[devin-ai-integration]

What and why

Second area of the documentation accuracy audit (the first is #1503, examples/): README.md, docs/guide, docs/manual, docs/clients and docs/reference, measured against binaries built with make build at develop 76c98694c (merge of #1483). Every fenced command and REPL block was replayed (per block and cumulatively per page), every fenced sysml/kerml snippet validated standalone / wrapped / cumulative / as a part-def body, and every CLI flag, % command, environment variable, Make target, gRPC capability and -print-config host name grepped in source or compared with the binary. Each mismatch below was reproduced on the v0.11.0 release binary as well, so all of them are documentation drift, not regressions. Docs-only; no model or code moved.

page	claimed	actual (proving command)
README.md runtime showcase	stage1DeltaV = 3037.69… [SI::'m/s'], margin = 70.65…, margin = -329.34…	one-element sequences [3037.6966629706967 [SI::'m/s']] … (./bin/sysml -quiet -analysis DeltaVBudget::saturnIBAscent examples/runtime-showcase/delta-v-budget.sysml, and with required = 9800 ['m/s'])
docs/guide/05-checking.md:534,629	3:36: error: unresolved reference: size …, 1:35: error: cannot bind …	error[unresolved]: …, error[binding-connector-type-conformance]: … (REPL replay of the blocks)
docs/guide/07-saving-and-rdf.md:46	3:3: error: expected a namespace member	3:3: error[syntax]: expected a namespace member
docs/guide/06-behavior.md:504-540 (Guard::Sentry)	no Use %events … / Use %step … hints, no choice-point line, log = "watch;watching(exit);abort;"	both hints, 1 choice point; %trace on to see them, and … (order-dependent; 1 choice point — explore shows the set) (replay of the block; the page's first transcript at line 90 already shows the hint form)
docs/reference/repl-commands.md:440-478, 530-556, 559-563, 623-637, 654-664 and docs/guide/04-repl.md:361-370	%render Demo::summary tree/Mermaid/DOT/PlantUML/D2 with qualified names and declared types (part def Demo::Vehicle, attribute mass : Real, view Demo::summary::detail)	unqualified names without types, the nested part wheel expanded, node ids renumbered, Mermaid themeCSS/themeVariables frontmatter (cumulative replay of the page's Demo package; identical on v0.11.0). The … elisions of the DOT/PlantUML/D2 blocks are kept; only the lines they show were re-recorded.
docs/reference/java-api.md:14	<version>0.10.1</version>	0.11.0 is on Maven Central (central.sonatype.com versions API lists 0.11.0, 0.10.1; client/java/pom.xml is 0.11.0; docs/guide/09-clients.md and client/java/README.md already say 0.11.0)

Checked and accurate (no edit): sysml-mcp -print-config blocks for all seven hosts (byte-identical to the binary's output with -print-config-command /usr/local/bin/sysml-mcp); 04-repl.md %clear dropped-instance note; repl-commands.md quoted-name suggestion; every % command in the REPL pages is in %help; every OPENSYSML_* variable named in docs/reference is in source or the installer; every make target named is in the Makefile; the jupyter-kernel.md 0.10.0 lines are illustrative, not stale.

Quick-starts run against bin/sysml-grpc (OPENSYSML_BINARY): Python (opensysml 0.11.0 from PyPI), Node (@openmbee/opensysml 0.11.0 from npm) and Java (mvn -f client/java/pom.xml install, then the page's Demo.java) each print 1800.0 as documented. Not run: the Rust, Julia and MATLAB quick-starts.

Left as written (evidence for a follow-up):

• docs/guide/05-checking.md:758 — the %check Box::crateUnset witness (height = 0.875 [metre], length = 1.09375, volume = 1.016845703125) is one z3 model among many; develop with z3 4.x prints 0.9375 / 1.0625 / 1.058349609375, and v0.11.0 fails the same block with value cannot be fixed for solving: a undetermined. Solver-dependent, so not re-recorded.
• docs/guide/modeling-product-lines.md transcripts depend on the downloaded pilot corpora and were not replayed.
• Command blocks that name files the page describes but does not ship (model.sysml, a.sysml, tmt.sysml, landing.sysml, …) cannot be replayed literally and were read instead.

Coverage: fully read and replayed — README.md, docs/guide/02, 04, 05, 06, 07, docs/reference/repl-commands.md (render section), java-api.md, sysml-mcp.md (config section), docs/clients/python/index.md, node.md, java.md; replayed and triaged by harness, prose skimmed — the remaining docs/guide, docs/manual and docs/reference pages; not reached beyond the names/figures grep — docs/clients/julia.md, matlab.md, rust.md, docs/reference/wire-contract.md (hash transcript, unedited).

Overlap (open PRs touching README.md): #1502 (304df41bb), #1484 (70dae7174), #1466 (e589bddc2) — each dry-merged onto this head with no conflict; none restates the showcase figures.

How it was verified

At e2db498fe (this head, on develop 76c98694c):

• python3 scripts/check-doc-links.py — rc 0
• python3 scripts/check-doc-ids.py — rc 0
• python3 scripts/check-doc-figures.py — rc 0
• make docs-landing-assets then make docs-check — 0 broken link(s), rc 0
• python -m mkdocs build --strict — rc 0
• python3 scripts/changelog.py check — rc 0

Checklist

• make test and make lint pass locally — docs-only change; the documentation gates above are the ones that cover it
• Tests added or updated for the change — not applicable, no code changed
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/docs-audit-user-docs.fixed.md
• baselines regenerated and make docs-counts run if a gate count moved — no gate count moved
• No internal work-item labels in the body, docs, or changelog

[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: 1 flag

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