Build NJS-GROWTH-07 benchmark harness, measured result charts, and AI-rule-safety carousel

[SebaSOFT]

Summary

Unblocks the NJS-GROWTH-07 proof assets by building a real benchmark harness, running it to produce measured actual_benchmark data, and publishing a results showcase generated from that data — plus a qualitative AI-rule-safety carousel. Grounded in the internal NJS-GROWTH-07 proof-assets brief: no src/ changes, no release, and the live playground / README GIF stay deferred until the playground is stable.

Benchmark harness (benchmarks/, consumes built dist/)

• 5 engine adapters — @sebasoft/neuron-js, json-rules-engine, json-logic-js, hand-coded-typescript, rule-engine-js — over 3 example-backed scenarios (pricing-discount, eligibility-approval, workflow-routing) × 3 sizes (smoke/small/medium).
• Fairness gate: every engine must reproduce the scenario's canonical decision before timing, so all engines are measured doing equivalent work.
• Metrics: batched throughput / p50 / p95, fresh-process cold start, esbuild minified bundle size, and Neuron-JS validateScript / explainExecution overhead deltas (competitors have no equivalent step → 0 with a row note).
• run.ts emits benchmarks/results/latest.actual.json (result_kind=actual_benchmark, is_placeholder=false, claims_allowed=true).
• Scripts: yarn benchmark, yarn benchmark:charts. Added esbuild devDependency; biome now lints benchmarks/.

Proof assets

• benchmarks/charts/generate.ts renders the 5 benchmark-chart-*.svg and the generated showcase page docs/benchmarks/results.md from the same source file, so charts and table can never drift from measured data. Every visible number exists verbatim in the result file.
• 5-slide AI-rule-safety carousel SVGs (qualitative, no numbers) + asset metadata; added the missing methodology-card metadata.
• README, VitePress nav/sidebar, and the benchmarks index link the results; methodology now documents how each metric is measured.

Privacy

Scrubbed private chaos-vault/... paths and internal agent notes from the 7 published docs/assets that were leaking them onto the live site (these rendered as visible page text, not just source comments).

Validation

• git diff --check ✅
• yarn lint ✅ (49 files, now including benchmarks/)
• yarn test ✅ (13 files / 68 tests, incl. new tests/contracts/benchmark-results.test.ts)
• yarn examples ✅
• yarn build ✅
• yarn docs:build ✅ — charts + carousel ship in the build; no chaos-vault strings in docs/.vitepress/dist

Guardrails

• No npm release. No src/ changes.
• Charts generated only from measured actual_benchmark output; placeholder fixture untouched.
• Playground app + README GIF/proof-strip remain deferred per the brief (blocked until the playground is stable).

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces a comprehensive benchmark harness for the neuron-js rules engine, comparing it against competitors across multiple scenarios and profiles, and integrates the generated charts and results into the documentation. The review feedback highlights several key improvement opportunities to make the harness more robust and maintainable, including resolving a constructor argument mismatch in neuron-plugins.ts, adding a default error-handling case to the scenario switch statement in the neuron-js adapter, replacing fragile dynamic export indexing with named exports in cold-start-child.ts, extracting duplicated state-reading logic into a shared utility, and refactoring scenario decision functions to avoid hardcoded values.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.