@lorekit/cli 1.59.1 → 1.61.0
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/README.md +55 -4
- package/bin/lorekit.mjs +47 -5
- package/package.json +1 -1
- package/src/commands/dedupe.mjs +20 -6
- package/src/commands/invariants.mjs +304 -0
- package/src/commands/obligations.mjs +28 -5
- package/src/commands.mjs +2 -0
- package/src/shared/candidates-pure.mjs +110 -0
- package/src/shared/completions.mjs +5 -1
- package/src/shared/deeplink-pure.mjs +18 -3
- package/src/shared/obligations-map.mjs +79 -23
- package/src/shared/obligations-pure.mjs +27 -10
- package/src/shared/recurrence-clusters.mjs +161 -0
- package/src/surfaces.generated.mjs +516 -0
|
@@ -0,0 +1,110 @@
|
|
|
1
|
+
// Pure core for `lorekit invariants candidates` — ranks near-duplicate
|
|
2
|
+
// clusters (the same Jaccard heuristic `dedupe` uses) as merge candidates for
|
|
3
|
+
// a hand-written compile-pipeline invariant declaration.
|
|
4
|
+
//
|
|
5
|
+
// This module is deliberately read-only in a stronger sense than "no writes":
|
|
6
|
+
// it never CLASSIFIES anything. It surfaces what a lesson already states about
|
|
7
|
+
// itself (a parsed meta comment, a seen_count, a resolved recurrence class)
|
|
8
|
+
// and ranks it — it does not decide whether a trigger-context is mechanically
|
|
9
|
+
// detectable, and it does not validate `status` against a fixed vocabulary
|
|
10
|
+
// (none is defined anywhere in this codebase yet). That judgment is the human
|
|
11
|
+
// step the compile pipeline's "never auto-compile, never auto-gate" rule
|
|
12
|
+
// protects; a scan that started making those calls would be the thing the
|
|
13
|
+
// rule exists to prevent.
|
|
14
|
+
//
|
|
15
|
+
// Operates on already-enriched cluster members — `{ scope, key, seenCount,
|
|
16
|
+
// value }` — leaving the store I/O and the seen_count/meta correlation to the
|
|
17
|
+
// command (`../commands/invariants.mjs`). Zero-dep, total: malformed input
|
|
18
|
+
// degrades to the empty/default case rather than throwing.
|
|
19
|
+
|
|
20
|
+
const META_COMMENT_RE = /<!--\s*meta:([\s\S]*?)-->/;
|
|
21
|
+
const META_FIELD_RE = /([\w-]+)=("(?:[^"\\]|\\.)*"|\S+)/g;
|
|
22
|
+
|
|
23
|
+
/**
|
|
24
|
+
* Extract the `<!-- meta: seen_count=1 status=active expires=<iso>
|
|
25
|
+
* trigger-context="<signal>" -->` convention documented in the lorekit-setup
|
|
26
|
+
* skill's `self-improvement-loops.md`. Nothing in the CLI parses this
|
|
27
|
+
* convention today; this is read-only extraction for a human's judgment, not
|
|
28
|
+
* a schema the scan enforces. Absent or malformed input yields `{}`, never a
|
|
29
|
+
* throw — a lesson written before (or without) the convention is not an
|
|
30
|
+
* error, just a candidate with no meta fields.
|
|
31
|
+
*/
|
|
32
|
+
export function parseMetaComment(value) {
|
|
33
|
+
if (typeof value !== 'string') return {};
|
|
34
|
+
const m = META_COMMENT_RE.exec(value);
|
|
35
|
+
if (!m) return {};
|
|
36
|
+
const out = {};
|
|
37
|
+
META_FIELD_RE.lastIndex = 0;
|
|
38
|
+
let mm;
|
|
39
|
+
while ((mm = META_FIELD_RE.exec(m[1]))) {
|
|
40
|
+
const [, k, rawV] = mm;
|
|
41
|
+
out[k] =
|
|
42
|
+
rawV.startsWith('"') && rawV.endsWith('"') ? rawV.slice(1, -1).replace(/\\"/g, '"') : rawV;
|
|
43
|
+
}
|
|
44
|
+
return out;
|
|
45
|
+
}
|
|
46
|
+
|
|
47
|
+
function totalSeen(members) {
|
|
48
|
+
return (members || []).reduce((n, m) => n + (Number.isFinite(m.seenCount) ? m.seenCount : 0), 0);
|
|
49
|
+
}
|
|
50
|
+
|
|
51
|
+
function distinctScopeCount(members) {
|
|
52
|
+
return new Set((members || []).map((m) => m.scope)).size;
|
|
53
|
+
}
|
|
54
|
+
|
|
55
|
+
/**
|
|
56
|
+
* A cluster is a candidate when its recurrence signal is real: either the
|
|
57
|
+
* summed `seenCount` across members crosses `minSeenCount` (default 3 — the
|
|
58
|
+
* kickoff's "seen_count >= 3" criterion, applied to the SUM across the
|
|
59
|
+
* cluster's members rather than any single one, since the whole pitch of a
|
|
60
|
+
* candidate is "these N sightings are really one entry"), or a member's own
|
|
61
|
+
* meta comment already declares a non-"active" status.
