@saulwade/swl-ses 1.4.0 → 1.4.2
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.md +4 -3
- package/README.md +15 -14
- package/agentes/nemesis-auditor-swl.md +161 -0
- package/bin/swl-mcp-server.js +187 -187
- package/comandos/swl/.evolved.json +22 -22
- package/comandos/swl/contribuir.md +233 -233
- package/comandos/swl/nemesis.md +122 -0
- package/comandos/swl/salud.md +34 -0
- package/comandos/swl/verificar.md +45 -0
- package/gateway/lib/event-channel.js +191 -191
- package/habilidades/backend-production-resilience/SKILL.md +288 -288
- package/habilidades/benchmark-memoria/SKILL.md +186 -186
- package/habilidades/diagrama-arquitectura/assets/template.html +276 -276
- package/habilidades/doubt-driven-review/SKILL.md +171 -171
- package/habilidades/doubt-driven-review/recursos/EXAMPLES.md +130 -130
- package/habilidades/eval-framework/SKILL.md +212 -212
- package/habilidades/feynman-auditor-swl/SKILL.md +123 -0
- package/habilidades/feynman-auditor-swl/recursos/preguntas-language-agnostic.md +108 -0
- package/habilidades/harness-claude-code/SKILL.md +299 -299
- package/habilidades/infra-github-actions/SKILL.md +166 -166
- package/habilidades/legacy-code-rescue/SKILL.md +267 -267
- package/habilidades/manejo-errores/.evolved.json +8 -8
- package/habilidades/meta-skills-estandar/recursos/convencion-examples.md +93 -93
- package/habilidades/meta-skills-estandar/recursos/skills-as-agents.md +163 -163
- package/habilidades/patrones-python/SKILL.md +229 -229
- package/habilidades/patrones-python/recursos/patrones-avanzados.md +469 -469
- package/habilidades/planear-fase/SKILL.md +319 -319
- package/habilidades/release-semver/.evolved.json +8 -8
- package/habilidades/state-inconsistency-auditor-swl/SKILL.md +166 -0
- package/habilidades/state-inconsistency-auditor-swl/recursos/coupled-state-patterns.md +147 -0
- package/habilidades/testing-python/SKILL.md +340 -340
- package/habilidades/web-fetcher-routing/SKILL.md +75 -0
- package/hooks/claudemd-bloat-detector.js +161 -161
- package/hooks/lib/agent-routing.js +107 -107
- package/hooks/lib/auto-consolidator.js +335 -335
- package/hooks/lib/error-classifier.js +308 -308
- package/hooks/lib/merkle-audit.js +96 -96
- package/hooks/lib/provenance-tracker.js +191 -191
- package/hooks/lib/rate-limit-tracker.js +253 -253
- package/hooks/lib/resource-quota.js +122 -122
- package/hooks/lib/retry-jitter.js +165 -165
- package/hooks/lib/security-net.js +201 -0
- package/hooks/lib/skill-auditor.js +588 -588
- package/hooks/lib/sync-status.js +228 -228
- package/hooks/lib/taint-tracker.js +107 -107
- package/hooks/lib/text-similarity.js +241 -241
- package/hooks/lib/toon-compressor.js +245 -245
- package/hooks/registro-turnos.js +209 -209
- package/hooks/sugerir-regenerar-inventario.js +170 -170
- package/hooks/validar-formato-post-subagente.js +140 -140
- package/hooks/validar-memoria-hook.js +218 -218
- package/instintos/prompt-appendices.yaml +57 -57
- package/manifiestos/agent-output-schemas.json +57 -57
- package/manifiestos/modulos.json +41 -6
- package/manifiestos/perfiles.json +2 -1
- package/manifiestos/skills-lock.json +30 -9
- package/package.json +2 -2
- package/plantillas/auditor-veto-template.md +105 -105
- package/plantillas/github-workflows/README.md +47 -47
- package/plantillas/github-workflows/release-please.yml +44 -44
- package/plantillas/github-workflows/swl-ci.yml +107 -107
- package/plantillas/github-workflows/swl-security.yml +51 -51
- package/plugin.json +10 -2
- package/reglas/analisis-previo-tareas-grandes.md +172 -172
- package/reglas/arreglar-al-detectar.md +147 -147
- package/reglas/fragmentos-compartidos.md +152 -152
- package/reglas/harness-claude-code.md +213 -213
- package/reglas/usar-context7.md +226 -226
- package/schemas/diary-entry.schema.json +80 -80
- package/scripts/audit-tools/audit-history.js +330 -0
- package/scripts/audit-tools/bundle-tracker.js +290 -0
- package/scripts/audit-tools/canary-monitor.js +352 -0
- package/scripts/audit-tools/code-profiler.js +605 -0
- package/scripts/audit-tools/dep-doctor.js +320 -0
- package/scripts/audit-tools/env-validator.js +206 -0
- package/scripts/audit-tools/lib/fs-walk.js +48 -0
- package/scripts/audit-tools/lib/output.js +23 -0
- package/scripts/audit-tools/migration-checker.js +392 -0
- package/scripts/audit-tools/pentest-scanner.js +1436 -0
- package/scripts/benchmark-memoria.js +167 -167
