@monoes/monomindcli 2.14.1 → 2.15.1

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.
Files changed (197) hide show
  1. package/.claude/helpers/control-start.cjs +10 -13
  2. package/.claude/skills/monodesign/scripts/context.mjs +4 -4
  3. package/.claude/skills/monodesign/scripts/critique-storage.mjs +1 -1
  4. package/.claude/skills/monodesign/scripts/detect-csp.mjs +1 -6
  5. package/.claude/skills/monodesign/scripts/detector/browser/injected/index.mjs +16 -18
  6. package/.claude/skills/monodesign/scripts/detector/cli/main.mjs +3 -3
  7. package/.claude/skills/monodesign/scripts/detector/detect-antipatterns-browser.js +67 -47
  8. package/.claude/skills/monodesign/scripts/detector/engines/regex/detect-text.mjs +6 -6
  9. package/.claude/skills/monodesign/scripts/detector/engines/static-html/css-cascade.mjs +6 -5
  10. package/.claude/skills/monodesign/scripts/detector/engines/static-html/detect-html.mjs +6 -6
  11. package/.claude/skills/monodesign/scripts/detector/fix/index.mjs +3 -3
  12. package/.claude/skills/monodesign/scripts/detector/node/file-system.mjs +1 -1
  13. package/.claude/skills/monodesign/scripts/detector/registry/antipatterns.mjs +1 -1
  14. package/.claude/skills/monodesign/scripts/detector/rules/checks.mjs +24 -25
  15. package/.claude/skills/monodesign/scripts/detector/shared/color.mjs +2 -2
  16. package/.claude/skills/monodesign/scripts/detector/shared/constants.mjs +1 -1
  17. package/.claude/skills/monodesign/scripts/detector/shared/inline-ignores.mjs +2 -2
  18. package/.claude/skills/monodesign/scripts/hook-admin.mjs +6 -6
  19. package/.claude/skills/monodesign/scripts/hook-before-edit.mjs +4 -4
  20. package/.claude/skills/monodesign/scripts/hook-lib.mjs +20 -11
  21. package/.claude/skills/monodesign/scripts/hook.mjs +1 -1
  22. package/.claude/skills/monodesign/scripts/lib/design-parser.mjs +9 -42
  23. package/.claude/skills/monodesign/scripts/lib/is-generated.mjs +1 -1
  24. package/.claude/skills/monodesign/scripts/lib/monodesign-config.mjs +5 -5
  25. package/.claude/skills/monodesign/scripts/lib/monodesign-paths.mjs +1 -1
  26. package/.claude/skills/monodesign/scripts/live/event-validation.mjs +13 -13
  27. package/.claude/skills/monodesign/scripts/live/manual-apply.mjs +4 -4
  28. package/.claude/skills/monodesign/scripts/live/manual-edits-buffer.mjs +1 -1
  29. package/.claude/skills/monodesign/scripts/live/session-store.mjs +5 -5
  30. package/.claude/skills/monodesign/scripts/live/svelte-component.mjs +13 -13
  31. package/.claude/skills/monodesign/scripts/live/sveltekit-adapter.mjs +4 -4
  32. package/.claude/skills/monodesign/scripts/live/ui-core.mjs +1 -1
  33. package/.claude/skills/monodesign/scripts/live-accept.mjs +24 -24
  34. package/.claude/skills/monodesign/scripts/live-browser-dom.js +8 -7
  35. package/.claude/skills/monodesign/scripts/live-browser-session.js +8 -7
  36. package/.claude/skills/monodesign/scripts/live-browser.js +405 -432
  37. package/.claude/skills/monodesign/scripts/live-commit-manual-edits.mjs +6 -6
  38. package/.claude/skills/monodesign/scripts/live-copy-edit-agent.mjs +2 -2
  39. package/.claude/skills/monodesign/scripts/live-discard-manual-edits.mjs +1 -1
  40. package/.claude/skills/monodesign/scripts/live-inject.mjs +7 -7
  41. package/.claude/skills/monodesign/scripts/live-insert.mjs +11 -11
  42. package/.claude/skills/monodesign/scripts/live-manual-edit-evidence.mjs +4 -4