|
|
62
|
+
*/
|
|
63
|
+
export function isCandidate(members, { minSeenCount = 3 } = {}) {
|
|
64
|
+
if (totalSeen(members) >= minSeenCount) return true;
|
|
65
|
+
return (members || []).some((m) => {
|
|
66
|
+
const status = parseMetaComment(m.value).status;
|
|
67
|
+
return typeof status === 'string' && status.length > 0 && status !== 'active';
|
|
68
|
+
});
|
|
69
|
+
}
|
|
70
|
+
|
|
71
|
+
/** Recurrence × distinct scopes — the ranking signal the kickoff names. */
|
|
72
|
+
export function scoreCandidate(members) {
|
|
73
|
+
return totalSeen(members) * distinctScopeCount(members);
|
|
74
|
+
}
|
|
75
|
+
|
|
76
|
+
/**
|
|
77
|
+
* Filter a set of dedupe-style clusters down to candidates and rank them,
|
|
78
|
+
* highest score first (ties broken by member count, then the first member's
|
|
79
|
+
* `scope::key` for determinism). Each input cluster is
|
|
80
|
+
* `{ members: [{scope,key,seenCount,value}], size, minSimilarity?,
|
|
81
|
+
* maxSimilarity? }`. Never mutates the input.
|
|
82
|
+
*
|
|
83
|
+
* `resolveClass`, when supplied, is called with the cluster's raw members to
|
|
84
|
+
* attach a resolved recurrence class (the `dedupe` join, reused here) — a
|
|
85
|
+
* candidate already resolving to a known class is a stronger case ("this
|
|
86
|
+
* should join an existing invariant") than one that doesn't ("this might be a
|
|
87
|
+
* new class"). Optional so the pure ranking logic stays testable without it.
|
|
88
|
+
*/
|
|
89
|
+
export function rankCandidates(clusters, { minSeenCount = 3, resolveClass } = {}) {
|
|
90
|
+
return (clusters || [])
|
|
91
|
+
.filter((cl) => isCandidate(cl.members, { minSeenCount }))
|
|
92
|
+
.map((cl) => {
|
|
93
|
+
const members = (cl.members || []).map((m) => ({ ...m, meta: parseMetaComment(m.value) }));
|
|
94
|
+
return {
|
|
95
|
+
members,
|
|
96
|
+
size: cl.size ?? members.length,
|
|
97
|
+
minSimilarity: cl.minSimilarity,
|
|
98
|
+
maxSimilarity: cl.maxSimilarity,
|
|
99
|
+
recurrenceClass: typeof resolveClass === 'function' ? resolveClass(cl.members) : null,
|
|
100
|
+
score: scoreCandidate(cl.members),
|
|
101
|
+
};
|
|
102
|
+
})
|
|
103
|
+
.sort((a, b) => {
|
|
104
|
+
if (b.score !== a.score) return b.score - a.score;
|
|
105
|
+
if (b.size !== a.size) return b.size - a.size;
|
|
106
|
+
const aKey = `${a.members[0]?.scope}::${a.members[0]?.key}`;
|
|
107
|
+
const bKey = `${b.members[0]?.scope}::${b.members[0]?.key}`;
|
|
108
|
+
return aKey < bKey ? -1 : aKey > bKey ? 1 : 0;
|
|
109
|
+
});
|
|
110
|
+
}
|
|
@@ -71,6 +71,8 @@ const FLAG = {
|
|
|
71
71
|
'retention-days': { desc: 'Only purge archived older than n days', arg: 'n' },
|
|
72
72
|
files: { desc: 'Changed files to check', arg: 'path' },
|
|
73
73
|
strict: { desc: 'Exit non-zero on any unmet obligation' },
|
|
74
|
+
'strict-all': { desc: 'Exit non-zero on ANY unmet obligation, advisory included' },
|
|
75
|
+
'min-seen-count': { desc: 'Minimum summed seen_count for a candidate', arg: 'n' },
|
|
74
76
|
'policy-id': { desc: 'Run/preview a saved policy', arg: 'id' },
|
|
75
77
|
'min-age-days': { desc: 'Match lessons at least n days old', arg: 'n' },
|
|
76
78
|
'unseen-days': { desc: 'Match lessons unseen for at least n days', arg: 'n' },
|
|
@@ -121,7 +123,9 @@ const COMMANDS = [
|
|
|
121
123
|
flags: ['dir', 'scope', 'threshold', 'cluster-by-key', 'json', 'endpoint', 'token', 'store'] },
|
|
122
124
|
{ name: 'obligations', summary: 'Check changed files against the surface-partner map',
|
|
123
125
|
positional: 'path',
|
|
124
|
-
flags: ['files', 'strict', 'json'] },
|
|
126
|
+
flags: ['files', 'strict', 'strict-all', 'json'] },
|
|
127
|
+
{ name: 'invariants', summary: "Compile pipeline's candidate scan (`invariants candidates`)",
|
|
128
|
+
flags: ['dir', 'scope', 'min-seen-count', 'json', 'endpoint', 'token', 'store'] },
|
|
125
129
|
{ name: 'link', summary: 'Print a shareable dashboard deep-link URL', aliases: ['url'],
|
|
126
130
|
positional: 'address',
|
|
127
131
|
flags: ['dir', 'scope', 'key', 'q', 'owner', 'tags', 'range', 'from', 'to', 'archived', 'base', 'json'] },
|
|
@@ -46,6 +46,9 @@ export const LORE_PARAM_DEFAULTS = {
|
|
|
46
46
|
// "unfiltered".