- package/scripts/configurar-branch-protection.js +418 -418
- package/scripts/detectar-aprendizajes-duplicados.js +151 -151
- package/scripts/field-report.js +199 -199
- package/scripts/generar-checklists-consolidados.js +273 -273
- package/scripts/generar-inventario.js +420 -420
- package/scripts/generar-matriz-lenguajes.js +271 -271
- package/scripts/lib/artefactos-python.js +43 -43
- package/scripts/lib/benchmark-metrics.js +160 -160
- package/scripts/lib/budget-enforcer.js +252 -252
- package/scripts/lib/configurar-ci.js +380 -380
- package/scripts/lib/contadores-inventario.js +217 -217
- package/scripts/lib/detectar-stack-detallado.js +307 -307
- package/scripts/lib/diary-entry.js +234 -234
- package/scripts/lib/eval-metrics-store.js +218 -218
- package/scripts/lib/eval-quality.js +171 -171
- package/scripts/lib/eval-schemas.js +144 -144
- package/scripts/lib/eval-self-correct.js +106 -106
- package/scripts/lib/eval-validator.js +185 -185
- package/scripts/lib/jaccard-similarity.js +98 -98
- package/scripts/lib/longmemeval-runner.js +125 -125
- package/scripts/lib/manifiestos.js +42 -1
- package/scripts/lib/npm-version.js +261 -261
- package/scripts/lib/paquetes-conocidos.js +50 -50
- package/scripts/lib/prompt-builder.js +264 -264
- package/scripts/lib/rrf-fusion.js +175 -175
- package/scripts/lib/scoring-instintos.js +277 -277
- package/scripts/lib/semantic-search.js +252 -252
- package/scripts/limpiar-artefactos-python.js +131 -131
- package/scripts/mcp-server/README.md +128 -128
- package/scripts/mcp-server/handlers.js +206 -206
- package/scripts/migrar-csv-a-array.js +168 -168
- package/scripts/migrar-fase-dominio.js +201 -201
- package/scripts/publicar.js +511 -511
- package/scripts/run-eval.js +141 -141
- package/scripts/validar-manifest.js +231 -195
- package/scripts/validar-userland-vacio.js +110 -110
|
@@ -1,171 +1,171 @@
|
|
|
1
|
-
'use strict';
|
|
2
|
-
|
|
3
|
-
/**
|
|
4
|
-
* eval-quality.js — Métricas de calidad para outputs estructurados de SWL.
|
|
5
|
-
*
|
|
6
|
-
* Patrón adoptado de `temp/agentmemory-main/src/eval/quality.ts`. Adaptado a
|
|
7
|
-
* swl-ses: scoring de aprendizajes, instintos, y resultados de búsqueda
|
|
8
|
-
* memoria.
|
|
9
|
-
*
|
|
10
|
-
* Cada función devuelve un score en [0, 100]. Los puntos se asignan por
|
|
11
|
-
* "presencia y calidad de campos clave" — un output con todos los campos
|
|
12
|
-
* tiene score 100, un output trivial tiene score bajo.
|
|
13
|
-
*
|
|
14
|
-
* Funciones puras zero-deps.
|
|
15
|
-
*
|
|
16
|
-
* @module scripts/lib/eval-quality
|
|
17
|
-
*/
|
|
18
|
-
|
|
19
|
-
// ── helpers ───────────────────────────────────────────────────────────────────
|
|
20
|
-
|
|
21
|
-
function clamp(n, min, max) {
|
|
22
|
-
if (Number.isNaN(n)) return min;
|
|
23
|
-
return Math.max(min, Math.min(max, n));
|
|
24
|
-
}
|
|
25
|
-
|
|
26
|
-
// ── scoring de outputs ────────────────────────────────────────────────────────
|
|
27
|
-
|
|
28
|
-
/**
|
|
29
|
-
* Score de calidad de una observación comprimida.
|
|
30
|
-
* Adaptado de scoreCompression() de agentmemory.
|
|
31
|
-
*
|
|
32
|
-
* Distribución de puntos:
|
|
33
|
-
* - facts presentes (no vacíos): 25
|
|
34
|
-
* - facts ≥ 3: +10
|
|
35
|
-
* - narrative ≥ 20 chars: 20
|
|
36
|
-
* - narrative ≥ 50 chars: +5
|
|
37
|
-
* - title 5-120 chars: 15
|
|
38
|
-
* - concepts presentes: 15
|
|
39
|
-
* - importance ∈ [1, 10]: 10
|
|
40
|
-
*
|
|
41
|
-
* @param {object} obs
|
|
42
|
-
* @returns {number} en [0, 100]
|
|
43
|
-
*/
|
|
44
|
-
function scoreObservacion(obs) {
|
|
45
|
-
let score = 0;
|
|
46
|
-
if (Array.isArray(obs?.facts) && obs.facts.length > 0) score += 25;
|
|
47
|
-
if (Array.isArray(obs?.facts) && obs.facts.length >= 3) score += 10;
|
|
48
|
-
if (typeof obs?.narrative === 'string' && obs.narrative.length >= 20) score += 20;
|
|
49
|
-
if (typeof obs?.narrative === 'string' && obs.narrative.length >= 50) score += 5;
|
|
50
|
-
if (typeof obs?.title === 'string' && obs.title.length >= 5 && obs.title.length <= 120) score += 15;
|
|
51
|
-
if (Array.isArray(obs?.concepts) && obs.concepts.length > 0) score += 15;
|
|
52
|
-
if (typeof obs?.importance === 'number' && obs.importance >= 1 && obs.importance <= 10) score += 10;
|
|
53
|
-
return clamp(score, 0, 100);
|
|
54
|
-
}
|
|
55
|
-
|
|
56
|
-
/**
|
|
57
|
-
* Score de calidad de un resumen de sesión.
|
|
58
|
-
* Adaptado de scoreSummary() de agentmemory.
