37c0d7a4a3890a75bff6ed59:

• pytest across the touched MCP test modules and native surface selection: 906 passed. Command: PYTHONPATH=$PWD/superset-core/src:$PWD /tmp/pr45504-pivot-venv/bin/python -m pytest <touched-test-files> -m 'not manual and not llm_compliance' -q (11 files total, including native surface).
• Targeted final regression set against the pre-follow-up source: 13 failed, 7 passed; with the follow-up: 20 passed. Preservation cases retain grouped-pivot uncertainty and production distribution-pipeline behavior.
• Native inventory regenerated using SUPERSET_MCP_UPDATE_TOOL_INVENTORY=1 pytest tests/unit_tests/mcp_service/test_native_tool_surface.py: 38 passed.
• Staged-file and all changed-branch-file pre-commit checks passed, including mypy, Ruff, and custom Pylint: pre-commit run and pre-commit run --files <changed-branch-files>.
• Real-LLM checks not run: OpenRouter disabled; no real provider API calls.

Manual checks:

1. Duplicate without duplicate_slices; expect independent chart IDs. Explicit false intentionally shares IDs.
2. Inspect the compact tool description for the deep-copy default and explicit sharing option.
3. Fetch an ungrouped 12-point timeseries under a 100-row limit; expect total_rows=12, truncated=false. Grouped pivots retain conservative input-coverage reporting.
4. Request more rows than the server permits; inspect the executed fetch limit and per-layer completeness.

ADDITIONAL INFORMATION

Omitting duplicate_slices creates independent chart objects. Unknown list arguments fail validation. The response metadata is additive, but total_rows and truncated have a clarified meaning: unknown input coverage is conservative uncertainty, not observed truncation. Histogram computation and authorization are unchanged. Propagating raw sql_rowcount for exact input-cap detection remains a separate follow-up.

• Has associated issue
• Required feature flags
• Changes UI
• Includes DB Migration
• Introduces new feature or API
• Removes existing feature or API

[netlify]

✅ Deploy Preview for superset-docs-preview ready!

Name	Link
🔨 Latest commit	8315981
🔍 Latest deploy log	https://app.netlify.com/projects/superset-docs-preview/deploys/6ad8bcbee2159cdcafc5af47
😎 Deploy Preview	https://deploy-preview-45504--superset-docs-preview.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[codeant-ai-for-open-source]

CodeAnt PR Risk: Low Risk

• The PR appears safe to merge; no actionable issues were identified in the review.

Assessed commit: 831598148220

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.05%. Comparing base (4e7e2b7) to head (8315981).
⚠️ Report is 13 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #45504      +/-   ##
==========================================
+ Coverage   83.98%   84.05%   +0.06%     
==========================================
  Files        3055     3054       -1     
  Lines      201017   200913     -104     
  Branches    46949    46950       +1     
==========================================
+ Hits       168828   168875      +47     
+ Misses      28943    28800     -143     
+ Partials     3246     3238       -8     

Flag	Coverage Δ
hive	34.92% <48.71%> (+0.03%)	⬆️
mysql	52.88% <48.71%> (+0.02%)	⬆️
postgres	52.89% <48.71%> (+0.03%)	⬆️
presto	36.65% <48.71%> (+<0.01%)	⬆️
python	87.32% <100.00%> (+0.12%)	⬆️
sqlite	52.65% <48.71%> (+0.02%)	⬆️
unit	81.57% <100.00%> (+0.16%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[gabotorresruiz]

Thanks @fitzee, this is a good set of fixes and the four behaviours all hold up when I drive them.

I ran the affected suites plus the duplication, ownership filter and histogram paths against a real instance on this branch, and reverted each behaviour one at a time to confirm the cited tests pin it. Everything in the body checks out, including the merge base count. Comments inline, none of them blocking.

Holding off on a formal approval only because unit-tests is red and the branch is conflicting. The failure is test_native_inventory_report_matches_registered_tools and it reproduces unchanged at the merge base 3cf94c87, so it is inherited rather than yours: master already refreshed those generate_chart, update_chart and get_chart_preview byte counts in #45311, which is also the conflict. A catch up to master plus a regenerated inventory should clear both.

[rebenitez1802]

Approve: the deep-copy default, the strict self-scoped list schemas and the currency-neutral guidance all look right, and the authorization model is unchanged. Duplicating still needs Dashboard write plus editor on the source, the same capability as REST /copy/. I left two Medium inline suggestions worth taking before merge:

1. pivot without a groupby is treated as row-collapsing, so most timeseries charts now report truncated=true.
2. The compact tool description no longer mentions the new deep-copy default.

Smaller follow-ups (non-blocking):

• 🟢 fetch_row_limit reports request.fetch_row_limit, not the limit that actually ran (query_context.queries[i].row_limit). They differ when SQL_MAX_ROW is lower than the request, and truncated can then be wrong too.
• 🟢 The histogram/box-plot note branch only fires when operation metadata is missing. It then returns total_rows=1, truncated=false next to a note saying those values prove nothing. The two tests for it build ChartData without metadata, so they cover that fallback, not the production path.
• 🟢 Stale descriptions on fetch_row_limit / total_rows in dashboard/schemas.py (they describe only a "capped fetch"), plus a capital "For" after a semicolon in the completeness note. The meaning change for truncated/total_rows deserves a line in UPDATING.md, since the PR calls the new fields "additive".
• 🟢 A few $,.2f examples remain: chart/schemas.py Big Number y_axis_format and d3_number_format, and update_dataset_metric.py (sum_revenue with $,.2f).
• 🟢 The ~370-char completeness_note repeats on every chart (~11 KB at max_charts=30). It could go once on DashboardData.