|
|
47
47
|
filters: null,
|
|
48
48
|
tags: [], // string[] — legacy label filter (AND across labels); [] means "no filter". Still READ by the app, superseded by `filters`
|
|
49
|
+
// RetentionConditions | null — the retention-preview trio (min age / unseen-for
|
|
50
|
+
// / seen-at-most). `null` means no narrowing; absent from the URL when unset.
|
|
51
|
+
retention: null,
|
|
49
52
|
// 'active' | 'archived' | 'expiring' | null — the Explorer's Status control.
|
|
50
53
|
// `null`, NOT 'active', is the default for `filters`' reason: the app has to
|
|
51
54
|
// tell "absent" from an explicit choice, because an absent `status` falls back
|
|
@@ -59,9 +62,21 @@ export const LORE_PARAM_DEFAULTS = {
|
|
|
59
62
|
|
|
60
63
|
// A stable, readable param order (also makes URLs deterministic for tests).
|
|
61
64
|
// Mirrors the `useUrlState` call order in `LoreExplorer.tsx` (+ the `lesson`
|
|
62
|
-
// param last), so `filters` and `tags` sit between `owner` and `status
|
|
63
|
-
//
|
|
64
|
-
|
|
65
|
+
// param last), so `filters` and `tags` sit between `owner` and `status`, and
|
|
66
|
+
// `retention` sits between `tags` and `status`. `scope` precedes `lesson` so a
|
|
67
|
+
// lesson link reads `?scope=…&lesson=…`.
|
|
68
|
+
const PARAM_ORDER = [
|
|
69
|
+
'scope',
|
|
70
|
+
'q',
|
|
71
|
+
'range',
|
|
72
|
+
'owner',
|
|
73
|
+
'filters',
|
|
74
|
+
'tags',
|
|
75
|
+
'retention',
|
|
76
|
+
'status',
|
|
77
|
+
'archived',
|
|
78
|
+
'lesson',
|
|
79
|
+
];
|
|
65
80
|
|
|
66
81
|
// Strip trailing slashes from a base URL, falling back to the default when the
|
|
67
82
|
// input is empty/absent. Pure.
|
|
@@ -31,7 +31,32 @@
|
|
|
31
31
|
// // when a name-derived module's real partner
|
|
32
32
|
// // could live under either of two sibling
|
|
33
33
|
// // directories)
|
|
34
|
-
//
|
|
34
|
+
// state: 'advisory'|'gating'|'retired',
|
|
35
|
+
// // advisory: reported, gates nothing.
|
|
36
|
+
// // gating: fails --strict. An entry may only
|
|
37
|
+
// // be `gating` if it has a `guard` — an
|
|
38
|
+
// // independent source that already asserts
|
|
39
|
+
// // the partnership. Without one, the entry
|
|
40
|
+
// // asserts only its own author's belief, and
|
|
41
|
+
// // `perf-index` / `error-code-doc` say as
|
|
42
|
+
// // much in their notes. Enforced in
|
|
43
|
+
// // obligations.test.mjs.
|
|
44
|
+
// // retired: not checked; kept for provenance.
|
|
45
|
+
// cluster: string, // the recurrence class this entry
|
|
46
|
+
// // instantiates (`./recurrence-clusters.mjs`).
|
|
47
|
+
// // Supplies the lesson key to cite, and
|
|
48
|
+
// // groups entries that share a root cause.
|
|
49
|
+
// lessonKey?: string, // an entry-specific memory key, overriding
|
|
50
|
+
// // its cluster's canonical one. Rare.
|
|
51
|
+
// owner: string, // who reviews this entry when it misfires.
|
|
52
|
+
// // An entry with no owner is cruft with a CI
|
|
53
|
+
// // job attached.
|
|
54
|
+
// added: string, // ISO date
|
|
55
|
+
// reviewBy: string, // ISO date. Memories expire; without this the
|
|
56
|
+
// // map is the one place in the system under
|
|
57
|
+
// // no decay pressure, and it accumulates the
|
|
58
|
+
// // way the lesson variants it replaced did.
|
|
59
|
+
// // Guard-less entries get a shorter horizon.