|
|
59
|
-
*
|
|
60
|
-
* @param {object} summary
|
|
61
|
-
* @returns {number} en [0, 100]
|
|
62
|
-
*/
|
|
63
|
-
function scoreResumen(summary) {
|
|
64
|
-
let score = 0;
|
|
65
|
-
if (typeof summary?.title === 'string' && summary.title.length >= 5) score += 20;
|
|
66
|
-
if (typeof summary?.narrative === 'string' && summary.narrative.length >= 20) score += 25;
|
|
67
|
-
if (typeof summary?.narrative === 'string' && summary.narrative.length >= 100) score += 5;
|
|
68
|
-
if (Array.isArray(summary?.keyDecisions) && summary.keyDecisions.length > 0) score += 20;
|
|
69
|
-
if (Array.isArray(summary?.filesModified) && summary.filesModified.length > 0) score += 15;
|
|
70
|
-
if (Array.isArray(summary?.concepts) && summary.concepts.length > 0) score += 15;
|
|
71
|
-
return clamp(score, 0, 100);
|
|
72
|
-
}
|
|
73
|
-
|
|
74
|
-
/**
|
|
75
|
-
* Score de relevancia de un contexto inyectado.
|
|
76
|
-
* Adaptado de scoreContextRelevance() de agentmemory.
|
|
77
|
-
*
|
|
78
|
-
* @param {string} context - Texto del contexto.
|
|
79
|
-
* @param {string} project - Nombre del proyecto.
|
|
80
|
-
* @returns {number} en [0, 100]
|
|
81
|
-
*/
|
|
82
|
-
function scoreRelevanciaContexto(context, project) {
|
|
83
|
-
if (typeof context !== 'string') return 0;
|
|
84
|
-
let score = 0;
|
|
85
|
-
if (context.length > 0) score += 20;
|
|
86
|
-
if (project && context.toLowerCase().includes(String(project).toLowerCase())) score += 20;
|
|
87
|
-
if (context.includes('<')) score += 15;
|
|
88
|
-
const sectionCount = (context.match(/<\w+>/g) || []).length;
|
|
89
|
-
if (sectionCount >= 2) score += 15;
|
|
90
|
-
if (sectionCount >= 4) score += 10;
|
|
91
|
-
if (context.length >= 100) score += 10;
|
|
92
|
-
if (context.length >= 500) score += 10;
|
|
93
|
-
return clamp(score, 0, 100);
|
|
94
|
-
}
|
|
95
|
-
|
|
96
|
-
/**
|
|
97
|
-
* Score de calidad de un aprendizaje SWL extraído de una sesión.
|
|
98
|
-
* Específico de swl-ses (no en agentmemory).
|
|
99
|
-
*
|
|
100
|
-
* Distribución:
|
|
101
|
-
* - título no vacío y descriptivo (≥ 10 chars): 20
|
|
102
|
-
* - título empieza con fecha [YYYY-MM-DD]: 10
|
|
103
|
-
* - cuerpo ≥ 100 chars: 20
|
|
104
|
-
* - cuerpo ≥ 300 chars: +10
|
|
105
|
-
* - tipo identificado (decisión|patrón|...): 15
|
|
106
|
-
* - menciona archivo o regla concreta: 15
|
|
107
|
-
* - tiene "trigger" o "criterio de disparo": 10
|
|
108
|
-
*
|
|
109
|
-
* @param {object} aprendizaje { titulo, contenido, tipo? }
|
|
110
|
-
* @returns {number}
|
|
111
|
-
*/
|
|
112
|
-
function scoreAprendizaje(aprendizaje) {
|
|
113
|
-
let score = 0;
|
|
114
|
-
const titulo = String(aprendizaje?.titulo || '');
|
|
115
|
-
const contenido = String(aprendizaje?.contenido || '');
|
|
116
|
-
|
|
117
|
-
if (titulo.length >= 10) score += 20;
|
|
118
|
-
if (/^\[\d{4}-\d{2}-\d{2}\]/.test(titulo)) score += 10;
|
|
119
|
-
if (contenido.length >= 100) score += 20;
|
|
120
|
-
if (contenido.length >= 300) score += 10;
|
|
121
|
-
|
|
122
|
-
const tipos = ['decisión', 'patrón', 'anti-patrón', 'bug-fix', 'descubrimiento', 'gotcha'];
|
|
123
|
-
if (aprendizaje?.tipo && tipos.some(t => String(aprendizaje.tipo).includes(t))) {
|
|
124
|
-
score += 15;
|
|
125
|
-
}
|
|
126
|
-
|
|
127
|
-
if (/`[^`]+\.(js|ts|md|py|json|yaml)`|`[^`]+\/[^`]+`/.test(contenido)) score += 15;
|
|
128
|
-
if (/trigger|criterio de disparo|cuando .+ entonces/i.test(contenido)) score += 10;
|
|
129
|
-
|
|
130
|
-
return clamp(score, 0, 100);
|
|
131
|
-
}
|
|
132
|
-
|
|
133
|
-
/**
|
|
134
|
-
* Score de calidad de un instinto.
|
|
135
|
-
* Específico de swl-ses.
|
|
136
|
-
*
|
|
137
|
-
* Distribución:
|
|
138
|
-
* - pattern presente: 25
|
|
139
|
-
* - pattern ≥ 30 chars: +10
|
|
140
|
-
* - confidence ∈ [0, 1]: 15
|
|
141
|
-
* - status válido (active|degraded|archived): 10
|
|
142
|
-
* - source_sessions o source_agents declarado: 15
|
|
143
|
-
* - evidence_count ≥ 1: 15
|
|
144
|
-
* - last_validated_at presente: 10
|
|
145
|
-
*
|
|
146
|
-
* @param {object} instinto
|
|
147
|
-
* @returns {number}
|
|
148
|
-
*/
|
|
149
|
-
function scoreInstinto(instinto) {
|
|
150
|
-
let score = 0;
|
|
151
|
-
const pattern = String(instinto?.pattern || '');
|
|
152
|
-
if (pattern.length > 0) score += 25;
|
|
153
|
-
if (pattern.length >= 30) score += 10;
|
|
154
|
-
if (typeof instinto?.confidence === 'number' && instinto.confidence >= 0 && instinto.confidence <= 1) score += 15;
|
|
155
|
-
if (['active', 'degraded', 'archived'].includes(instinto?.status)) score += 10;
|
|
156
|
-
if ((Array.isArray(instinto?.source_sessions) && instinto.source_sessions.length > 0) ||
|
|
157
|
-
(Array.isArray(instinto?.source_agents) && instinto.source_agents.length > 0)) score += 15;
|
|
158
|
-
if (typeof instinto?.evidence_count === 'number' && instinto.evidence_count >= 1) score += 15;
|
|
159
|
-
if (instinto?.last_validated_at || instinto?.last_validated) score += 10;
|
|
160
|
-
return clamp(score, 0, 100);
|
|
161
|
-
}
|
|
162
|
-
|
|
163
|
-
// ── exports ───────────────────────────────────────────────────────────────────
|
|
164
|
-
|
|
165
|
-
module.exports = {
|
|
166
|
-
scoreObservacion,
|
|
167
|
-
scoreResumen,
|
|
168
|
-
scoreRelevanciaContexto,
|
|
169
|
-
scoreAprendizaje,
|
|
170
|
-
scoreInstinto,
|
|
171
|
-
};
|
|
1
|
+
'use strict';
|
|
2
|
+
|
|
3
|
+
/**
|
|
4
|
+
* eval-quality.js — Métricas de calidad para outputs estructurados de SWL.
