0.1.0 shakedown: fold the pending changes into the entry, demo the solver and views, and expose an element once

[devin-ai-integration]

Summary

The release notes carried two sections for one unreleased version — ## Unreleased and ## 0.1.0 — 2026-08-17 — and nothing has been tagged, so both are the same release. This folds them into one dated ## 0.1.0 — 2026-08-18 entry, leading with the rename (it decides how every other entry spells things), and reconciles the statements the merge made false.

Then, shaking the release down against that entry, two of its claims wanted demos and one turned out to be a bug, so this PR also carries the fix and the two examples:

• fix(view): a wildcard or filtered expose showed an element once per scope tree it was reached through. A document's own scope tree and the global index build a *symbols.Symbol each for one declaration, so deduping by pointer deduped nothing across them: expose Fleet::**[@Fleet::Safety] rendered part def Fleet::Airbag twice. Identity is now the declaration a symbol was built from, in one place the four callers share:

  // internal/core/symbols/identity.go
  func KeyOf(sym *Symbol) ElementKey   // {DocName, DeclSpan}, or the pointer for a cached library symbol
  func SameElement(a, b *Symbol) bool
  
  -seen map[*symbols.Symbol]bool   // resolve.elementList, semantics.exposedSet
  +seen map[symbols.ElementKey]bool

  internal/lsp/rename.go and internal/core/edit/locate.go each had their own sameSymbol doing the DocName+DeclSpan comparison locally; both now call symbols.SameElement, and view.dedupe is deleted since Model.ExposedElements no longer returns a duplicate. Regression: internal/repl/view_wildcard_test.go asserts one occurrence per element in %view and %render for expose P::* and expose P::**[@T] — it fails on main's behavior (verified by stashing the resolve/semantics change).

• examples/solver-demo.sysml + walkthrough — a rover budget written so each solver command has something to answer: asserted constraints, a requirement whose conditions conflict (%explain names all three), a satisfaction assertion, two interacting variation points (%configure … all permits 3 of 4), and three analyses (one objective, a quantity-valued one, two improved lexicographically). Every command and every output in the walkthrough is transcribed from the built binary under z3, including cvc5's typed refusal of %optimize.

• examples/views-demo.sysml + walkthrough — a lander with one view per rendering kind (tree, interconnection, state, action, table), a nested view, a framed concern whose viewpoint conformance is violated by one exposed part, and a filtered recursive expose. This is the model the duplicate above showed up on.

• an action rendering, not a action rendering — Kind.article(), used by the empty-rendering message and the wrong-form error.

Two features that landed this cycle had no changelog entry at all, and no row in the README capability table. The code and the reference docs are complete, only the release-facing text was missing:

• the SMT solving path — %check, %explain, %solve, %configure, %optimize, the backend capability model and SMT-LIB 2.6 logic selection, the differential and portability gates, and the OPENSYSML_SMT* variables (OPENSYSML_SMT_MAX_CONFIGURATIONS was missing from the rename's variable list too);
• the source-preserving edit path — the ApplyEdits RPC, internal/core/edit, and opensysml's model.edit() / set_value / rename / apply() behind the apply_edits capability.

Reconciled, because within one release a reader never saw the intermediate state:

the 0.1.0 entry said	it now says
pysysml raises ExperimentalFeatureWarning, pysysml.is_experimental	opensysml, as the rename entry above it states
RDF covers structure only, 71 of 120 models convert, 49 refused	structure and behavior, expressions excepted; the counts have one home, in the behavior entry (102/18)
SYSTEMICA_REQUIRE_TRAINING_CORPUS=1 in the tag pipeline	OPENSYSML_REQUIRE_TRAINING_CORPUS=1
known limitation: a model stating behavior is refused by the mapping	closed by the behavior coverage; the limitations that do stand are named individually instead of by reference to 0.0.9's list

The known-limitations list also gained the ones the new entries imply: package-owned bindings, %invoke on a calc/constraint, tool-defined rendering, %optimize needing z3 (optimization is a z3 extension cvc5 does not implement), and the two edit operations offered.

docs/project/releasing.md had stale pre-rename versions in the PyPI runbook: the rehearsal declares VERSION = "0.1.0rc1" and installs opensysml==0.1.0rc1 while tagging opensysml-v0.3.0rc1, and the verification step asserts opensysml.__version__ prints 0.1.0. The client's declared version is 0.3.0 (python/opensysml/_version.py), so following the runbook as written would fail its own check — the examples are now 0.3.0.

Verification

The whole release gate, on this branch with the fix: gofmt -l . (empty), go build ./..., go vet ./..., make lint, go test ./..., both SMT suites (OPENSYSML_REQUIRE_SMT=1 against z3 and cvc5 over ./internal/core/solve ./internal/repl), the corpus gate (OPENSYSML_REQUIRE_TRAINING_CORPUS=1 ./internal/core/model -run TestTrainingExamples), TestExamplesAnalyseCleanly and sysml -validate over both new models, and scripts/check-doc-links.py (0 broken links).

Beyond the gates, the release was shaken down through the REPL and the gRPC service against the changelog's claims — every rendering kind and form, nested and empty views, %print/%features/%slots/%invoke, action and state debugging across declarations, binding propagation with its conflict and cycle errors, named control nodes, math-import gating, %check/%explain under both solvers, feature_values over gRPC with the opensysml client and its deprecated aliases, and the SYSTEMICA_* → OPENSYSML_* migration. The duplicate exposure above was the only defect found; not exercised: %step on a state machine, collection- and enum-valued features.

Open question for the release, not addressed here

docs/project/roadmap.md § T1 says the deprecated spellings (%slots, Instance.slots/SlotValue, raw_slots/get_slot) must be removed before 0.1.0 ships, on the pre-release clean-break rule — while the changelog entry documents keeping them as deprecated aliases. One of the two has to give before the tag; this PR changes neither.

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/c3c88c095dc9480c813ccf30359ed271
Requested by: @HuiJun

---

[devin-ai-integration]

Original prompt from jason.han

@JPL-Devin/OpenSysML Help me do a shakedown for the 0.1.0 release. We need to make sure there are no obvious bugs and that all implemented features and fixes are working now.

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 1 additional finding.