|
|
35
60
|
// guard?: string, // the existing CI spec/script that enforces
|
|
36
61
|
// // this partnership, if any
|
|
37
62
|
// note?: string, // why this entry exists / its limitations
|
|
@@ -56,19 +81,6 @@
|
|
|
56
81
|
|
|
57
82
|
import { mirrorPairs } from './mirror-pairs.mjs';
|
|
58
83
|
|
|
59
|
-
// The flagship recurrence class: a partner copies a CLAIM (a mirrored
|
|
60
|
-
// module's behavior, a generated artifact's content, a documented mechanism)
|
|
61
|
-
// and goes stale when the source changes. Cited by every entry below whose
|
|
62
|
-
// obligation is "this partner copies what you just changed."
|
|
63
|
-
const COPIES_A_CLAIM_LESSON =
|
|
64
|
-
'implement-suggestion-lessons::a-mechanism-clause-you-correct-in-the-pr-body-must-be-corrected-in-every-doc-that-copies-it';
|
|
65
|
-
|
|
66
|
-
// The registered-everywhere / sibling-set recurrence class: adding or moving
|
|
67
|
-
// something that a SET of surfaces enumerates (a doc listing every command,
|
|
68
|
-
// a generated mirror listing every file) re-flags every surface that lists
|
|
69
|
-
// the set and now has a hole.
|
|
70
|
-
const SIBLING_SET_LESSON = 'aw-lessons::docs-drift-grep-must-search-names-not-invocation';
|
|
71
|
-
|
|
72
84
|
const EDGE_MIRROR_GUARD = 'packages/mcp-core/src/edge/edge-parity.spec.ts';
|
|
73
85
|
|
|
74
86
|
// One `{ match, obliges }` row PER KNOWN PAIR, in both directions — see the
|
|
@@ -76,17 +88,25 @@ const EDGE_MIRROR_GUARD = 'packages/mcp-core/src/edge/edge-parity.spec.ts';
|
|
|
76
88
|
const EDGE_MIRROR_ENTRIES = mirrorPairs.flatMap(({ core, edge }) => [
|
|
77
89
|
{
|
|
78
90
|
id: 'edge-mirror',
|
|
91
|
+
state: 'gating',
|
|
92
|
+
owner: '@mthines',
|
|
93
|
+
added: '2026-08-26',
|
|
94
|
+
reviewBy: '2027-02-26',
|
|
79
95
|
match: edge,
|
|
80
96
|
obliges: [core],
|
|
81
|
-
|
|
97
|
+
cluster: 'copies-a-claim',
|
|
82
98
|
guard: EDGE_MIRROR_GUARD,
|
|
83
99
|
note: 'An edge (Deno) module mirrored self-contained from mcp-core — edit one, mirror the other. Partner looked up from the shared mirror-pairs inventory, never reconstructed from an assumed-symmetric path.',
|
|
84
100
|
},
|
|
85
101
|
{
|
|
86
102
|
id: 'edge-mirror-core',
|
|
103
|
+
state: 'gating',
|
|
104
|
+
owner: '@mthines',
|
|
105
|
+
added: '2026-08-26',
|
|
106
|
+
reviewBy: '2027-02-26',
|
|
87
107
|
match: core,
|
|
88
108
|
obliges: [edge],
|
|
89
|
-
|
|
109
|
+
cluster: 'copies-a-claim',
|
|
90
110
|
guard: EDGE_MIRROR_GUARD,
|
|
91
111
|
note: 'The reverse direction of edge-mirror — a mcp-core source file changed, its known edge mirror (from the same mirror-pairs inventory) is the partner.',
|
|
92
112
|
},
|
|
@@ -96,6 +116,10 @@ export const SURFACE_PARTNER_MAP = [
|
|
|
96
116
|
...EDGE_MIRROR_ENTRIES,
|
|
97
117
|
{
|
|
98
118
|
id: 'tool-catalog',
|
|
119
|
+
state: 'gating',
|
|
120
|
+
owner: '@mthines',
|
|
121
|
+
added: '2026-08-26',
|
|
122
|
+
reviewBy: '2027-02-26',
|
|
99
123
|
match: 'packages/schemas/src/shared/tool-catalog.ts',
|
|
100
124
|
obliges: [
|
|
101
125
|
'supabase/functions/mcp/tool-dispatch.generated.ts',
|
|
@@ -103,52 +127,84 @@ export const SURFACE_PARTNER_MAP = [
|
|
|
103
127
|
'packages/web/public/llms.txt',
|
|
104
128
|
'run:pnpm nx generate:llms schemas',
|
|
105
129
|
],
|
|
106
|
-
|
|
130
|
+
cluster: 'copies-a-claim',
|
|
107
131
|
guard: 'packages/mcp-core/src/mcp-guards/tool-catalog-parity.spec.ts, scripts/codegen/gen-surfaces.mjs --check',
|
|
108
132
|
note: 'The catalog is the single origin of the operation surface — every generated projection of it must be regenerated in the same commit.',
|
|
109
133
|
},
|
|
110
134
|
{
|
|
111
135
|
id: 'llms-generated',
|
|
136
|
+
state: 'gating',
|
|
137
|
+
owner: '@mthines',
|
|
138
|
+
added: '2026-08-26',
|
|
139
|
+
reviewBy: '2027-02-26',
|
|
112
140
|
