harness-everything 0.22.0 → 0.22.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.claude-plugin/marketplace.json +1 -1
- package/.claude-plugin/plugin.json +1 -1
- package/CHANGELOG.md +6 -0
- package/README.md +30 -0
- package/benchmarks/results/system-one/cua-s1-forms-resident-async-cpu-2026-09-23/README.md +87 -0
- package/benchmarks/results/system-one/cua-s1-forms-resident-async-cpu-2026-09-23/paired-latency.json +107 -0
- package/benchmarks/results/system-one/cua-s1-forms-resident-async-cpu-2026-09-23/report.json +856 -0
- package/benchmarks/results/system-one/pre-merge-review-2026-09-23/README.md +92 -0
- package/benchmarks/results/system-one/pre-merge-review-2026-09-23/haiku-spec-conformance.md +241 -0
- package/benchmarks/results/system-one/pre-merge-review-2026-09-23/luna-adversarial.md +212 -0
- package/benchmarks/results/system-one/pre-merge-review-2026-09-23/skill-top1-probe.js +83 -0
- package/benchmarks/results/system-one/pre-merge-review-2026-09-23/skill-top1-probe.json +7700 -0
- package/ci/fixtures/system-one-worker-probe.js +15 -0
- package/ci/mechanism-system-one-resident.test.js +114 -3
- package/ci/mechanism-system-one-router.test.js +25 -0
- package/docs/platform-capabilities.md +7 -2
- package/docs/system-one-routing.md +23 -4
- package/evals/pilot-router-nonsoftware/graders/no-tier-language.md +6 -0
- package/evals/pilot-router-nonsoftware/graders/skill-not-fired.md +9 -0
- package/evals/pilot-router-nonsoftware/prompt.md +8 -0
- package/evals/pilot-router-tier3/graders/skill-fired.md +6 -0
- package/evals/pilot-router-tier3/prompt.md +8 -0
- package/evals/pilot-tdd-bugfix/graders/skill-fired.md +6 -0
- package/evals/pilot-tdd-bugfix/prompt.md +8 -0
- package/evals/pilot-tdd-trivia/graders/no-red-phase-claim.md +7 -0
- package/evals/pilot-tdd-trivia/graders/skill-not-fired.md +9 -0
- package/evals/pilot-tdd-trivia/prompt.md +8 -0
- package/evals/pilot-verify-discussion/graders/no-verification-claim.md +7 -0
- package/evals/pilot-verify-discussion/graders/skill-not-fired.md +9 -0
- package/evals/pilot-verify-discussion/prompt.md +8 -0
- package/evals/pilot-verify-preship/graders/skill-fired.md +6 -0
- package/evals/pilot-verify-preship/prompt.md +8 -0
- package/harness-everything/SKILL.md +1 -1
- package/harness-everything/scripts/system-one/provider.js +14 -2
- package/harness-everything/scripts/system-one/resident.js +26 -2
- package/harness-everything/scripts/system-one/router.js +17 -1
- package/harness-everything/scripts/tier-router.js +13 -4
- package/opencode-plugin/plugin.json +1 -1
- package/package.json +1 -1
- package/plugins/harness-everything/.codex-plugin/plugin.json +1 -1
- package/plugins/harness-everything/plugin.json +1 -1
- package/plugins/harness-everything/skills/harness-everything/SKILL.md +1 -1
- package/plugins/harness-everything/skills/harness-everything/scripts/system-one/provider.js +14 -2
- package/plugins/harness-everything/skills/harness-everything/scripts/system-one/resident.js +26 -2
- package/plugins/harness-everything/skills/harness-everything/scripts/system-one/router.js +17 -1
- package/plugins/harness-everything/skills/harness-everything/scripts/tier-router.js +13 -4
- package/sbom.cdx.json +7 -7
- package/scripts/evaluate-system-one.js +6 -6
|
@@ -5,6 +5,7 @@
|
|
|
5
5
|
const fs = require('node:fs');
|
|
6
6
|
const path = require('node:path');
|
|
7
7
|
const { createHash } = require('node:crypto');
|
|
8
|
+
const net = require('node:net');
|
|
8
9
|
const { spawn } = require('node:child_process');
|
|
9
10
|
const { Worker } = require('node:worker_threads');
|
|
10
11
|
const { validateRequest } = require('./contract');
|
|
@@ -81,6 +82,20 @@ function callSync(port, message, timeoutMs = CALL_TIMEOUT_MS) {
|
|
|
81
82
|
try { return { reply: JSON.parse(Buffer.from(out, 0, Atomics.load(signal, 1)).toString('utf8')) }; }
|
|
82
83
|
catch (_) { return { error: 'json' }; }
|
|
83
84
|
}
|
|
85
|
+
// The same round trip for async callers (the hook entry): no worker thread to bootstrap.