|
|
5
|
+
*
|
|
6
|
+
* Patrón adoptado de `temp/agentmemory-main/src/eval/quality.ts`. Adaptado a
|
|
7
|
+
* swl-ses: scoring de aprendizajes, instintos, y resultados de búsqueda
|
|
8
|
+
* memoria.
|
|
9
|
+
*
|
|
10
|
+
* Cada función devuelve un score en [0, 100]. Los puntos se asignan por
|
|
11
|
+
* "presencia y calidad de campos clave" — un output con todos los campos
|
|
12
|
+
* tiene score 100, un output trivial tiene score bajo.
|
|
13
|
+
*
|
|
14
|
+
* Funciones puras zero-deps.
|
|
15
|
+
*
|
|
16
|
+
* @module scripts/lib/eval-quality
|
|
17
|
+
*/
|
|
18
|
+
|
|
19
|
+
// ── helpers ───────────────────────────────────────────────────────────────────
|
|
20
|
+
|
|
21
|
+
function clamp(n, min, max) {
|
|
22
|
+
if (Number.isNaN(n)) return min;
|
|
23
|
+
return Math.max(min, Math.min(max, n));
|
|
24
|
+
}
|
|
25
|
+
|
|
26
|
+
// ── scoring de outputs ────────────────────────────────────────────────────────
|
|
27
|
+
|
|
28
|
+
/**
|
|
29
|
+
* Score de calidad de una observación comprimida.
|
|
30
|
+
* Adaptado de scoreCompression() de agentmemory.
|
|
31
|
+
*
|
|
32
|
+
* Distribución de puntos:
|
|
33
|
+
* - facts presentes (no vacíos): 25
|
|
34
|
+
* - facts ≥ 3: +10
|
|
35
|
+
* - narrative ≥ 20 chars: 20
|
|
36
|
+
* - narrative ≥ 50 chars: +5
|
|
37
|
+
* - title 5-120 chars: 15
|
|
38
|
+
* - concepts presentes: 15
|
|
39
|
+
* - importance ∈ [1, 10]: 10
|
|
40
|
+
*
|
|
41
|
+
* @param {object} obs
|
|
42
|
+
* @returns {number} en [0, 100]
|
|
43
|
+
*/
|
|
44
|
+
function scoreObservacion(obs) {
|
|
45
|
+
let score = 0;
|
|
46
|
+
if (Array.isArray(obs?.facts) && obs.facts.length > 0) score += 25;
|
|
47
|
+
if (Array.isArray(obs?.facts) && obs.facts.length >= 3) score += 10;
|
|
48
|
+
if (typeof obs?.narrative === 'string' && obs.narrative.length >= 20) score += 20;
|
|
49
|
+
if (typeof obs?.narrative === 'string' && obs.narrative.length >= 50) score += 5;
|
|
50
|
+
if (typeof obs?.title === 'string' && obs.title.length >= 5 && obs.title.length <= 120) score += 15;
|
|
51
|
+
if (Array.isArray(obs?.concepts) && obs.concepts.length > 0) score += 15;
|
|
52
|
+
if (typeof obs?.importance === 'number' && obs.importance >= 1 && obs.importance <= 10) score += 10;
|
|
53
|
+
return clamp(score, 0, 100);
|
|
54
|
+
}
|
|
55
|
+
|
|
56
|
+
/**
|
|
57
|
+
* Score de calidad de un resumen de sesión.
|
|
58
|
+
* Adaptado de scoreSummary() de agentmemory.
|
|
59
|
+
*
|
|
60
|
+
* @param {object} summary
|
|
61
|
+
* @returns {number} en [0, 100]
|
|
62
|
+
*/
|
|
63
|
+
function scoreResumen(summary) {
|
|
64
|
+
let score = 0;
|
|
65
|
+
if (typeof summary?.title === 'string' && summary.title.length >= 5) score += 20;
|
|
66
|
+
if (typeof summary?.narrative === 'string' && summary.narrative.length >= 20) score += 25;
|
|
67
|
+
if (typeof summary?.narrative === 'string' && summary.narrative.length >= 100) score += 5;
|
|
68
|
+
if (Array.isArray(summary?.keyDecisions) && summary.keyDecisions.length > 0) score += 20;
|
|
69
|
+
if (Array.isArray(summary?.filesModified) && summary.filesModified.length > 0) score += 15;
|
|
70
|
+
if (Array.isArray(summary?.concepts) && summary.concepts.length > 0) score += 15;
|
|
71
|
+
return clamp(score, 0, 100);
|
|
72
|
+
}
|
|
73
|
+
|
|
74
|
+
/**
|
|
75
|
+
* Score de relevancia de un contexto inyectado.