match: ['packages/schemas/src/llms/template.md', 'packages/schemas/src/shared/tool-catalog.ts'],
|
|
113
141
|
obliges: ['packages/web/public/llms.txt', 'run:pnpm nx generate:llms schemas'],
|
|
114
|
-
|
|
142
|
+
cluster: 'copies-a-claim',
|
|
115
143
|
guard: 'packages/schemas/src/llms/render.spec.ts',
|
|
116
144
|
note: 'llms.txt is GENERATED — never hand-edited; the committed file must be what the generator produces from these two sources.',
|
|
117
145
|
},
|
|
118
146
|
{
|
|
119
147
|
id: 'docs-section',
|
|
148
|
+
state: 'gating',
|
|
149
|
+
owner: '@mthines',
|
|
150
|
+
added: '2026-08-26',
|
|
151
|
+
reviewBy: '2027-02-26',
|
|
120
152
|
match: 're:^packages/web/src/content/docs/[^/]+\\.mdx$',
|
|
121
153
|
obliges: ['packages/web/src/lib/docs/sections.ts'],
|
|
122
|
-
|
|
154
|
+
cluster: 'sibling-set',
|
|
123
155
|
guard: 'packages/web/src/lib/docs/sections.spec.ts',
|
|
124
156
|
note: 'A new/removed docs page needs its DOCS_SECTIONS entry, or the site index and the page itself drift apart.',
|
|
125
157
|
},
|
|
126
158
|
{
|
|
127
159
|
id: 'plugin-skill',
|
|
160
|
+
state: 'gating',
|
|
161
|
+
owner: '@mthines',
|
|
162
|
+
added: '2026-08-26',
|
|
163
|
+
reviewBy: '2027-02-26',
|
|
128
164
|
match: 'packages/cli/skill/**',
|
|
129
165
|
obliges: ['run:node scripts/codegen/sync-plugin-skill.mjs', 'plugins/lorekit-claude/skills/**'],
|
|
130
|
-
|
|
166
|
+
cluster: 'sibling-set',
|
|
131
167
|
guard: 'scripts/codegen/sync-plugin-skill.mjs --check',
|
|
132
168
|
note: 'The Claude plugin vendors a copy of every skill/* source — regenerate the mirror, never hand-edit it.',
|
|
133
169
|
},
|
|
170
|
+
{
|
|
171
|
+
id: 'cli-flag-doc',
|
|
172
|
+
state: 'advisory',
|
|
173
|
+
owner: '@mthines',
|
|
174
|
+
added: '2026-08-31',
|
|
175
|
+
reviewBy: '2026-11-30',
|
|
176
|
+
match: 'packages/cli/bin/lorekit.mjs',
|
|
177
|
+
obliges: ['docs/cli.md', 'packages/cli/README.md', 'CLAUDE.md'],
|
|
178
|
+
cluster: 'copies-a-claim',
|
|
179
|
+
guard: null,
|
|
180
|
+
note: 'The flag table and per-command help in bin/lorekit.mjs is the origin of every flag claim; docs/cli.md, the package README (prose AND its flag table) and CLAUDE.md each restate it. Found by running this command against its own changed-set while changing the meaning of --strict: nothing fired, because the map covered every generated surface and no hand-written one. A path proxy, not a content predicate — bin/lorekit.mjs changes for reasons unrelated to flags — so advisory by construction, like error-code-doc.',
|
|
181
|
+
},
|
|
134
182
|
{
|
|
135
183
|
id: 'perf-index',
|
|
184
|
+
state: 'advisory',
|
|
185
|
+
owner: '@mthines',
|
|
186
|
+
added: '2026-08-26',
|
|
187
|
+
reviewBy: '2026-11-26',
|
|
136
188
|
match: 're:^supabase/migrations/.*index.*\\.sql$',
|
|
137
189
|
obliges: ['supabase/tests/migrations.test.sql'],
|
|
138
|
-
|
|
190
|
+
cluster: 'sibling-set',
|
|
139
191
|
guard: null,
|
|
140
192
|
note: 'Convention only — a real gap. A new index migration has no automated nudge to add its coverage to the migrations test.',
|
|
141
193
|
},
|
|
142
194
|
{
|
|
143
195
|
id: 'error-code-doc',
|
|
196
|
+
state: 'advisory',
|
|
197
|
+
owner: '@mthines',
|
|
198
|
+
added: '2026-08-26',
|
|
199
|
+
reviewBy: '2026-11-26',
|
|
144
200
|
match: ['supabase/functions/mcp/mcp-handler.ts', 'packages/mcp-core/src/auth/account-wide-tools.ts'],
|
|
145
201
|
obliges: [
|
|
146
202
|
'docs/mcp-tools.md',
|
|
147
203
|
'packages/schemas/src/llms/template.md',
|
|
148
204
|
'packages/web/public/llms.txt',
|
|
149
205
|
],
|
|
150
|
-
|
|
206
|
+
cluster: 'copies-a-claim',
|
|
151
207
|
guard: null,
|
|
152
|
-
note: 'A documented path-proxy, not a content predicate — obligations sees file paths, not diffs, so it cannot tell an error-const edit from an unrelated one in the same file and may over-flag.