|
|
86
|
+
function callAsync(port, message, timeoutMs = CALL_TIMEOUT_MS) {
|
|
87
|
+
return new Promise(resolve => {
|
|
88
|
+
const chunks = []; let size = 0; let done = false; let socket = null;
|
|
89
|
+
const finish = value => { if (done) return; done = true; clearTimeout(timer); if (socket) socket.destroy(); resolve(value); };
|
|
90
|
+
const timer = setTimeout(() => finish({ error: 'timeout' }), timeoutMs);
|
|
91
|
+
socket = net.connect({ port, host: '127.0.0.1' }, () => socket.end(`${JSON.stringify(message)}\n`));
|
|
92
|
+
socket.on('data', c => { size += c.length; if (size > MAX_REPLY_BYTES) finish({ error: 'limit' }); else chunks.push(c); });
|
|
93
|
+
socket.on('end', () => {
|
|
94
|
+
try { finish({ reply: JSON.parse(Buffer.concat(chunks).toString('utf8')) }); } catch (_) { finish({ error: 'json' }); }
|
|
95
|
+
});
|
|
96
|
+
socket.on('error', e => finish({ error: e.code === 'ECONNREFUSED' ? 'refused' : 'io' }));
|
|
97
|
+
});
|
|
98
|
+
}
|
|
84
99
|
|
|
85
100
|
function start(manifest, files, deps) {
|
|
86
101
|
const lock = lockInfo(files.lock);
|
|
@@ -114,7 +129,16 @@ function score(request, manifest, manifestPath, deps = {}) {
|
|
|
114
129
|
const files = stateFiles(manifest, manifestPath);
|
|
115
130
|
const state = liveState(files);
|
|
116
131
|
if (!state) return start(manifest, files, deps);
|
|
117
|
-
|
|
132
|
+
return outcome(callSync(state.port, { token: state.token, op: 'score', request }), files, state, manifest, deps);
|
|
133
|
+
}
|
|
134
|
+
async function scoreAsync(request, manifest, manifestPath, deps = {}) {
|
|
135
|
+
validateRequest(request);
|
|
136
|
+
const files = stateFiles(manifest, manifestPath);
|
|
137
|
+
const state = liveState(files);
|
|
138
|
+
if (!state) return start(manifest, files, deps);
|
|
139
|
+
return outcome(await callAsync(state.port, { token: state.token, op: 'score', request }), files, state, manifest, deps);
|
|
140
|
+
}
|
|
141
|
+
function outcome(result, files, state, manifest, deps) {
|
|
118
142
|
if (result.reply) {
|
|
119
143
|
if (result.reply.ok === true && Object.hasOwn(result.reply, 'response')) return { status: 'scored', response: result.reply.response };
|
|
120
144
|
if (result.reply.ok === false) return unavailable(result.reply.reason === 'unauthorized' ? 'provider-unavailable' : 'provider-exit');
|
|
@@ -173,4 +197,4 @@ if (require.main === module) {
|
|
|
173
197
|
if (op === 'stop') stop(manifest, manifestPath);
|
|
174
198
|
process.stdout.write(`${JSON.stringify(status(manifest, manifestPath))}\n`);
|
|
175
199
|
}
|
|
176
|
-
module.exports = { stateFiles, readState, callSync, score, ensureReady, status, stop, LOCK_TTL_MS, CALL_TIMEOUT_MS };
|
|
200
|
+
module.exports = { stateFiles, readState, callSync, callAsync, score, scoreAsync, ensureReady, status, stop, LOCK_TTL_MS, CALL_TIMEOUT_MS };
|
|
@@ -34,4 +34,20 @@ function selectTier(input, env = process.env, scorer = provider.score) {
|
|
|
34
34
|
const reason = input.explicit && accepted ? 'explicit-precedence' : belowFloor ? 'structural-floor' : decision.reason;
|
|
35
35
|
return { tier: applied ? LABELS[decision.selectedId] : input.tier, diagnostic: { mode, ...decision, applied, reason } };
|
|
36
36
|
}
|
|
37
|
-
|
|
37
|
+
// Hook entry: score the same fixed-catalog request asynchronously before routing (no sync worker bridge).
|
|
38
|
+
// Returns a scorer for selectTier, or null when System One is off/invalid; nothing is called then.