  43. package/.claude/skills/monodesign/scripts/live-poll.mjs +5 -5
  44. package/.claude/skills/monodesign/scripts/live-server.mjs +10 -10
  45. package/.claude/skills/monodesign/scripts/live-wrap.mjs +31 -41
  46. package/.claude/skills/monodesign/scripts/live.mjs +5 -5
  47. package/.claude/skills/monodesign/scripts/palette.mjs +3 -1
  48. package/.claude/skills/monodesign/scripts/pin.mjs +1 -1
  49. package/dist/src/commands/doc-filters.d.ts.map +1 -1
  50. package/dist/src/commands/doc-filters.js +6 -0
  51. package/dist/src/commands/doc-filters.js.map +1 -1
  52. package/dist/src/commands/doc-library.d.ts +2 -1
  53. package/dist/src/commands/doc-library.d.ts.map +1 -1
  54. package/dist/src/commands/doc-library.js +66 -2
  55. package/dist/src/commands/doc-library.js.map +1 -1
  56. package/dist/src/commands/doc.d.ts.map +1 -1
  57. package/dist/src/commands/doc.js +2 -0
  58. package/dist/src/commands/doc.js.map +1 -1
  59. package/dist/src/commands/org.d.ts.map +1 -1
  60. package/dist/src/commands/org.js +42 -4
  61. package/dist/src/commands/org.js.map +1 -1
  62. package/dist/src/commands/ui.d.ts.map +1 -1
  63. package/dist/src/commands/ui.js +60 -0
  64. package/dist/src/commands/ui.js.map +1 -1
  65. package/dist/src/knowledge/capture-envelope.d.ts +5 -0
  66. package/dist/src/knowledge/capture-envelope.d.ts.map +1 -1
  67. package/dist/src/knowledge/capture-envelope.js +2 -0
  68. package/dist/src/knowledge/capture-envelope.js.map +1 -1
  69. package/dist/src/knowledge/document-chunking.d.ts +42 -0
  70. package/dist/src/knowledge/document-chunking.d.ts.map +1 -0
  71. package/dist/src/knowledge/document-chunking.js +288 -0
  72. package/dist/src/knowledge/document-chunking.js.map +1 -0
  73. package/dist/src/knowledge/document-index.d.ts +124 -0
  74. package/dist/src/knowledge/document-index.d.ts.map +1 -0
  75. package/dist/src/knowledge/document-index.js +314 -0
  76. package/dist/src/knowledge/document-index.js.map +1 -0
  77. package/dist/src/knowledge/document-ingest.d.ts +23 -0
  78. package/dist/src/knowledge/document-ingest.d.ts.map +1 -0
  79. package/dist/src/knowledge/document-ingest.js +368 -0
  80. package/dist/src/knowledge/document-ingest.js.map +1 -0
  81. package/dist/src/knowledge/document-pipeline.d.ts +26 -208
  82. package/dist/src/knowledge/document-pipeline.d.ts.map +1 -1
  83. package/dist/src/knowledge/document-pipeline.js +25 -1104
  84. package/dist/src/knowledge/document-pipeline.js.map +1 -1
  85. package/dist/src/knowledge/document-search.d.ts +27 -0
  86. package/dist/src/knowledge/document-search.d.ts.map +1 -0
  87. package/dist/src/knowledge/document-search.js +138 -0
  88. package/dist/src/knowledge/document-search.js.map +1 -0
  89. package/dist/src/knowledge/document-store.d.ts +28 -0
  90. package/dist/src/knowledge/document-store.d.ts.map +1 -0
  91. package/dist/src/knowledge/document-store.js +61 -0
  92. package/dist/src/knowledge/document-store.js.map +1 -0
  93. package/dist/src/knowledge/document-types.d.ts +103 -0
  94. package/dist/src/knowledge/document-types.d.ts.map +1 -0
  95. package/dist/src/knowledge/document-types.js +12 -0
  96. package/dist/src/knowledge/document-types.js.map +1 -0
  97. package/dist/src/knowledge/highlights.d.ts +168 -0
  98. package/dist/src/knowledge/highlights.d.ts.map +1 -0
  99. package/dist/src/knowledge/highlights.js +311 -0