|
|
208
|
+
note: 'A documented path-proxy, not a content predicate — obligations sees file paths, not diffs, so it cannot tell an error-const edit from an unrelated one in the same file and may over-flag. State is `advisory` for that reason: it is reported on every hit and gates nothing, not even under --strict (only --strict-all).',
|
|
153
209
|
},
|
|
154
210
|
];
|
|
@@ -45,6 +45,8 @@
|
|
|
45
45
|
// `stemOf` (below) is the simpler, general-purpose primitive — a plain
|
|
46
46
|
// basename-without-extension — exported for standalone use and unit testing.
|
|
47
47
|
|
|
48
|
+
import { clusterForEntry, lessonKeyForEntry } from './recurrence-clusters.mjs';
|
|
49
|
+
|
|
48
50
|
export const RUN_PREFIX = 'run:';
|
|
49
51
|
export const REGEX_PREFIX = 're:';
|
|
50
52
|
|
|
@@ -217,10 +219,15 @@ export function checkObligations({ changedFiles = [], map = [] } = {}) {
|
|
|
217
219
|
|
|
218
220
|
let bucket = byId.get(entry.id);
|
|
219
221
|
if (!bucket) {
|
|
222
|
+
const cluster = clusterForEntry(entry);
|
|
220
223
|
bucket = {
|
|
221
224
|
id: entry.id,
|
|
222
|
-
|
|
225
|
+
state: entry.state ?? 'advisory',
|
|
226
|
+
cluster: cluster ? { id: cluster.id, name: cluster.name, why: cluster.why } : null,
|
|
227
|
+
lessonKey: lessonKeyForEntry(entry),
|
|
223
228
|
guard: entry.guard ?? null,
|
|
229
|
+
owner: entry.owner ?? null,
|
|
230
|
+
reviewBy: entry.reviewBy ?? null,
|
|
224
231
|
note: entry.note ?? null,
|
|
225
232
|
obliges: new Map(),
|
|
226
233
|
};
|
|
@@ -232,15 +239,25 @@ export function checkObligations({ changedFiles = [], map = [] } = {}) {
|
|
|
232
239
|
}
|
|
233
240
|
}
|
|
234
241
|
|
|
235
|
-
const matched = [...byId.values()]
|
|
236
|
-
|
|
237
|
-
|
|
238
|
-
|
|
239
|
-
|
|
240
|
-
|
|
241
|
-
|
|
242
|
+
const matched = [...byId.values()]
|
|
243
|
+
.filter((b) => b.state !== 'retired')
|
|
244
|
+
.map((b) => ({
|
|
245
|
+
id: b.id,
|
|
246
|
+
state: b.state,
|
|
247
|
+
cluster: b.cluster,
|
|
248
|
+
lessonKey: b.lessonKey,
|
|
249
|
+
guard: b.guard,
|
|
250
|
+
owner: b.owner,
|
|
251
|
+
reviewBy: b.reviewBy,
|
|
252
|
+
note: b.note,
|
|
253
|
+
obliges: [...b.obliges.values()],
|
|
254
|
+
}));
|
|
255
|
+
|
|
256
|
+
const countUnmet = (entries) =>
|
|
257
|
+
entries.reduce((n, e) => n + e.obliges.filter((o) => o.met === false).length, 0);
|
|
242
258
|
|
|
243
|
-
const unmet = matched
|
|
259
|
+
const unmet = countUnmet(matched);
|
|
260
|
+
const unmetGating = countUnmet(matched.filter((e) => e.state === 'gating'));
|
|
244
261
|
|
|
245
|
-
return { files, matched, unmet, ok: unmet === 0 };
|
|
262
|
+
return { files, matched, unmet, unmetGating, ok: unmet === 0, okGating: unmetGating === 0 };
|
|
246
263
|
}
|
|
@@ -0,0 +1,161 @@
|
|
|
1
|
+
// Recurrence clusters — the named classes of repeated failure that the
|
|
2
|
+
// Surface-Partner Map's entries instantiate.
|
|
3
|
+
//
|
|
4
|
+
// This module formalizes something the map already did informally. Before it,
|
|
5
|
+
// `obligations-map.mjs` held two bare string constants —
|
|
6
|
+
// `COPIES_A_CLAIM_LESSON` and `SIBLING_SET_LESSON` — whose own comments called
|
|
7
|
+
// them "the flagship recurrence class" and "the registered-everywhere /
|
|
8
|
+
// sibling-set recurrence class". Eight entries, two classes: the clustering had
|
|
9
|
+
// been done by hand and encoded as a variable name.
|
|
10
|
+
//
|
|
11
|
+
// Naming it makes three things possible that a bare constant does not:
|
|
12
|
+
//
|
|
13
|
+
// 1. A cluster can carry WHY it recurs and WHAT ITS SHAPE IS, not just which
|
|
14
|
+
// lesson key to cite. That text is what a human hitting the check needs.