|
|
76
|
+
* Adaptado de scoreContextRelevance() de agentmemory.
|
|
77
|
+
*
|
|
78
|
+
* @param {string} context - Texto del contexto.
|
|
79
|
+
* @param {string} project - Nombre del proyecto.
|
|
80
|
+
* @returns {number} en [0, 100]
|
|
81
|
+
*/
|
|
82
|
+
function scoreRelevanciaContexto(context, project) {
|
|
83
|
+
if (typeof context !== 'string') return 0;
|
|
84
|
+
let score = 0;
|
|
85
|
+
if (context.length > 0) score += 20;
|
|
86
|
+
if (project && context.toLowerCase().includes(String(project).toLowerCase())) score += 20;
|
|
87
|
+
if (context.includes('<')) score += 15;
|
|
88
|
+
const sectionCount = (context.match(/<\w+>/g) || []).length;
|
|
89
|
+
if (sectionCount >= 2) score += 15;
|
|
90
|
+
if (sectionCount >= 4) score += 10;
|
|
91
|
+
if (context.length >= 100) score += 10;
|
|
92
|
+
if (context.length >= 500) score += 10;
|
|
93
|
+
return clamp(score, 0, 100);
|
|
94
|
+
}
|
|
95
|
+
|
|
96
|
+
/**
|
|
97
|
+
* Score de calidad de un aprendizaje SWL extraído de una sesión.
|
|
98
|
+
* Específico de swl-ses (no en agentmemory).
|
|
99
|
+
*
|
|
100
|
+
* Distribución:
|
|
101
|
+
* - título no vacío y descriptivo (≥ 10 chars): 20
|
|
102
|
+
* - título empieza con fecha [YYYY-MM-DD]: 10
|
|
103
|
+
* - cuerpo ≥ 100 chars: 20
|
|
104
|
+
* - cuerpo ≥ 300 chars: +10
|
|
105
|
+
* - tipo identificado (decisión|patrón|...): 15
|
|
106
|
+
* - menciona archivo o regla concreta: 15
|
|
107
|
+
* - tiene "trigger" o "criterio de disparo": 10
|
|
108
|
+
*
|
|
109
|
+
* @param {object} aprendizaje { titulo, contenido, tipo? }
|
|
110
|
+
* @returns {number}
|
|
111
|
+
*/
|
|
112
|
+
function scoreAprendizaje(aprendizaje) {
|
|
113
|
+
let score = 0;
|
|
114
|
+
const titulo = String(aprendizaje?.titulo || '');
|
|
115
|
+
const contenido = String(aprendizaje?.contenido || '');
|
|
116
|
+
|
|
117
|
+
if (titulo.length >= 10) score += 20;
|
|
118
|
+
if (/^\[\d{4}-\d{2}-\d{2}\]/.test(titulo)) score += 10;
|
|
119
|
+
if (contenido.length >= 100) score += 20;
|
|
120
|
+
if (contenido.length >= 300) score += 10;
|
|
121
|
+
|
|
122
|
+
const tipos = ['decisión', 'patrón', 'anti-patrón', 'bug-fix', 'descubrimiento', 'gotcha'];
|
|
123
|
+
if (aprendizaje?.tipo && tipos.some(t => String(aprendizaje.tipo).includes(t))) {
|
|
124
|
+
score += 15;
|
|
125
|
+
}
|
|
126
|
+
|
|
127
|
+
if (/`[^`]+\.(js|ts|md|py|json|yaml)`|`[^`]+\/[^`]+`/.test(contenido)) score += 15;
|
|
128
|
+
if (/trigger|criterio de disparo|cuando .+ entonces/i.test(contenido)) score += 10;
|
|
129
|
+
|
|
130
|
+
return clamp(score, 0, 100);
|
|
131
|
+
}
|
|
132
|
+
|
|
133
|
+
/**
|
|
134
|
+
* Score de calidad de un instinto.
|
|
135
|
+
* Específico de swl-ses.
|
|
136
|
+
*
|
|
137
|
+
* Distribución:
|
|
138
|
+
* - pattern presente: 25
|
|
139
|
+
* - pattern ≥ 30 chars: +10
|
|
140
|
+
* - confidence ∈ [0, 1]: 15
|
|
141
|
+
* - status válido (active|degraded|archived): 10
|
|
142
|
+
* - source_sessions o source_agents declarado: 15
|
|
143
|
+
* - evidence_count ≥ 1: 15
|
|
144
|
+
* - last_validated_at presente: 10
|
|
145
|
+
*
|
|
146
|
+
* @param {object} instinto
|
|
147
|
+
* @returns {number}
|
|
148
|
+
*/
|
|
149
|
+
function scoreInstinto(instinto) {
|
|
150
|
+
let score = 0;
|
|
151
|
+
const pattern = String(instinto?.pattern || '');
|
|
152
|
+
if (pattern.length > 0) score += 25;
|
|
153
|
+
if (pattern.length >= 30) score += 10;
|
|
154
|
+
if (typeof instinto?.confidence === 'number' && instinto.confidence >= 0 && instinto.confidence <= 1) score += 15;
|
|
155
|
+
if (['active', 'degraded', 'archived'].includes(instinto?.status)) score += 10;
|
|
156
|
+
if ((Array.isArray(instinto?.source_sessions) && instinto.source_sessions.length > 0) ||
|
|
157
|
+
(Array.isArray(instinto?.source_agents) && instinto.source_agents.length > 0)) score += 15;
|
|
158
|
+
if (typeof instinto?.evidence_count === 'number' && instinto.evidence_count >= 1) score += 15;
|
|
159
|
+
if (instinto?.last_validated_at || instinto?.last_validated) score += 10;
|
|
160
|
+
return clamp(score, 0, 100);
|
|
161
|
+
}
|
|
162
|
+
|
|
163
|
+
// ── exports ───────────────────────────────────────────────────────────────────
|
|
164
|
+
|
|
165
|
+
module.exports = {
|
|
166
|
+
scoreObservacion,
|
|
167
|
+
scoreResumen,
|
|
168
|
+
scoreRelevanciaContexto,
|
|
169
|
+
scoreAprendizaje,
|
|
170
|
+
scoreInstinto,
|
|
171
|
+
};
|
|
@@ -1,144 +1,144 @@
|
|
|
1
|
-
'use strict';
|
|
2
|
-
|
|
3
|
-
/**
|
|
4
|
-
* eval-schemas.js — Schemas JSON-lite para evaluación de outputs SWL.