  100. package/dist/src/knowledge/highlights.js.map +1 -0
  101. package/dist/src/knowledge/lookup.d.ts +80 -0
  102. package/dist/src/knowledge/lookup.d.ts.map +1 -0
  103. package/dist/src/knowledge/lookup.js +156 -0
  104. package/dist/src/knowledge/lookup.js.map +1 -0
  105. package/dist/src/knowledge/okf-bundle.d.ts +19 -0
  106. package/dist/src/knowledge/okf-bundle.d.ts.map +1 -0
  107. package/dist/src/knowledge/okf-bundle.js +106 -0
  108. package/dist/src/knowledge/okf-bundle.js.map +1 -0
  109. package/dist/src/knowledge/profile-store.d.ts +134 -0
  110. package/dist/src/knowledge/profile-store.d.ts.map +1 -0
  111. package/dist/src/knowledge/profile-store.js +237 -0
  112. package/dist/src/knowledge/profile-store.js.map +1 -0
  113. package/dist/src/orgrt/agent-runner.d.ts +4 -0
  114. package/dist/src/orgrt/agent-runner.d.ts.map +1 -1
  115. package/dist/src/orgrt/agent-runner.js +22 -0
  116. package/dist/src/orgrt/agent-runner.js.map +1 -1
  117. package/dist/src/orgrt/antigravity-runner.d.ts.map +1 -1
  118. package/dist/src/orgrt/antigravity-runner.js +2 -1
  119. package/dist/src/orgrt/antigravity-runner.js.map +1 -1
  120. package/dist/src/orgrt/authority-mask.d.ts +32 -0
  121. package/dist/src/orgrt/authority-mask.d.ts.map +1 -0
  122. package/dist/src/orgrt/authority-mask.js +135 -0
  123. package/dist/src/orgrt/authority-mask.js.map +1 -0
  124. package/dist/src/orgrt/codex-runner.d.ts.map +1 -1
  125. package/dist/src/orgrt/codex-runner.js +2 -1
  126. package/dist/src/orgrt/codex-runner.js.map +1 -1
  127. package/dist/src/orgrt/copilot-runner.d.ts.map +1 -1
  128. package/dist/src/orgrt/copilot-runner.js +2 -1
  129. package/dist/src/orgrt/copilot-runner.js.map +1 -1
  130. package/dist/src/orgrt/crush-runner.d.ts.map +1 -1
  131. package/dist/src/orgrt/crush-runner.js +6 -1
  132. package/dist/src/orgrt/crush-runner.js.map +1 -1
  133. package/dist/src/orgrt/daemon.d.ts +4 -0
  134. package/dist/src/orgrt/daemon.d.ts.map +1 -1
  135. package/dist/src/orgrt/daemon.js +13 -2
  136. package/dist/src/orgrt/daemon.js.map +1 -1
  137. package/dist/src/orgrt/decisions.d.ts +9 -0
  138. package/dist/src/orgrt/decisions.d.ts.map +1 -1
  139. package/dist/src/orgrt/decisions.js +32 -5
  140. package/dist/src/orgrt/decisions.js.map +1 -1
  141. package/dist/src/orgrt/file-roots.d.ts +2 -1
  142. package/dist/src/orgrt/file-roots.d.ts.map +1 -1
  143. package/dist/src/orgrt/file-roots.js +7 -3
  144. package/dist/src/orgrt/file-roots.js.map +1 -1
  145. package/dist/src/orgrt/grok-runner.d.ts.map +1 -1
  146. package/dist/src/orgrt/grok-runner.js +2 -1
  147. package/dist/src/orgrt/grok-runner.js.map +1 -1
  148. package/dist/src/orgrt/hermes-runner.d.ts.map +1 -1
  149. package/dist/src/orgrt/hermes-runner.js +2 -1
  150. package/dist/src/orgrt/hermes-runner.js.map +1 -1
  151. package/dist/src/orgrt/inbox.d.ts.map +1 -1
  152. package/dist/src/orgrt/inbox.js +70 -4
  153. package/dist/src/orgrt/inbox.js.map +1 -1
  154. package/dist/src/orgrt/kimicode-runner.d.ts.map +1 -1
  155. package/dist/src/orgrt/kimicode-runner.js +2 -1
  156. package/dist/src/orgrt/kimicode-runner.js.map +1 -1
  157. package/dist/src/orgrt/opencode-runner.d.ts.map +1 -1
  158. package/dist/src/orgrt/opencode-runner.js +3 -2
  159. package/dist/src/orgrt/opencode-runner.js.map +1 -1
  160. package/dist/src/orgrt/pi-rpc-runner.d.ts.map +1 -1
  161. package/dist/src/orgrt/pi-rpc-runner.js +3 -2
