Verification Evidence

This ledger records what was actually inspected or executed for the maintained documentation baseline. It distinguishes a successful release check from a discovered blocker and gives maintainers a repeatable starting point.

Golden Path evidence last verified: 2026-07-13. Official version drift last checked: 2026-07-15.

Baseline Sources

Component	Version	Authority	Repeatable check
Generator Server	20260710.110004	Official latest-version endpoint	npm run check:latest-versions
teaql-java	1.525-RELEASE	Latest-version endpoint, official source, and TeaQL Maven artifacts	Drift check plus inspect the matching POM/JAR/source tag or commit.
teaql-rs	4.1.1	Latest-version endpoint and generated Cargo dependency set	Drift check plus cargo tree in clean generated output.
Cargo client	2.0.8	crates.io index, installed version/help, and live service calls	cargo teaql --version, cargo teaql --help, and cargo teaql services
Maven plugin	1.1.0	Official plugin source, Maven metadata, installed help, and live service calls	Fully qualified help and list-services goals.

The latest-version endpoint does not return Cargo or Maven client versions, so the automated drift check also reads crates.io and TeaQL Maven metadata.

Client and Target Evidence

Evidence ID	Scope	Result	Reproduction
CLIENT-CARGO-2.0.8-HELP	Cargo commands, flags, defaults, and configuration precedence	Verified from installed help and client source.	Run cargo teaql --help, cargo teaql --version, and inspect effective configuration with secrets redacted.
CLIENT-MAVEN-1.1.0-HELP	Maven goals, parameters, dynamic service, and output default	Verified from generated plugin help/source.	Run mvn io.teaql:teaql-maven-plugin:1.1.0:help -Ddetail=true.
TARGET-CATALOG-20260713	Live generation targets	Verified; included Java application/library targets and Rust library/console targets.	Run Cargo services and Maven list-services immediately before generation.
REPORT-MARKDOWN	Evaluation and generation failure output	The same warning-only and blocking-error models were evaluated through cargo-teaql 2.0.8 and Maven plugin 1.1.0 on 2026-07-14. Both returned the same evaluator body; Cargo generation also reused it for invalid KSML. Maven generation separately reproduced local input-path rejection and a server Bad Request (400) template report.	Preserve complete reports and exit statuses; redact environment/account details before sharing.

Rust Golden Path Evidence

Step	Result	Evidence/command shape
Evaluate model	Passed with cargo-teaql 2.0.8.	cargo teaql evaluate --input <model>
Generate library and console	Passed for rust-lib-core and rust-app-console.	Run each dynamic target with explicit input/output; keep outputs as siblings.
Compile/test	Both generated targets passed cargo test.	Run from each generated crate/workspace.
Start provider-backed app	Passed with SQLite.	Set the generated database URL variable and run the console.
Intent-declared query	Passed.	End the chain with comment → purpose → execute_for_xxx.
Audited mutation	Passed.	Use an emitted update_xxx, then audit_as(...).save(...).
Regeneration	Passed after version 1.0.0 → 1.0.1, adding email, and declaring _audit_mask_fields="email".	Re-evaluate, regenerate, inspect API/audit metadata diff, then test offline.

Known Rust evidence boundary: generated sample_data.rs currently contains two queries with purpose but no comment. The generated file was not edited; handwritten examples retain the required order.

Java Golden Path Evidence

Step	Result	Evidence/command shape
Evaluate model	Passed with Maven plugin 1.1.0 after giving the root model a concrete version.	Fully qualified eval goal with explicit input.
Generate java-lib-core	Passed and emitted Java 21, Spring Boot 4.1.0, and TeaQL 1.525-RELEASE.	Dynamic generate -Dservice=java-lib-core.
Regeneration	Passed after version 1.0.0 → 1.0.1, adding email, and declaring audit masking.	Inspect emitted entity/request/expression/checker/metadata diff.
Generate java-app-console	Blocked by missing server template application.properties.stg.	Preserve the Markdown report and current target catalog.
Resolve SQLite dependency	Generated POM is blocked by nonexistent teaql-data-service-sqlite; published artifact is io.teaql:teaql-sqlite:1.525-RELEASE.	Compare generated POM with Maven repository metadata.
Compile generated entities	Still blocked after using the published artifact: generated internalSet/internalGet do not match runtime __internalSet/__internalGet.	Compile without editing generated source; inspect first compiler errors and runtime signatures.
Query and audited mutation	Not executed because compilation is blocked upstream.	Re-run only after both generator defects are fixed.

These Java blockers are generator/runtime alignment evidence, not permission to patch generated output.

Documentation Checks

Run the maintained checks from the site root:

npm run check:docs-contracts
npm run check:latest-versions
npm run build
git diff --check

The contract check covers query intent order, mutation audit declarations, stale Rust provider names, sidebar reachability, thin pages, metadata, and a focused set of unsupported absolute claims. The drift check covers the version endpoint, Cargo client index, and Maven plugin metadata.

CI Artifact Retention

The documentation workflow uploads a docs-evidence-<run>-<attempt> artifact for 30 days, even when a check fails. It contains:

• manifest.json and a Markdown summary with repository, commit, ref, runner, run ID, attempt, and step outcomes;
• complete npm ci, contract-check, contract-test, live version-check, offline version-test, evidence-writer test, and Docusaurus build logs for the steps that ran;
• the generated static build/ directory when the production build succeeds.

This closes the documentation-check retention gap.

The current manifest.json uses schema version 3. It records install, documentation contract, contract-test, live version, offline version-test, evidence-writer-test, and Docusaurus build outcomes independently.

The manually dispatched TeaQL Golden Path evidence workflow now provides separate Rust and Java jobs using the committed minimal model fixture. It archives client help, live target catalogs, evaluation/generation logs, generated dependency manifests, and build/test output. The Rust job generates and tests both current targets. The Java job classifies a successful build, the known generator blockers, or an unexpected failure without editing generated source.

The workflow has not yet been run from the repository's GitHub Actions page, so do not treat its presence as retained CI proof. After the first run, record the run/artifact identifier here. Query/mutation runtime probes remain future work; until they are added, the dated local Golden Path results above remain the runtime evidence and are not a release support guarantee.