|
|
15
|
+
// 2. `checkObligations` can report the cluster alongside the entry, so two
|
|
16
|
+
// unmet obligations from the same root cause read as one problem.
|
|
17
|
+
// 3. It is the join point for the compile pipeline: a cluster is what a
|
|
18
|
+
// candidates scan over the memory store produces (N near-duplicate
|
|
19
|
+
// memories that are really one class), and an invariant entry is what a
|
|
20
|
+
// human writes from a cluster. The `sourceKeys` field records that origin
|
|
21
|
+
// in the only direction currently available — from the entry back to the
|
|
22
|
+
// memories.
|
|
23
|
+
//
|
|
24
|
+
// Pure and zero-dep by design, like `obligations-pure.mjs`: no filesystem, no
|
|
25
|
+
// cwd, no network. The registry is data.
|
|
26
|
+
//
|
|
27
|
+
// Cluster schema:
|
|
28
|
+
// {
|
|
29
|
+
// id: string, // stable slug, referenced by map entries' `cluster`
|
|
30
|
+
// name: string, // one line, the shape of the recurrence
|
|
31
|
+
// lessonKey: string, // canonical memory key an entry in this class cites
|
|
32
|
+
// why: string, // why this class recurs — shown on a hit
|
|
33
|
+
// sourceKeys?: string[] // other memory keys this class subsumes, if known
|
|
34
|
+
// }
|
|
35
|
+
|
|
36
|
+
/**
|
|
37
|
+
* The named recurrence classes. Adding one is a deliberate act: a cluster
|
|
38
|
+
* asserts that several distinct memories are really one class, which is the
|
|
39
|
+
* claim a groom-merge would act on. Two entries sharing a `lessonKey` but not
|
|
40
|
+
* a cluster is a smell — they are the same class and should say so.
|
|
41
|
+
*/
|
|
42
|
+
export const RECURRENCE_CLUSTERS = [
|
|
43
|
+
{
|
|
44
|
+
id: 'copies-a-claim',
|
|
45
|
+
name: 'A partner copies a claim and goes stale when the source changes',
|
|
46
|
+
lessonKey:
|
|
47
|
+
'implement-suggestion-lessons::a-mechanism-clause-you-correct-in-the-pr-body-must-be-corrected-in-every-doc-that-copies-it',
|
|
48
|
+
why: 'Some surface — a mirrored module, a generated artifact, a doc paragraph — restates a claim it does not own. Changing the origin does not change the copy, and nothing in the edit itself points at the copy.',
|
|
49
|
+
},
|
|
50
|
+
{
|
|
51
|
+
id: 'sibling-set',
|
|
52
|
+
name: 'A set-enumerating surface gains a hole when a member is added or moved',
|
|
53
|
+
lessonKey: 'aw-lessons::docs-drift-grep-must-search-names-not-invocation',
|
|
54
|
+
why: 'Several surfaces enumerate a set (every command, every docs page, every mirrored file). Adding or moving a member leaves each enumeration silently incomplete, and a grep for the invocation rather than the name misses them.',
|
|
55
|
+
},
|
|
56
|
+
];
|
|
57
|
+
|
|
58
|
+
const BY_ID = new Map(RECURRENCE_CLUSTERS.map((cl) => [cl.id, cl]));
|
|
59
|
+
const BY_LESSON_KEY = new Map(RECURRENCE_CLUSTERS.map((cl) => [cl.lessonKey, cl]));
|
|
60
|
+
|
|
61
|
+
/** The cluster with this id, or null. Total: an unknown id is not an error. */
|
|
62
|
+
export function clusterById(id) {
|
|
63
|
+
if (typeof id !== 'string' || !id) return null;
|
|
64
|
+
return BY_ID.get(id) ?? null;
|
|
65
|
+
}
|
|
66
|
+
|
|
67
|
+
/**
|
|
68
|
+
* The cluster that owns this lesson key, or null. Lets an entry that still
|
|
69
|
+
* carries a bare `lessonKey` (rather than a `cluster` id) resolve to its class,
|
|
70
|
+
* so the two forms coexist during the migration.
|
|
71
|
+
*/
|
|
72
|
+
export function clusterForLessonKey(lessonKey) {
|
|
73
|
+
if (typeof lessonKey !== 'string' || !lessonKey) return null;
|
|
74
|
+
return BY_LESSON_KEY.get(lessonKey) ?? null;
|
|
75
|
+
}
|
|
76
|
+
|
|
77
|
+
/**
|
|
78
|
+
* Resolve an entry's cluster from either form — an explicit `cluster` id wins,
|
|
79
|
+
* a bare `lessonKey` falls back to reverse lookup. Returns null when the entry
|
|
80
|
+
* belongs to no known class, which is legal but worth reporting: an entry with
|
|
81
|
+
* no cluster is a one-off, and a one-off is weak evidence for a check.