  162. package/dist/src/orgrt/pi-rpc-runner.js.map +1 -1
  163. package/dist/src/orgrt/pi-runner.d.ts.map +1 -1
  164. package/dist/src/orgrt/pi-runner.js +2 -1
  165. package/dist/src/orgrt/pi-runner.js.map +1 -1
  166. package/dist/src/orgrt/policy.d.ts.map +1 -1
  167. package/dist/src/orgrt/policy.js +3 -0
  168. package/dist/src/orgrt/policy.js.map +1 -1
  169. package/dist/src/orgrt/qwen-rpc-runner.d.ts.map +1 -1
  170. package/dist/src/orgrt/qwen-rpc-runner.js +3 -2
  171. package/dist/src/orgrt/qwen-rpc-runner.js.map +1 -1
  172. package/dist/src/orgrt/qwen-runner.d.ts.map +1 -1
  173. package/dist/src/orgrt/qwen-runner.js +2 -1
  174. package/dist/src/orgrt/qwen-runner.js.map +1 -1
  175. package/dist/src/orgrt/role-sandbox.d.ts +21 -0
  176. package/dist/src/orgrt/role-sandbox.d.ts.map +1 -1
  177. package/dist/src/orgrt/role-sandbox.js +31 -6
  178. package/dist/src/orgrt/role-sandbox.js.map +1 -1
  179. package/dist/src/orgrt/session.d.ts.map +1 -1
  180. package/dist/src/orgrt/session.js +21 -2
  181. package/dist/src/orgrt/session.js.map +1 -1
  182. package/dist/src/orgrt/task-dag.d.ts +3 -0
  183. package/dist/src/orgrt/task-dag.d.ts.map +1 -1
  184. package/dist/src/orgrt/task-dag.js +1 -0
  185. package/dist/src/orgrt/task-dag.js.map +1 -1
  186. package/dist/src/orgrt/types.d.ts +5 -0
  187. package/dist/src/orgrt/types.d.ts.map +1 -1
  188. package/dist/src/orgrt/types.js +23 -1
  189. package/dist/src/orgrt/types.js.map +1 -1
  190. package/dist/src/ui/dashboard.html +28 -848
  191. package/dist/src/ui/human-auth.mjs +103 -0
  192. package/dist/src/ui/org-hil.mjs +86 -37
  193. package/dist/src/ui/org-runtime.mjs +21 -13
  194. package/dist/src/ui/routes-org.mjs +49 -445
  195. package/dist/src/ui/server.mjs +103 -4
  196. package/dist/tsconfig.tsbuildinfo +1 -1
  197. package/package.json +6 -6
@@ -0,0 +1,42 @@
1
+ /**
2
+ * Document chunking — turning a document's extracted text into the enriched
3
+ * chunks that get embedded and stored.
4
+ *
5
+ * Two passes, in this order:
6
+ * 1. CHUNK — split on heading/paragraph boundaries (shared `@monoes/memory`
7
+ * chunker, with an inline fallback kept byte-identical to it).
8
+ * 2. ENRICH — replace each chunk's leaf-heading prefix with a situating blurb
9
+ * (doc title + full heading path + summary) so the embedding model can tell
10
+ * two identically-titled sections of different documents apart.
11
+ *
12
+ * Split out of document-pipeline.ts. Pure string work: no fs, no network, no
13
+ * store access — which is what makes chunk spans reproducible at citation time
14
+ * (`chunkSpans`) without re-reading anything.
15
+ *
16
+ * @module v1/cli/knowledge/document-chunking
17
+ */
18
+ import type { ChunkSpan } from './citation.js';
19
+ export interface TextChunk {
20
+ chunkId: string;
21
+ docId: string;
22
+ text: string;
23
+ startChar: number;
24
+ endChar: number;
25
+ chunkIndex: number;
26
+ }
27
+ export declare function chunkDocument(docId: string, text: string): Promise<TextChunk[]>;
28
+ /**
29
+ * Replace each chunk's `§ heading` prefix with a richer situating blurb:
30
+ * § <doc title> · <full heading path>
31
+ * <doc summary for non-first chunks>
32
+ *
33
+ * First chunks that start with their own heading are left untouched (the
34
+ * heading IS the context). The summary line is omitted for the first
35
+ * chunk since it is adjacent to the summary text anyway.
