Back-merge master into dev (conflicts resolved)

[QuiltSimon]

Resolves the conflicts blocking #5242 (and #5213 before it). Supersedes #5242 — that PR is the workflow's automated attempt with the conflicts unresolved.

dev was 15 commits behind master with master no longer an ancestor of it, so every later dev→master merge would conflict on every file both branches touched. This restores the ancestry.

The five conflicts and how each was resolved

App.jsx — took master. It renames DataProductsRoute → DataProductsGate and wraps it in a Suspense boundary (#5209's fix); a strict superset of dev's version.

PackageFilters.tsx, model.ts, model.spec.ts — took master. This is the metadata-ordering feature that existed on both sides:

• dev had catalog: order the package-metadata filter list, with a reader-chosen order #5211 (two controls: an "Order by" select plus a direction select)
• Promote dev to master: front door, volumes delight, data-product fixes #5216 promoted catalog: order the package-metadata filter list, with a reader-chosen order #5211 to master, then catalog: sort the package metadata list from one control, next to the list #5222 refined it there into one combined "Sort by" control

So master's version contains and supersedes dev's — 54 lines vs dev's 56 from the same base. @fiskus confirmed this read.

Why this does not discard @kevinemoore's Wave 2 work, which was the open question: his ee84a1e53 ("adopt PackageOrdering unified-ordering contract") and 739560dd0 ("decouple sort-eligibility from facet render type") rewrite model.ts heavily but touch zero FACET_ORDER* lines. Inside the four conflict hunks, the only mentions of Wave 2 identifiers are two prose comments referencing PRESET_ORDERINGS. The conflict is entirely the FACET_ORDER* UI question; Wave 2 sits outside every hunk and is intact in the merged tree (orderingToResultOrder ×4, the Ordering type, PRESET_ORDERINGS).

Why not resolve by hand: git's auto-merge of PackageFilters.tsx produced a file that would have compiled and passed tests while rendering duplicate sort controls — two ordering.offered blocks and two FacetsFilteringState.match blocks, with master's single "Sort by" reduced to dead code and its constants unused. That is why this needed a deliberate resolution rather than accepting the merge.

CHANGELOG.md — both sides' entries kept, master's above dev's, per merge-time ordering.

Verification

• TZ=UTC npx tsc --noEmit -p . — clean
• TZ=UTC npx vitest run (full catalog suite) — 174 files, 1637 passed, 1 skipped
• npx oxfmt --check on the changed dirs — clean
• npm run gql:generate — runs successfully, so every query validates against the merged schema

Ancestry checks: origin/master is now an ancestor of this merge; both Wave 2 commits remain ancestors; Alexei's theme/NavBar work and the data-products commits are all preserved.

⚠️ One thing deliberately left undone

gql:generate regenerates app/model/graphql/{schema,types}.generated.ts with 1566 deletions — it wants to strip the entire data-products GraphQL contract, because master's #5223 removed that contract from the schema while dev still carries the generated artifacts.

I reverted that output rather than commit it here. It is a real inconsistency in dev (generated files disagreeing with their own schema), but deleting a contract hinges on whether enterprise #1125 — the registry half of data-products — targets this release or the next. @kevinemoore, that question is still open. Nothing here depends on the answer: the tree builds and all tests pass either way, and the generated files are not hand-edited.

Worth doing as its own change once #1125's target is known.

🤖 Generated with Claude Code

Greptile Summary

This back-merge combines master’s recent Python SDK, Catalog UI, search, preview, volume-list, workflow, and GraphQL-contract changes into dev. The merge preserves the newer unified metadata-ordering UI and route loading boundary, but leaves generated Catalog GraphQL artifacts inconsistent with the changed shared schema.

• Removes the experimental Python agent-context API and its tests.
• Refines Catalog search ordering, error boundaries, Perspective cleanup, volume rendering, admin bucket controls, and front-door examples.
• Removes the unserved data-products GraphQL schema while retaining the fixture-backed preview UI.
• Updates the automatic master-to-dev back-merge workflow with drift escalation.

Confidence Score: 4/5

The pull request should not merge until the generated Catalog GraphQL artifacts are reconciled with the removed schema so the required validation job can pass.

Removing the data-products schema while retaining generated types and introspection guarantees that Catalog code generation changes the working tree, and the required CI job rejects that state.

Files Needing Attention: shared/graphql/schema.graphql, catalog/app/model/graphql/schema.generated.ts, catalog/app/model/graphql/types.generated.ts

Important Files Changed

Filename	Overview
shared/graphql/schema.graphql	Removes the complete data-products service contract without updating generated Catalog artifacts, causing the codegen clean-tree gate to fail.
catalog/app/model/graphql/types.generated.ts	Remains unchanged and still contains types for the removed data-products schema.
catalog/app/model/graphql/schema.generated.ts	Remains unchanged and still exposes removed data-products fields through generated introspection.
catalog/app/containers/Search/model.ts	Adds the combined facet-ordering contract and safer malformed-filter parsing with corresponding tests.
catalog/app/utils/perspective.ts	Strengthens asynchronous Perspective resource cleanup and routes load failures through React error boundaries.
.github/workflows/back-merge-dev.yaml	Adds bounded drift escalation when an existing conflict-resolution pull request remains open.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Schema[shared GraphQL schema] --> Generate[npm run gql:generate]
  Generate --> Types[Catalog generated types]
  Generate --> Introspection[Catalog generated introspection]
  Types --> Clean{Git tree clean?}
  Introspection --> Clean
  Clean -->|Yes| RemainingCI[Lint, typecheck, tests, build]
  Clean -->|No| Failure[Catalog CI fails]

Loading

_{Reviews (1): Last reviewed commit: "Merge remote-tracking branch 'origin/mas..." | Re-trigger Greptile}

Greptile also left 1 inline comment on this PR.

Context used (5)

• Knowledge Base — Python SDK and GraphQL client
• Knowledge Base — Python SDK package operations
• Knowledge Base — Catalog client features
• Knowledge Base — Catalog server and build runtime
• Knowledge Base — API contracts and schemas

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.09%. Comparing base (28dbd70) to head (4467e5b).
⚠️ Report is 11 commits behind head on dev.

Additional details and impacted files

@@             Coverage Diff             @@
##              dev    #5252       +/-   ##
===========================================
+ Coverage   54.39%   92.09%   +37.70%     
===========================================
  Files         871      129      -742     
  Lines       36406    12351    -24055     
  Branches     6420        0     -6420     
===========================================
- Hits        19804    11375     -8429     
+ Misses      14841      976    -13865     
+ Partials     1761        0     -1761     

Flag	Coverage Δ
api-python	93.64% <100.00%> (-0.19%)	⬇️
catalog	?
lambda	85.37% <ø> (ø)
py-shared	98.02% <ø> (ø)

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.

[greptile-apps]

Generated GraphQL contract is stale

When Catalog CI runs, gql:generate removes the data-product declarations that remain in the checked-in generated types and introspection, causing the required git diff --exit-code check to fail.

Knowledge Base Used:

• API contracts and schemas
• Catalog server and build runtime
• Build, test, and deployment automation

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!