|
|
82
|
+
*/
|
|
83
|
+
export function clusterForEntry(entry) {
|
|
84
|
+
if (!entry || typeof entry !== 'object') return null;
|
|
85
|
+
return clusterById(entry.cluster) ?? clusterForLessonKey(entry.lessonKey);
|
|
86
|
+
}
|
|
87
|
+
|
|
88
|
+
/**
|
|
89
|
+
* The lesson key an entry should cite: its own explicit `lessonKey` if it has
|
|
90
|
+
* one, otherwise its cluster's canonical key.
|
|
91
|
+
*/
|
|
92
|
+
export function lessonKeyForEntry(entry) {
|
|
93
|
+
if (!entry || typeof entry !== 'object') return null;
|
|
94
|
+
if (typeof entry.lessonKey === 'string' && entry.lessonKey) return entry.lessonKey;
|
|
95
|
+
return clusterForEntry(entry)?.lessonKey ?? null;
|
|
96
|
+
}
|
|
97
|
+
|
|
98
|
+
/** Every map entry belonging to a cluster, by id. Order preserved. */
|
|
99
|
+
export function clusterMembers(clusterId, map = []) {
|
|
100
|
+
if (!Array.isArray(map)) return [];
|
|
101
|
+
const seen = new Set();
|
|
102
|
+
const out = [];
|
|
103
|
+
for (const entry of map) {
|
|
104
|
+
if (clusterForEntry(entry)?.id !== clusterId) continue;
|
|
105
|
+
if (!entry.id || seen.has(entry.id)) continue;
|
|
106
|
+
seen.add(entry.id);
|
|
107
|
+
out.push(entry.id);
|
|
108
|
+
}
|
|
109
|
+
return out;
|
|
110
|
+
}
|
|
111
|
+
|
|
112
|
+
/**
|
|
113
|
+
* The join to `dedupe`: does a group of memory keys (a dedupe cluster's
|
|
114
|
+
* members) already belong to a NAMED recurrence class? A member's `key`
|
|
115
|
+
* resolves to a class when it equals that class's canonical `lessonKey` or
|
|
116
|
+
* appears in its `sourceKeys` — the other sightings the class is known to
|
|
117
|
+
* subsume.
|
|
118
|
+
*
|
|
119
|
+
* This is a stronger signal than lexical similarity alone: a dedupe cluster
|
|
120
|
+
* that resolves here isn't just "these look alike", it's "this is (at least
|
|
121
|
+
* partly) another sighting of a class we already named and can cite."
|
|
122
|
+
*
|
|
123
|
+
* `pure: true` only when EVERY member resolves to the SAME single class — the
|
|
124
|
+
* strongest case, where merging under the class's canonical key loses no
|
|
125
|
+
* stragglers. A mixed match (some members resolve, some don't, or they split
|
|
126
|
+
* across classes) still reports the majority class via `matched`, so a
|
|
127
|
+
* partial hit isn't silently discarded — ties break by registry order.
|
|
128
|
+
*
|
|
129
|
+
* Total: no members, or no member resolving to any class, returns the null
|
|
130
|
+
* shape rather than throwing or returning undefined fields.
|
|
131
|
+
*/
|
|
132
|
+
export function resolveRecurrenceClass(members = [], clusters = RECURRENCE_CLUSTERS) {
|
|
133
|
+
const list = Array.isArray(members) ? members : [];
|
|
134
|
+
const counts = new Map();
|
|
135
|
+
const matched = [];
|
|
136
|
+
for (const m of list) {
|
|
137
|
+
const key = m?.key;
|
|
138
|
+
if (typeof key !== 'string' || !key) continue;
|
|
139
|
+
for (const cl of clusters) {
|
|
140
|
+
const sourceKeys = Array.isArray(cl.sourceKeys) ? cl.sourceKeys : [];
|
|
141
|
+
if (key !== cl.lessonKey && !sourceKeys.includes(key)) continue;
|
|
142
|
+
matched.push(key);
|
|
143
|
+
counts.set(cl.id, (counts.get(cl.id) ?? 0) + 1);
|
|
144
|
+
break;
|
|
145
|
+
}
|
|
146
|
+
}
|
|
147
|
+
if (matched.length === 0) return { classId: null, className: null, matched: [], pure: false };
|
|
148
|
+
|
|
149
|
+
let bestId = null;
|
|
150
|
+
let bestCount = -1;
|
|
151
|
+
for (const cl of clusters) {
|
|
152
|
+
const n = counts.get(cl.id) ?? 0;
|
|
153
|
+
if (n > bestCount) {
|
|
154
|
+
bestCount = n;
|
|
155
|
+
bestId = cl.id;
|
|
156
|
+
}
|
|
157
|
+
}
|
|
158
|
+
const best = clusters.find((cl) => cl.id === bestId) ?? null;
|
|
159
|
+
const pure = matched.length === list.length && counts.size === 1;
|
|
160
|
+
return { classId: best?.id ?? null, className: best?.name ?? null, matched, pure };
|
|
161
|
+
}
|