36
+ */
37
+ export declare function enrichChunks(chunks: TextChunk[], fullText: string, filePath: string): TextChunk[];
38
+ /** Chunk spans for a text, using the SAME chunker the ingest used — this is
39
+ * what makes a chunk index resolvable back to a character range without
40
+ * storing the text twice. */
41
+ export declare function chunkSpans(text: string): Promise<ChunkSpan[]>;
42
+ //# sourceMappingURL=document-chunking.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"document-chunking.d.ts","sourceRoot":"","sources":["../../../src/knowledge/document-chunking.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAGH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,eAAe,CAAC;AAE/C,MAAM,WAAW,SAAS;IACxB,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC;IACd,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;CACpB;AA6GD,wBAAsB,aAAa,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,EAAE,CAAC,CAOrF;AAwGD;;;;;;;;GAQG;AACH,wBAAgB,YAAY,CAAC,MAAM,EAAE,SAAS,EAAE,EAAE,QAAQ,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,SAAS,EAAE,CAoCjG;AAED;;8BAE8B;AAC9B,wBAAsB,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,EAAE,CAAC,CAOnE"}
@@ -0,0 +1,288 @@
1
+ /**
2
+ * Document chunking — turning a document's extracted text into the enriched
3
+ * chunks that get embedded and stored.
4
+ *
5
+ * Two passes, in this order:
6
+ * 1. CHUNK — split on heading/paragraph boundaries (shared `@monoes/memory`
7
+ * chunker, with an inline fallback kept byte-identical to it).
8
+ * 2. ENRICH — replace each chunk's leaf-heading prefix with a situating blurb
9
+ * (doc title + full heading path + summary) so the embedding model can tell
10
+ * two identically-titled sections of different documents apart.
11
+ *
12
+ * Split out of document-pipeline.ts. Pure string work: no fs, no network, no
13
+ * store access — which is what makes chunk spans reproducible at citation time
14
+ * (`chunkSpans`) without re-reading anything.
15
+ *
16
+ * @module v1/cli/knowledge/document-chunking
17
+ */
18
+ import * as path from 'node:path';
19
+ const DEFAULT_CHUNK_SIZE = 3200;
20
+ const DEFAULT_OVERLAP = 400;
21
+ // Inline fallback identical to @monoes/memory's knowledge/document-chunker.ts —
22
+ // used only if the dynamic import below fails (package not installed/built).
23
+ // Keep in sync if the shared chunker's boundary-snapping logic changes.
24
+ const HEADING_LINE_RE = /^#{1,6} /;
25
+ const FENCE_LINE_RE = /^\s{0,3}(`{3,}|~{3,})/;
26
+ function fenceTogglesInline(text) {
27
+ const toggles = [];
28
+ let lineStart = 0;
29
+ while (lineStart <= text.length) {
30
+ const eol = text.indexOf('\n', lineStart);
31
+ const line = text.slice(lineStart, eol === -1 ? undefined : eol);
32
+ if (FENCE_LINE_RE.test(line))
33
+ toggles.push(lineStart);
34
+ if (eol === -1)
35
+ break;
36
+ lineStart = eol + 1;
37
+ }
38
+ return toggles;
39
+ }
40
+ function inFenceInline(toggles, pos) {
41
+ let lo = 0, hi = toggles.length;
42
+ while (lo < hi) {
43
+ const mid = (lo + hi) >> 1;
44
+ if (toggles[mid] <= pos)
45
+ lo = mid + 1;
46
+ else
47
+ hi = mid;
48
+ }
49
+ return (lo & 1) === 1;
50
+ }
51
+ function lastHeadingBefore(text, pos, toggles) {
52
+ let i = text.lastIndexOf('\n#', pos - 1);
53
+ while (i !== -1) {
54
+ const eol = text.indexOf('\n', i + 1);
55
+ const line = text.slice(i + 1, eol === -1 ? undefined : eol);
56
+ if (HEADING_LINE_RE.test(line) && !inFenceInline(toggles, i + 1))