|
|
5
|
-
*
|
|
6
|
-
* Patrón adoptado de `temp/agentmemory-main/src/eval/schemas.ts`. Adaptado a
|
|
7
|
-
* swl-ses: sin Zod (sería dep externa). Uso JSON Schema-lite con validador
|
|
8
|
-
* propio en `eval-validator.js`. Funciones puras zero-deps.
|
|
9
|
-
*
|
|
10
|
-
* Cada schema describe la estructura esperada de un output evaluable. El
|
|
11
|
-
* validador devuelve `{ valid, errors[] }` para que el caller decida.
|
|
12
|
-
*
|
|
13
|
-
* @module scripts/lib/eval-schemas
|
|
14
|
-
*/
|
|
15
|
-
|
|
16
|
-
// ── tipos enumerados ──────────────────────────────────────────────────────────
|
|
17
|
-
|
|
18
|
-
const TIPOS_OBSERVACION = [
|
|
19
|
-
'file_read', 'file_write', 'file_edit',
|
|
20
|
-
'command_run', 'search', 'web_fetch',
|
|
21
|
-
'conversation', 'error', 'decision',
|
|
22
|
-
'discovery', 'subagent', 'notification',
|
|
23
|
-
'task', 'other',
|
|
24
|
-
];
|
|
25
|
-
|
|
26
|
-
const TIPOS_MEMORIA = [
|
|
27
|
-
'pattern', 'preference', 'architecture',
|
|
28
|
-
'bug', 'workflow', 'fact',
|
|
29
|
-
];
|
|
30
|
-
|
|
31
|
-
const TIPOS_RELACION = [
|
|
32
|
-
'supersedes', 'extends', 'derives', 'contradicts', 'related',
|
|
33
|
-
];
|
|
34
|
-
|
|
35
|
-
// ── schemas JSON-lite ─────────────────────────────────────────────────────────
|
|
36
|
-
|
|
37
|
-
/**
|
|
38
|
-
* Schema para output de compresión de observación.
|
|
39
|
-
* Estructura compatible con `CompressOutputSchema` de agentmemory.
|
|
40
|
-
*/
|
|
41
|
-
const COMPRESS_OUTPUT_SCHEMA = {
|
|
42
|
-
type: 'object',
|
|
43
|
-
required: ['type', 'title', 'facts', 'narrative', 'concepts', 'files', 'importance'],
|
|
44
|
-
properties: {
|
|
45
|
-
type: { type: 'string', enum: TIPOS_OBSERVACION },
|
|
46
|
-
title: { type: 'string', minLength: 1, maxLength: 120 },
|
|
47
|
-
subtitle: { type: 'string' },
|
|
48
|
-
facts: { type: 'array', minItems: 1, items: { type: 'string' } },
|
|
49
|
-
narrative: { type: 'string', minLength: 10 },
|
|
50
|
-
concepts: { type: 'array', items: { type: 'string' } },
|
|
51
|
-
files: { type: 'array', items: { type: 'string' } },
|
|
52
|
-
importance: { type: 'integer', minimum: 1, maximum: 10 },
|
|
53
|
-
},
|
|
54
|
-
};
|
|
55
|
-
|
|
56
|
-
/**
|
|
57
|
-
* Schema para output de resumen de sesión.
|
|
58
|
-
*/
|
|
59
|
-
const SUMMARY_OUTPUT_SCHEMA = {
|
|
60
|
-
type: 'object',
|
|
61
|
-
required: ['title', 'narrative', 'keyDecisions', 'filesModified', 'concepts'],
|
|
62
|
-
properties: {
|
|
63
|
-
title: { type: 'string', minLength: 1 },
|
|
64
|
-
narrative: { type: 'string', minLength: 20 },
|
|
65
|
-
keyDecisions: { type: 'array', items: { type: 'string' } },
|
|
66
|
-
filesModified: { type: 'array', items: { type: 'string' } },
|
|
67
|
-
concepts: { type: 'array', items: { type: 'string' } },
|
|
68
|
-
},
|
|
69
|
-
};
|
|
70
|
-
|
|
71
|
-
/**
|
|
72
|
-
* Schema para input de búsqueda.
|
|
73
|
-
*/
|
|
74
|
-
const SEARCH_INPUT_SCHEMA = {
|
|
75
|
-
type: 'object',
|
|
76
|
-
required: ['query'],
|
|
77
|
-
properties: {
|
|
78
|
-
query: { type: 'string', minLength: 1 },
|
|
79
|
-
limit: { type: 'integer', minimum: 1 },
|
|
80
|
-
},
|
|
81
|
-
};
|
|
82
|
-
|
|
83
|
-
/**
|
|
84
|
-
* Schema para input de "remember" (guardar memoria).
|
|
85
|
-
*/
|
|
86
|
-
const REMEMBER_INPUT_SCHEMA = {
|
|
87
|
-
type: 'object',
|
|
88
|
-
required: ['content'],
|
|
89
|
-
properties: {
|
|
90
|
-
content: { type: 'string', minLength: 1 },
|
|
91
|
-
type: { type: 'string', enum: TIPOS_MEMORIA },
|
|
92
|
-
concepts: { type: 'array', items: { type: 'string' } },
|
|
93
|
-
files: { type: 'array', items: { type: 'string' } },
|
|
94
|
-
},
|
|
95
|
-
};
|
|
96
|
-
|
|
97
|
-
/**
|
|
98
|
-
* Schema para resultado de evaluación.
