docs(roadmap): re-baseline on v0.9.1 and account for open and in-progress work

[devin-ai-integration]

What and why

docs/project/roadmap.md was still on the v0.8.0 baseline. v0.9.0 (267 PRs) and v0.9.1 (139 PRs) have shipped since then, and nine more PRs have merged to develop. This PR re-baselines the roadmap on v0.9.1 (b9de3b070, #614) and goes through every track again.

• Header and status words. Gives the exact metadata for each tag. Lists the nine PRs merged to develop after v0.9.1 (feat(state-machines): express pseudostates and deferred events as standard metadata #639, refactor(lang): remove the defer state-body extension #694, feat(export): name the standard library elements a converted model references (#731) #746, feat(grpc): Convert takes the id form, as sysml -id does (#732) #749–fix(grpc): report the parser's warnings as the workspace does #751, chore(release): merge main (0.9.1 + Windows hotfixes) back into develop #754–fix(quantity): absorb MeasurementReferences::one as the identity of the unit product #756); those are the only items now marked landed but unreleased. Adds the in progress status for work being implemented that has no PR yet. Replaces the closed-work summaries for 0.6→0.7 and 0.7→0.8 with summaries for 0.8→0.9.0 and 0.9.0→0.9.1.

• Gate table. Re-measured at the tag, with the v0.8.0 figure in brackets:

  • corpus round trips: RDF 351/358 stable plus 7 whitespace-only; api-json 349/358 plus 2 graph-diff;
  • pilot corpora: 9 of 213 files report a diagnostic;
  • stdlib: 106/106 files clean;
  • 1224 execution conformance cases and 554 trace goldens;
  • rejection oracle: 310 cases;
  • validation census: 164/217;
  • PSSM referee: 60/12/30/1; fUML referee: 23/0/28/4.

  The suite counts come from doc-counts -site-blocks on the v0.9.1 tree. A new paragraph covers the three gates that read differently at develop's head (rejection 311; PSSM 52/12/34/5 after refactor(lang): remove the defer state-body extension #694), and points out that the README's stdlib prose still says 105/105.

• Coverage table removed. Since P2 moved the suites under tests/, coverage measured from each package's own tests understates exactly the packages those suites exercise. The text now points to make coverage / SonarCloud instead.

• Release follow-through. R2 is rewritten: the Node, Java and Rust clients now publish from the core tag (ci(release): publish the Node client to npm from the core v* tag #735, ci(release): publish the Java client to Maven Central from the core v* tag #738, ci(release): publish the Rust client to crates.io from the core v* tag #739). npm and crates.io serve 0.9.1. Maven Central's 0.9.1 artifact is not confirmed, because the registry rate-limited the check. Also adds R4's wizard (feat(msi): install through a setup wizard with a destination-folder page #350), the rehearsal (ci(release): add a no-publish release rehearsal #743) and the provenance attestation (ci(release): attest SLSA provenance over the release manifest's artifacts #606).

• Track status fixes:

  • Track P is landed except P7.
  • L7, L8, Q2, Q3, A4 and A7 are landed in v0.9.0.
  • Track W's richer shapes and compartments landed (feat(view): fit a stated box's label to it, draw symbol kinds as their notation, and leave migration bookkeeping undrawn #560, fix(view): keep containment edges off compartment rows and decode quoted-name escapes in labels #589, feat(view): draw Cameo behaviour notation for action and state views #623); fix(view): draw a part's typed ports on the interconnection and end connectors at them #757 is the only open item.
  • I2 is landed as conformance/; what remains is getting the Python client onto the shared runner.
  • Julia and MATLAB landed in v0.9.1.
  • The large-model design has three of its four steps landed (perf(model): parse, index and validate a load's files on a pool of workers #312, feat(repl): analyze each loaded file as a workspace document of its own #309, feat(symbols): hold closed documents as interface records #621, feat(workspace): hydrate and demote interface records, write them from every analysis #637). feat(stressmodel): generate the satellite network as a fleet of occurrences #308, feat(repl): analyze each loaded file as a workspace document of its own #309 and perf(model): parse, index and validate a load's files on a pool of workers #312 had been described as "open".
  • Behavioral editing: the insertion operations landed in v0.9.1. Rewriting an existing declaration in place is not started.
  • The PSSM fails include no runtime gap any more (fix(runtime): draw a due do step against sibling entry units #525).
  • The census tail is now 52 unknown rows.
  • X8's harness halves are not started rather than "open".

• Open PRs and in-progress work. Each open PR against develop is named where it touches an item: feat(client/java): parity with the Python client's newer surface #762, fix(migrate): write an initial into an orthogonal region as that region's entry #763, fix(export): first a then b owns its connector ends (#760) #761, feat(mdk): rename the Cameo plugin to OpenSysML MDK and bridge it into MDK's DocGen #759, fix(view): draw a part's typed ports on the interconnection and end connectors at them #757, fix(check): conformance of an implicit subsetting (#726) #748, feat(query): report the elementId Convert writes (#733) #747 and fix(editors): align SysON plugin with real 2026.9.2 artifacts #483. Work in progress with no PR is also named: parity for the Node, Rust, Julia and MATLAB clients; the SysML v1 -migrate verb; the size of the WebAssembly build.

• Sequencing. Rewritten. The new step 1 is client parity, then the R package, then the C client. After that come X8, D13/D8, B1/B2, M1/M2, and finally the shared C ABI.

How it was verified

• At v0.9.1 with the pilot material fetched: go test -count=1 ./tests/corpus/... -v produces the round-trip, training and pilot-corpora figures, and go run -C tools ./cmd/doc-counts -site-blocks produces the suite figures.
• The referee, census and rejection figures were read from the committed baselines at v0.9.1 and at origin/develop.
• Each PR's tag membership was checked with git merge-base --is-ancestor. Open PRs come from the PR listing against develop.
• Registries checked: npm, crates.io and PyPI JSON APIs. Maven Central returned 429.
• python3 scripts/check-doc-links.py (0 broken), mkdocs build --strict, and go run -C tools ./cmd/doc-counts -check ("already current").

Checklist

• make test and make lint pass locally (docs-only; link check, strict site build and doc-counts check run)
• Tests added or updated for the change (n/a, docs only)
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md (n/a, docs only)
• baselines regenerated and make docs-counts run if a gate count moved (no gate count moved)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/eb03e4319deb4f19bce2ef7d3ab1f0ed
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/eb03e4319deb4f19bce2ef7d3ab1f0ed?variant=devin
Requested by: @HuiJun

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

🔍 Devin Review: 2 flags

Not posted on this PR by your GitHub settings — view them in Devin Review. (Configure)