57
+ return line.replace(/^#+ /, '').trim();
58
+ i = i > 0 ? text.lastIndexOf('\n#', i - 1) : -1; // fromIndex -1 clamps to 0 — would loop on a match at 0
59
+ }
60
+ const firstEol = text.indexOf('\n');
61
+ const firstLine = firstEol === -1 ? text : text.slice(0, firstEol);
62
+ return HEADING_LINE_RE.test(firstLine) &&
63
+ !inFenceInline(toggles, 0) &&
64
+ firstEol !== -1 &&
65
+ firstEol < pos
66
+ ? firstLine.replace(/^#+ /, '').trim()
67
+ : null;
68
+ }
69
+ function chunkDocumentInline(docId, text) {
70
+ if (text.includes('\r\n'))
71
+ text = text.replace(/\r\n/g, '\n');
72
+ if (text.length === 0)
73
+ return [];
74
+ const toggles = fenceTogglesInline(text);
75
+ const chunks = [];
76
+ let startChar = 0;
77
+ let chunkIndex = 0;
78
+ while (startChar < text.length) {
79
+ let endChar = Math.min(startChar + DEFAULT_CHUNK_SIZE, text.length);
80
+ let brokeAtHeading = false;
81
+ if (endChar < text.length) {
82
+ const windowStart = Math.max(startChar, endChar - Math.floor(DEFAULT_CHUNK_SIZE * 0.2));
83
+ const window = text.slice(windowStart, endChar);
84
+ let h = window.lastIndexOf('\n#');
85
+ while (h !== -1) {
86
+ const eol = window.indexOf('\n', h + 1);
87
+ const line = window.slice(h + 1, eol === -1 ? undefined : eol);
88
+ if (HEADING_LINE_RE.test(line) &&
89
+ windowStart + h > startChar &&
90
+ !inFenceInline(toggles, windowStart + h + 1))
91
+ break;
92
+ h = h > 0 ? window.lastIndexOf('\n#', h - 1) : -1;
93
+ }
94
+ if (h !== -1 && windowStart + h > startChar) {
95
+ endChar = windowStart + h + 1;
96
+ brokeAtHeading = true;
97
+ }
98
+ else {
99
+ let lastParagraph = window.lastIndexOf('\n\n');
100
+ while (lastParagraph > 0 && inFenceInline(toggles, windowStart + lastParagraph + 1)) {
101
+ lastParagraph = window.lastIndexOf('\n\n', lastParagraph - 1);
102
+ }
103
+ if (lastParagraph === 0 && inFenceInline(toggles, windowStart + 1))
104
+ lastParagraph = -1;
105
+ if (lastParagraph !== -1)
106
+ endChar = windowStart + lastParagraph + 2;
107
+ }
108
+ }
109
+ let chunkText = text.slice(startChar, endChar);
110
+ const heading = lastHeadingBefore(text, startChar + 1, toggles);
111
+ if (heading && !HEADING_LINE_RE.test(chunkText.trimStart()))
112
+ chunkText = `§ ${heading}\n${chunkText}`;
113
+ chunks.push({
114
+ chunkId: `${docId}:${chunkIndex}`,
115
+ docId,
116
+ text: chunkText,
117
+ startChar,
118
+ endChar,
119
+ chunkIndex,
120
+ });
121
+ chunkIndex++;
122
+ if (endChar >= text.length)
123
+ break;
124
+ startChar += brokeAtHeading
125
+ ? Math.max(1, endChar - startChar)
126
+ : Math.max(1, endChar - startChar - DEFAULT_OVERLAP);
127
+ }
128
+ return chunks;
129
+ }
130
+ export async function chunkDocument(docId, text) {
131
+ try {
132
+ const mod = await import('@monoes/memory');
133
+ return mod.chunkDocument(docId, text, DEFAULT_CHUNK_SIZE, DEFAULT_OVERLAP);
134
+ }
135
+ catch {
136
+ return chunkDocumentInline(docId, text);
137
+ }
138
+ }
139
+ // ── Contextual chunk enrichment (item 6a) ─────────────────────────
140
+ // Prepend a situating blurb per chunk before embedding: full heading
141
+ // path + doc title + doc summary. No LLM, no network.
142
+ //
143
+ // The `§ heading` prefix `chunkDocumentInline` writes provides only the nearest
144
+ // leaf heading. This replaces it with doc-level context so the embedding
145
+ // model can distinguish "Memory Coordination" in a hooks doc from
146
+ // "Memory Coordination" in a concepts doc.