|
|
99
|
-
*/
|
|
100
|
-
const EVAL_RESULT_SCHEMA = {
|
|
101
|
-
type: 'object',
|
|
102
|
-
required: ['valid', 'qualityScore', 'latencyMs', 'functionId'],
|
|
103
|
-
properties: {
|
|
104
|
-
valid: { type: 'boolean' },
|
|
105
|
-
errors: { type: 'array', items: { type: 'string' } },
|
|
106
|
-
qualityScore: { type: 'number', minimum: 0, maximum: 100 },
|
|
107
|
-
latencyMs: { type: 'number', minimum: 0 },
|
|
108
|
-
functionId: { type: 'string', minLength: 1 },
|
|
109
|
-
metadata: { type: 'object' },
|
|
110
|
-
},
|
|
111
|
-
};
|
|
112
|
-
|
|
113
|
-
/**
|
|
114
|
-
* Schema para resultado de búsqueda en memoria SWL.
|
|
115
|
-
*/
|
|
116
|
-
const MEMORY_SEARCH_RESULT_SCHEMA = {
|
|
117
|
-
type: 'object',
|
|
118
|
-
required: ['id', 'tipo', 'titulo', 'fecha', 'relevancia'],
|
|
119
|
-
properties: {
|
|
120
|
-
id: { type: 'string', minLength: 1 },
|
|
121
|
-
tipo: { type: 'string', enum: ['aprendizaje', 'sesion', 'instinto'] },
|
|
122
|
-
titulo: { type: 'string' },
|
|
123
|
-
fecha: { type: 'string' },
|
|
124
|
-
relevancia: { type: 'number', minimum: 0, maximum: 1 },
|
|
125
|
-
combinedScore: { type: 'number', minimum: 0 },
|
|
126
|
-
confidence: { type: 'number', minimum: 0, maximum: 1 },
|
|
127
|
-
},
|
|
128
|
-
};
|
|
129
|
-
|
|
130
|
-
// ── exports ───────────────────────────────────────────────────────────────────
|
|
131
|
-
|
|
132
|
-
module.exports = {
|
|
133
|
-
// Schemas
|
|
134
|
-
COMPRESS_OUTPUT_SCHEMA,
|
|
135
|
-
SUMMARY_OUTPUT_SCHEMA,
|
|
136
|
-
SEARCH_INPUT_SCHEMA,
|
|
137
|
-
REMEMBER_INPUT_SCHEMA,
|
|
138
|
-
EVAL_RESULT_SCHEMA,
|
|
139
|
-
MEMORY_SEARCH_RESULT_SCHEMA,
|
|
140
|
-
// Enums
|
|
141
|
-
TIPOS_OBSERVACION,
|
|
142
|
-
TIPOS_MEMORIA,
|
|
143
|
-
TIPOS_RELACION,
|
|
144
|
-
};
|
|
1
|
+
'use strict';
|
|
2
|
+
|
|
3
|
+
/**
|
|
4
|
+
* eval-schemas.js — Schemas JSON-lite para evaluación de outputs SWL.
|
|
5
|
+
*
|
|
6
|
+
* Patrón adoptado de `temp/agentmemory-main/src/eval/schemas.ts`. Adaptado a
|
|
7
|
+
* swl-ses: sin Zod (sería dep externa). Uso JSON Schema-lite con validador
|
|
8
|
+
* propio en `eval-validator.js`. Funciones puras zero-deps.
|
|
9
|
+
*
|
|
10
|
+
* Cada schema describe la estructura esperada de un output evaluable. El
|
|
11
|
+
* validador devuelve `{ valid, errors[] }` para que el caller decida.
|
|
12
|
+
*
|
|
13
|
+
* @module scripts/lib/eval-schemas
|
|
14
|
+
*/
|
|
15
|
+
|
|
16
|
+
// ── tipos enumerados ──────────────────────────────────────────────────────────
|
|
17
|
+
|
|
18
|
+
const TIPOS_OBSERVACION = [
|
|
19
|
+
'file_read', 'file_write', 'file_edit',
|
|
20
|
+
'command_run', 'search', 'web_fetch',
|
|
21
|
+
'conversation', 'error', 'decision',
|
|
22
|
+
'discovery', 'subagent', 'notification',
|
|
23
|
+
'task', 'other',
|
|
24
|
+
];
|
|
25
|
+
|
|
26
|
+
const TIPOS_MEMORIA = [
|
|
27
|
+
'pattern', 'preference', 'architecture',
|
|
28
|
+
'bug', 'workflow', 'fact',
|
|
29
|
+
];
|
|
30
|
+
|
|
31
|
+
const TIPOS_RELACION = [
|
|
32
|
+
'supersedes', 'extends', 'derives', 'contradicts', 'related',
|
|
33
|
+
];
|
|
34
|
+
|
|
35
|
+
// ── schemas JSON-lite ─────────────────────────────────────────────────────────
|
|
36
|
+
|
|
37
|
+
/**
|
|
38
|
+
* Schema para output de compresión de observación.
|
|
39
|
+
* Estructura compatible con `CompressOutputSchema` de agentmemory.
