feat: carry each chart mapping's framework

[jfrench9]

Summary

Picks up robosystems#1608, which anchors each chart mapping to a mapping taxonomy the chart owns and records the framework it maps into. The regenerated models carry the new optional field, and LedgerClient now returns each mapping's framework.

Release order: publish only after robosystems#1608 is deployed. list_mappings and list_structures now select framework, and a server without that field rejects the query.

Changes

• Regenerated REST models (robosystems_client/models/): TaxonomyBlockStructureRequest and TaxonomyBlockStructure gain an optional target_framework, used only on coa_mapping structures; omitted means the book framework. The regen picked up only this change, with no other drift since the last generation.
• GraphQL (robosystems_client/graphql/):
  • schema.graphql is refreshed; the Structure type gains framework.
  • The ListLedgerMappings and ListLedgerStructures operations select it.
  • generated/ is regenerated from those operations.
• Facade (clients/ledger_client.py):
  • list_mappings and list_structures return framework.
  • The list_mappings docstring says the book mapping is listed first.
  • Both are sync-only facade methods; there is no async twin to cover.
• Tests (tests/test_ledger_client.py): the two fixtures carry framework, and list_mappings asserts it.

Compatibility

ADDITIVE: a client minor. The new REST model fields are optional, and framework is a new, nullable field on two facade return types. No export, signature or existing field changed.

Deploy-order dependency: the two facade queries request a field that only exists once robosystems#1608 is deployed. Against an older server they fail GraphQL validation, so this has to be published after that deploy.

Testing

• Regenerated with just generate-sdk against the local stack running robosystems#1608's branch. The branch was current with main, and the new fields were confirmed in the live /openapi.json and GraphQL schema first.
• just generate-graphql was run after the operation edits.
• just test-all: 595 passed, 17 skipped; ruff and format clean; basedpyright 0 errors.
• just build-package was not run (no packaging change).

🤖 Generated with Claude Code

[jfrench9]

@claude please review this PR

[claude]

Claude finished @jfrench9's task in 58s —— View job

---

Review complete

• Gather context (PR description, changed files)
• Read repo CLAUDE.md for conventions (none exists in this repo)
• Review diff against origin/main
• Check regenerated models/GraphQL for consistency
• Check facade + tests
• Post review feedback

Verdict: LGTM. This is a clean, well-scoped additive change.

What I checked:

• schema.graphql / Structure.framework is correctly documented as coa_mapping-only and nullable, matching ListLedgerMappings.graphql and ListLedgerStructures.graphql, which are the only two operations that select the shared id name description blockType taxonomyId isActive field set — no other operation was missed.
• The generated Pydantic models (list_ledger_mappings.py, list_ledger_structures.py) add framework: Optional[str] with no alias, consistent with the existing description: Optional[str] field — codegen output looks correctly regenerated, not hand-edited.
• TaxonomyBlockStructure / TaxonomyBlockStructureRequest REST models gain target_framework as None | str | Unset, following the exact to_dict/from_dict pattern used for the neighboring role_uri field. Consistent and additive.
• ledger_client.py: list_mappings docstring now documents both the framework field and the "book mapping first" ordering — worth double-checking that claim is actually guaranteed server-side (robosystems#1608) since nothing here enforces or tests that ordering; if it's not a hard guarantee it may be worth softening the wording.
• Tests: both fixtures (list_mappings, list_structures) were updated with framework, and only list_mappings asserts on the new value, which matches the facade surface described in the PR body. The incidental taxonomyId fixture change (tax_usgaap → tax_map_gaap) in test_list_mappings_unwraps_structures is harmless — nothing else in the test depends on that value.
• Confirmed no async twin exists for list_mappings/list_structures in ledger_client.py, so the "sync-only, no async twin" claim in the PR body checks out.
• Deploy-order dependency is clearly called out in the PR body; nothing in the diff undermines that (both new GraphQL selections are unconditional, so they will indeed fail validation against a server without the field, as stated).

Nit (non-blocking): the list_mappings docstring's "book mapping first" claim reads as an API guarantee but isn't covered by a test — consider adding an assertion (or at least a comment on the mock ordering) once robosystems#1608 lands, or softening the docstring if ordering isn't contractually guaranteed.

Note: I was unable to run just test-all/ruff/basedpyright in this sandbox (shell commands required approval that wasn't available), so I relied on static review of the diff. The PR description states just test-all passed locally (595 passed, 17 skipped; ruff/format/basedpyright clean), which is consistent with what I see in the diff.