|
|
39
|
+
async function prescoreTier(prompt, env = process.env, scoreAsync = provider.scoreAsync) {
|
|
40
|
+
if (!['shadow', 'prefer'].includes(env.HARNESS_SYSTEM_ONE_MODE)) return null;
|
|
41
|
+
let request;
|
|
42
|
+
try { request = createRequest('tier', prompt, TIER_OPTIONS); }
|
|
43
|
+
catch (_) { return null; }
|
|
44
|
+
let result; let failed = false;
|
|
45
|
+
try { result = await scoreAsync(request, env.HARNESS_SYSTEM_ONE_CONFIG); }
|
|
46
|
+
catch (_) { failed = true; }
|
|
47
|
+
return (req, config) => {
|
|
48
|
+
if (req.requestHash !== request.requestHash) return provider.score(req, config);
|
|
49
|
+
if (failed) throw new Error('provider-unavailable');
|
|
50
|
+
return result;
|
|
51
|
+
};
|
|
52
|
+
}
|
|
53
|
+
module.exports = { selectTier, prescoreTier, TIER_OPTIONS };
|
|
@@ -231,7 +231,7 @@ function emitDynamicSkills(promptLower, context, recommendedGuides) {
|
|
|
231
231
|
}
|
|
232
232
|
}
|
|
233
233
|
|
|
234
|
-
function run(userPrompt, context) {
|
|
234
|
+
function run(userPrompt, context, options = {}) {
|
|
235
235
|
console.log(`[Tier Routing Pre-check]`);
|
|
236
236
|
|
|
237
237
|
const promptLower = userPrompt.toLowerCase();
|
|
@@ -317,7 +317,7 @@ function run(userPrompt, context) {
|
|
|
317
317
|
const structuralFloor = hasMacroSignal ? 'tier3'
|
|
318
318
|
: !isTrivialDocsEdit && hasMultipleTasks && hasMultipleSentences ? 'tier2' : null;
|
|
319
319
|
const semantic = selectTier({ prompt: userPrompt, tier: recommendedTier, floor: structuralFloor,
|
|
320
|
-
explicit: Boolean(detectExplicitStrategy(userPrompt, detectFableModel(userPrompt))) });
|
|
320
|
+
explicit: Boolean(detectExplicitStrategy(userPrompt, detectFableModel(userPrompt))) }, process.env, options.systemOneScorer || undefined);
|
|
321
321
|
if (semantic.diagnostic) console.log(`\n=> SYSTEM ONE: ${JSON.stringify(semantic.diagnostic)}`);
|
|
322
322
|
if (semantic.diagnostic?.applied) {
|
|
323
323
|
recommendedTier = semantic.tier;
|
|
@@ -460,12 +460,21 @@ function run(userPrompt, context) {
|
|
|
460
460
|
return contract;
|
|
461
461
|
}
|
|
462
462
|
|
|
463
|
+
// Entry point: pre-score System One asynchronously so routing never needs the sync worker bridge.
|
|
464
|
+
async function main(prompt, context) {
|
|
465
|
+
let systemOneScorer = null;
|
|
466
|
+
if (['shadow', 'prefer'].includes(process.env.HARNESS_SYSTEM_ONE_MODE)) {
|
|
467
|
+
try { systemOneScorer = await require('./system-one/router').prescoreTier(prompt); } catch (_) { /* the sync path still scores */ }
|
|
468
|
+
}
|
|
469
|
+
run(prompt, context, { systemOneScorer });
|
|
470
|
+
}
|
|
471
|
+
|
|
463
472
|
if (require.main === module) {
|
|
464
473
|
let userPrompt = process.argv[2] || '';
|
|
465
474
|
let hookContext = null;
|
|
466
475
|
|
|
467
476
|
if (process.argv[2]) {
|
|
468
|
-
|
|
477
|
+
main(userPrompt);
|
|
469
478
|
} else if (process.stdin.isTTY) {
|
|
470
479
|
run('');
|
|
471
480
|
} else {
|
|
@@ -482,7 +491,7 @@ if (process.argv[2]) {
|
|
|
482
491
|
} catch (err) {
|
|
483
492
|
// Not valid JSON on stdin - fall back to argv/manual behavior.
|
|
484
493
|
}
|
|
485
|
-
|
|
494
|
+
main(userPrompt, hookContext);
|
|
486
495
|
});
|
|
487
496
|
}
|
|
488
497
|
}
|
package/sbom.cdx.json
CHANGED
|
@@ -2,23 +2,23 @@
|
|
|
2
2
|
"$schema": "https://cyclonedx.org/schema/bom-1.7.schema.json",
|
|
3
3
|
"bomFormat": "CycloneDX",
|
|
4
4
|
"specVersion": "1.7",
|
|
5
|
-
"serialNumber": "urn:uuid:
|
|
5
|
+
"serialNumber": "urn:uuid:3cb9e1e8-a9f0-5bc4-87ab-794acb1e79c3",
|
|
6
6
|
"version": 1,
|
|
7
7
|
"metadata": {
|
|
8
|
-
"timestamp": "2026-09-
|
|
8
|
+
"timestamp": "2026-09-23T11:35:56.699Z",
|
|
9
9
|
"tools": [
|
|
10
10
|
{
|
|
11
11
|
"vendor": "Harness Everything",
|
|
12
12
|
"name": "generate-sbom.js",
|
|
13
|
-
"version": "0.22.
