docs: add sparq development notice to the top of the README

[jeswr]

🚧 Draft — agent-generated, for @jeswr's wording/figures review before landing

Adds a > [!NOTE] banner at the very top of the README (directly under the H1) announcing that @jeswr is now actively developing sparq, for eye-js users to find the successor project. Nothing else in the README is touched.

Direct eye-js-vs-sparq DeepTaxonomy comparison (+ the caveat)

The speed line no longer reuses sparq's sparq-vs-native-EYE figure ("orders of magnitude faster than the native EYE reasoner… ≈62×"). That 62× compared sparq against the standalone native EYE binary running a full forward --pass closure — a different task from what eye-js's own bench actually runs, so quoting it against eye-js would have been misleading.

Instead the notice now states a direct eye-js-vs-sparq DeepTaxonomy-1k comparison with measured numbers:

• sparq (native Rust): ≈7 ms engine time / ≈70 ms end-to-end incl. process startup, from sparq's published benchmarks (idle-machine wall-clock dt1k 0.068 s; engine-internal min-of-5 ≈0.005–0.007 s).
• eye-js (WASM build of EYE): ≈390 ms end-to-end, measured on-box (eyereasoner v21.1.10, node 22, 2-core CI box, gated) on the canonical linear DeepTaxonomy-1k shape from perf/bench.ts — of which ≈270 ms is one-time WASM module instantiation and ≈120 ms is reasoning.

⚠️ Caveat, stated explicitly in the notice: this is a native-vs-WASM comparison. sparq publishes no WASM DeepTaxonomy timing anywhere, so the honest framing is sparq-native vs eye-js-WASM, and the notice says so — the gap reflects both the engine and the native-vs-WASM runtime cost (WASM boot alone is ≈270 ms of eye-js's ≈390 ms). Framed as "roughly an order of magnitude on reasoning time and several-fold end-to-end," not "orders of magnitude." When sparq publishes a WASM DT number this should become a true wasm-vs-wasm line (tracked in #1965).

Every other figure is verified

Claim	Source
Rust-native RDF triplestore, SPARQL 1.1/1.2	sparq README lead
N3 reasoning (RDFS / OWL 2 RL / Notation3), native + WASM	sparq README features; skills/inference/SKILL.md; crates/sparq-reason-wasm, deployed at sparq.jeswr.org/wasm/reason/
sparq reasoning bundle ≈2.6 MB wasm / ≈0.9 MB gzip	re-measured from the deployed artifact 2026-07-04: sparq_reason_wasm_bg.wasm = 2,572,368 B raw, 909,549 B gzip -9
eye-js ≈3.8 MB wasm + data / ≈4.1 MB browser bundle	measured from this repo @ fa76971: swipl-web.wasm 2,190,887 B + swipl-web.data 1,642,703 B = 3,833,590 B; webpack bundle/index.js 4,051,911 B
DeepTaxonomy-1k: sparq ≈7 ms/≈70 ms vs eye-js ≈390 ms	sparq: bench/inference/eye-comparison.md; eye-js: on-box measurement (eyereasoner v21.1.10, canonical perf/bench.ts linear DT-1000 shape)
experimental / pre-1.0 / evolving quickly	sparq README status banner
migration npm package "tracked in sparq-org/sparq#1499"	sparq-org/sparq#1499 — does not exist on npm yet, hence tracked, not available

Deliberately not claimed: any identical-closure guarantee between eye-js and sparq (verified only sparq-vs-native-EYE in the linked file; eye-js's bench uses a goal-directed --query, so its derivation may differ), and any published sparq npm package.

Trackers

• Keep the sparq comparison in the README current as sparq evolves #1965 — broader tracker: keep the notice's figures (bundle size, DT timings) and state (pre-1.0 → stable, native-vs-WASM framing) current as sparq evolves.
• Update the sparq migration pointer in the README once the npm package ships #1960 — narrow tracker: make the migration pointer concrete (npm install <name> + drop-in snippet) once sparq's npm package publishes.

Adjacency note

⚠️ #1954 (browser dist e2e test + RDF 1.2 support level) also edits README.md. This change is at the very top (under the H1); #1954's edit is in the RDF-1.2 section — different region, so a conflict is unlikely, but rebase after #1954 lands if the merge queue complains.

🤖 Generated with Claude Code

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'EYE JS Benchmark'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.25.

Benchmark suite	Current: a82637b	Previous: d129aee	Ratio
Run socrates query	15.33 ops/sec (±6.69%)	22.2 ops/sec (±2.39%)	1.45
Executing the socrates query	115 ops/sec (±20.74%)	3890 ops/sec (±1.67%)	33.83
Run deep taxonomy benchmark [10]	6.51 ops/sec (±2.61%)	18.83 ops/sec (±6.56%)	2.89
Run deep taxonomy benchmark [50]	0.24 ops/sec (±1.60%)	19.63 ops/sec (±7.26%)	81.79
Run deep taxonomy benchmark [100]	0.04 ops/sec (±1.32%)	18.67 ops/sec (±6.95%)	466.75
Run deep taxonomy benchmark [10] [reasoning only]	7.07 ops/sec (±16.51%)	4107 ops/sec (±1.65%)	580.91
Run deep taxonomy benchmark [50] [reasoning only]	0.56 ops/sec (±21.44%)	4147 ops/sec (±1.85%)	7405.36
Run deep taxonomy benchmark [100] [reasoning only]	0.15 ops/sec (±23.05%)	4136 ops/sec (±1.60%)	27573.33
Run timbl + foaf + rdfs rules	1.95 ops/sec (±2.47%)	17.23 ops/sec (±6.18%)	8.84
Run timbl + foaf + owl2rl rules	1.12 ops/sec (±1.60%)	17.03 ops/sec (±3.91%)	15.21
Run timbl + rdfs rules	4.43 ops/sec (±2.59%)	18.28 ops/sec (±6.60%)	4.13
Run timbl + owl2rl rules	5.66 ops/sec (±0.72%)	17.1 ops/sec (±6.85%)	3.02
Run timbl + foaf + rdfs rules [string]	2 ops/sec (±0.46%)	18.45 ops/sec (±7.08%)	9.22
Run timbl + foaf + owl2rl rules [string]	1.15 ops/sec (±0.46%)	18.54 ops/sec (±6.64%)	16.12
Run timbl + rdfs rules [string]	4.52 ops/sec (±0.30%)	18.77 ops/sec (±6.77%)	4.15
Run timbl + owl2rl rules [string]	5.82 ops/sec (±0.29%)	18.98 ops/sec (±6.47%)	3.26

This comment was automatically generated by workflow using github-action-benchmark.

CC: @jeswr