[bot] Merge master/b57834ce into rel/dev - #1833
Conversation
The expectation's `type` selects everything that differs: `dashboard` reads the drafted document, `dashboardPatch` applies the streamed patch to the base it shipped with and scores the result. An edit that routes to the builder, or a creation answered by the editor, fails as the routing failure it is. `date_range` and `filters` are opt-in. A fixture that states neither is not checked on them, which is what lets an edit case assert a rename without claiming anything about the range the saved dashboard already had. A check the case cannot apply is not published at all. Publishing it as passed lifts quality_score -- the fraction of true booleans in the detail -- above what the run earned, and would make a creation case read better than the same case scored before editing existed. A patch carrying no operations fails. jsonpatch applies an empty list happily and hands the document back untouched, so the run would otherwise score the saved dashboard and pass every case the saved dashboard already satisfied. The date range is checked per document shape. A drafted dashboard carries one filter per tab and every tab must have it; a saved dashboard can hold a dashboard-wide range and a dataset-scoped one with nothing to tell them apart, so one matching entry satisfies the check there. A run that never routed skills passes the skill gate. gen-ai omits set_skills for a pinned agent, and a run continuing a conversation can have routed in a turn it never saw -- absence of the call is absence of evidence, not a mis-route. Verified against the eight-item Langfuse dataset on e2e-llm-static: five creation and two editing cases pass, including the shape and opt-in changes above. jira: QA-29477 risk: nonprod Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
feat(gooddata-eval): score dashboard editing alongside creation
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Advanced Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## rel/dev #1833 +/- ##
===========================================
+ Coverage 83.08% 83.19% +0.11%
===========================================
Files 330 330
Lines 21763 21931 +168
===========================================
+ Hits 18082 18246 +164
- Misses 3681 3685 +4 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
🚀 Automated PR to perform merge from master into rel/dev with changes up to b57834c (created by https://github.com/gooddata/gooddata-python-sdk/actions/runs/36407179018).