|
|
13
|
+
"version": "0.22.1"
|
|
14
14
|
}
|
|
15
15
|
],
|
|
16
16
|
"component": {
|
|
17
17
|
"type": "application",
|
|
18
|
-
"bom-ref": "pkg:npm/harness-everything@0.22.
|
|
18
|
+
"bom-ref": "pkg:npm/harness-everything@0.22.1",
|
|
19
19
|
"name": "harness-everything",
|
|
20
|
-
"version": "0.22.
|
|
21
|
-
"purl": "pkg:npm/harness-everything@0.22.
|
|
20
|
+
"version": "0.22.1",
|
|
21
|
+
"purl": "pkg:npm/harness-everything@0.22.1"
|
|
22
22
|
}
|
|
23
23
|
},
|
|
24
24
|
"components": [
|
|
@@ -1999,7 +1999,7 @@
|
|
|
1999
1999
|
]
|
|
2000
2000
|
},
|
|
2001
2001
|
{
|
|
2002
|
-
"ref": "pkg:npm/harness-everything@0.22.
|
|
2002
|
+
"ref": "pkg:npm/harness-everything@0.22.1",
|
|
2003
2003
|
"dependsOn": [
|
|
2004
2004
|
"pkg:npm/%40commitlint/cli@19.8.1",
|
|
2005
2005
|
"pkg:npm/%40commitlint/config-conventional@19.8.1",
|
|
@@ -4,12 +4,12 @@ const fs = require('node:fs');
|
|
|
4
4
|
const os = require('node:os');
|
|
5
5
|
const { performance } = require('node:perf_hooks');
|
|
6
6
|
const { validateCorpus, evaluate } = require('../harness-everything/scripts/system-one/evaluate');
|
|
7
|
-
const {
|
|
7
|
+
const { scoreAsync, provenance, readManifest } = require('../harness-everything/scripts/system-one/provider');
|
|
8
8
|
const resident = require('../harness-everything/scripts/system-one/resident');
|
|
9
9
|
const { decide } = require('../harness-everything/scripts/system-one/contract');
|
|
10
10
|
const { run: route } = require('../harness-everything/scripts/tier-router');
|
|
11
11
|
|
|
12
|
-
function run(corpus, manifest) {
|
|
12
|
+
async function run(corpus, manifest) {
|
|
13
13
|
validateCorpus(corpus);
|
|
14
14
|
let config = null;
|
|
15
15
|
try { config = readManifest(manifest); } catch (_) { /* Missing provider remains an explicit unavailable measurement. */ }
|
|
@@ -28,7 +28,7 @@ function run(corpus, manifest) {
|
|
|
28
28
|
const runs = [];
|
|
29
29
|
for (let i = 0; i < 2; i++) {
|
|
30
30
|
const start = performance.now();
|
|
31
|
-
const result =
|
|
31
|
+
const result = await scoreAsync(c.request, manifest);
|
|
32
32
|
const decision = result.status === 'scored' ? decide(c.request, result.response) : result;
|
|
33
33
|
if (decision.selectedId === 'unclassified') Object.assign(decision, { status: 'abstain', reason: 'unclassified', selectedId: null });
|
|
34
34
|
const scores = ['accepted', 'abstain'].includes(decision.status) ? result.response.scores.map(s => s.probability) : null;
|
|
@@ -55,12 +55,12 @@ function run(corpus, manifest) {
|
|
|
55
55
|
limitations: ['No independently verified holdout review', ...(warm ? [] : ['No warm inference measurement']), 'No live-host or policy evidence'] }, records };
|
|
56
56
|
}
|
|
57
57
|
if (require.main === module) {
|
|
58
|
-
|
|
58
|
+
(async () => {
|
|
59
59
|
const [input, manifest, output, extra] = process.argv.slice(2);
|
|
60
60
|
if (!input || !manifest || !output || extra) throw new Error('usage');
|
|
61
|
-
const report = run(JSON.parse(fs.readFileSync(input, 'utf8')), manifest);
|
|
61
|
+
const report = await run(JSON.parse(fs.readFileSync(input, 'utf8')), manifest);
|
|
62
62
|
fs.writeFileSync(output, JSON.stringify(report, null, 2) + '\n');
|
|
63
63
|
console.log(JSON.stringify({ cases: report.cases, coverage: report.model.coverage, rolloutReady: report.rolloutReady }));
|
|
64
|
-
}
|
|
64
|
+
})().catch(() => { console.error('System One evaluation failed: invalid corpus, arguments, or output path.'); process.exitCode = 1; });
|
|
65
65
|
}
|
|
66
66
|
module.exports = { run };
|