|
|
40
|
+
*/
|
|
41
|
+
const COMPRESS_OUTPUT_SCHEMA = {
|
|
42
|
+
type: 'object',
|
|
43
|
+
required: ['type', 'title', 'facts', 'narrative', 'concepts', 'files', 'importance'],
|
|
44
|
+
properties: {
|
|
45
|
+
type: { type: 'string', enum: TIPOS_OBSERVACION },
|
|
46
|
+
title: { type: 'string', minLength: 1, maxLength: 120 },
|
|
47
|
+
subtitle: { type: 'string' },
|
|
48
|
+
facts: { type: 'array', minItems: 1, items: { type: 'string' } },
|
|
49
|
+
narrative: { type: 'string', minLength: 10 },
|
|
50
|
+
concepts: { type: 'array', items: { type: 'string' } },
|
|
51
|
+
files: { type: 'array', items: { type: 'string' } },
|
|
52
|
+
importance: { type: 'integer', minimum: 1, maximum: 10 },
|
|
53
|
+
},
|
|
54
|
+
};
|
|
55
|
+
|
|
56
|
+
/**
|
|
57
|
+
* Schema para output de resumen de sesión.
|
|
58
|
+
*/
|
|
59
|
+
const SUMMARY_OUTPUT_SCHEMA = {
|
|
60
|
+
type: 'object',
|
|
61
|
+
required: ['title', 'narrative', 'keyDecisions', 'filesModified', 'concepts'],
|
|
62
|
+
properties: {
|
|
63
|
+
title: { type: 'string', minLength: 1 },
|
|
64
|
+
narrative: { type: 'string', minLength: 20 },
|
|
65
|
+
keyDecisions: { type: 'array', items: { type: 'string' } },
|
|
66
|
+
filesModified: { type: 'array', items: { type: 'string' } },
|
|
67
|
+
concepts: { type: 'array', items: { type: 'string' } },
|
|
68
|
+
},
|
|
69
|
+
};
|
|
70
|
+
|
|
71
|
+
/**
|
|
72
|
+
* Schema para input de búsqueda.
|
|
73
|
+
*/
|
|
74
|
+
const SEARCH_INPUT_SCHEMA = {
|
|
75
|
+
type: 'object',
|
|
76
|
+
required: ['query'],
|
|
77
|
+
properties: {
|
|
78
|
+
query: { type: 'string', minLength: 1 },
|
|
79
|
+
limit: { type: 'integer', minimum: 1 },
|
|
80
|
+
},
|
|
81
|
+
};
|
|
82
|
+
|
|
83
|
+
/**
|
|
84
|
+
* Schema para input de "remember" (guardar memoria).
|
|
85
|
+
*/
|
|
86
|
+
const REMEMBER_INPUT_SCHEMA = {
|
|
87
|
+
type: 'object',
|
|
88
|
+
required: ['content'],
|
|
89
|
+
properties: {
|
|
90
|
+
content: { type: 'string', minLength: 1 },
|
|
91
|
+
type: { type: 'string', enum: TIPOS_MEMORIA },
|
|
92
|
+
concepts: { type: 'array', items: { type: 'string' } },
|
|
93
|
+
files: { type: 'array', items: { type: 'string' } },
|
|
94
|
+
},
|
|
95
|
+
};
|
|
96
|
+
|
|
97
|
+
/**
|
|
98
|
+
* Schema para resultado de evaluación.
|
|
99
|
+
*/
|
|
100
|
+
const EVAL_RESULT_SCHEMA = {
|
|
101
|
+
type: 'object',
|
|
102
|
+
required: ['valid', 'qualityScore', 'latencyMs', 'functionId'],
|
|
103
|
+
properties: {
|
|
104
|
+
valid: { type: 'boolean' },
|
|
105
|
+
errors: { type: 'array', items: { type: 'string' } },
|
|
106
|
+
qualityScore: { type: 'number', minimum: 0, maximum: 100 },
|
|
107
|
+
latencyMs: { type: 'number', minimum: 0 },
|
|
108
|
+
functionId: { type: 'string', minLength: 1 },
|
|
109
|
+
metadata: { type: 'object' },
|
|
110
|
+
},
|
|
111
|
+
};
|
|
112
|
+
|
|
113
|
+
/**
|
|
114
|
+
* Schema para resultado de búsqueda en memoria SWL.
|
|
115
|
+
*/
|
|
116
|
+
const MEMORY_SEARCH_RESULT_SCHEMA = {
|
|
117
|
+
type: 'object',
|
|
118
|
+
required: ['id', 'tipo', 'titulo', 'fecha', 'relevancia'],
|
|
119
|
+
properties: {
|
|
120
|
+
id: { type: 'string', minLength: 1 },
|
|
121
|
+
tipo: { type: 'string', enum: ['aprendizaje', 'sesion', 'instinto'] },
|
|
122
|
+
titulo: { type: 'string' },
|
|
123
|
+
fecha: { type: 'string' },
|
|
124
|
+
relevancia: { type: 'number', minimum: 0, maximum: 1 },
|
|
125
|
+
combinedScore: { type: 'number', minimum: 0 },
|
|
126
|
+
confidence: { type: 'number', minimum: 0, maximum: 1 },
|
|
127
|
+
},
|
|
128
|
+
};
|
|
129
|
+
|
|
130
|
+
// ── exports ───────────────────────────────────────────────────────────────────
|
|
131
|
+
|
|
132
|
+
module.exports = {
|
|
133
|
+
// Schemas
|
|
134
|
+
COMPRESS_OUTPUT_SCHEMA,
|
|
135
|
+
SUMMARY_OUTPUT_SCHEMA,
|
|
136
|
+
SEARCH_INPUT_SCHEMA,
|
|
137
|
+
REMEMBER_INPUT_SCHEMA,
|
|
138
|
+
EVAL_RESULT_SCHEMA,
|
|
139
|
+
MEMORY_SEARCH_RESULT_SCHEMA,
|
|
140
|
+
// Enums
|
|
141
|
+
TIPOS_OBSERVACION,
|
|
142
|
+
TIPOS_MEMORIA,
|
|
143
|
+
TIPOS_RELACION,
|
|
144
|
+
};
|