147
+ //
148
+ // Applied at INGEST TIME (after chunking, before embedding), so:
149
+ // - Works identically regardless of which chunker ran (inline or @monoes/memory)
150
+ // - Works identically for both better-sqlite3 and sql.js (pure string ops)
151
+ // - Zero dependencies, zero network
152
+ const SECTION_PREFIX_RE = /^§ [^\n]+\n/;
153
+ /** Cap on the situating summary prepended to each chunk, ellipsis included. */
154
+ const SUMMARY_MAX_CHARS = 120;
155
+ function extractDocTitle(text, filePath) {
156
+ const eol = text.indexOf('\n');
157
+ const first = eol === -1 ? text : text.slice(0, eol);
158
+ return HEADING_LINE_RE.test(first)
159
+ ? first.replace(/^#+ /, '').trim()
160
+ : path.basename(filePath, path.extname(filePath)).replace(/[-_]/g, ' ');
161
+ }
162
+ function extractDocSummary(text) {
163
+ const lines = text.split('\n');
164
+ let inFence = false;
165
+ const parts = [];
166
+ for (const line of lines) {
167
+ if (FENCE_LINE_RE.test(line)) {
168
+ inFence = !inFence;
169
+ continue;
170
+ }
171
+ if (inFence)
172
+ continue;
173
+ if (HEADING_LINE_RE.test(line)) {
174
+ if (parts.length > 0)
175
+ break;
176
+ continue;
177
+ }
178
+ const t = line.trim();
179
+ if (!t || /^[|=-]/.test(t)) {
180
+ if (parts.length > 0)
181
+ break;
182
+ continue;
183
+ }
184
+ parts.push(t.startsWith('>') ? t.replace(/^>\s*/, '') : t);
185
+ }
186
+ const joined = parts.join(' ');
187
+ // Truncate with an ellipsis so a clipped summary is visibly clipped. The bare
188
+ // 150-char slice this replaces gave no signal that anything was cut, which
189
+ // read as a complete sentence to both a human and the embedding model.
190
+ return joined.length > SUMMARY_MAX_CHARS
191
+ ? `${joined.slice(0, SUMMARY_MAX_CHARS - 3).trimEnd()}...`
192
+ : joined;
193
+ }
194
+ function buildHeadingHierarchy(text, toggles) {
195
+ const out = [];
196
+ const eol0 = text.indexOf('\n');
197
+ const line0 = eol0 === -1 ? text : text.slice(0, eol0);
198
+ const firstLevel = line0.match(/^(#{1,6}) /)?.[1]?.length;
199
+ if (firstLevel !== undefined && !inFenceInline(toggles, 0)) {
200
+ out.push({
201
+ level: firstLevel,
202
+ text: line0.replace(/^#+ /, '').trim(),
203
+ offset: 0,
204
+ });
205
+ }
206
+ let i = text.indexOf('\n#', 0);
207
+ while (i !== -1) {
208
+ const ls = i + 1;
209
+ const e = text.indexOf('\n', ls);
210
+ const line = text.slice(ls, e === -1 ? undefined : e);
211
+ const level = line.match(/^(#{1,6}) /)?.[1]?.length;
212
+ if (level !== undefined && !inFenceInline(toggles, ls)) {
213
+ out.push({
214
+ level,
215
+ text: line.replace(/^#+ /, '').trim(),
216
+ offset: ls,
217
+ });
218
+ }
219
+ i = text.indexOf('\n#', ls);
220
+ }
221
+ return out;
222
+ }
223
+ function headingPathAt(hierarchy, pos) {
224
+ const stack = [];
225
+ for (const h of hierarchy) {
226
+ if (h.offset >= pos)
227
+ break;
228
+ while (stack.length > 0 && stack[stack.length - 1].level >= h.level)
229
+ stack.pop();
230
+ stack.push(h);
231
+ }
232
+ return stack.map((s) => s.text);
233
+ }
234
+ /**
235
+ * Replace each chunk's `§ heading` prefix with a richer situating blurb:
236
+ * § <doc title> · <full heading path>
237
+ * <doc summary for non-first chunks>
238
+ *
239
+ * First chunks that start with their own heading are left untouched (the
240
+ * heading IS the context). The summary line is omitted for the first
241
+ * chunk since it is adjacent to the summary text anyway.
242
+ */
243
+ export function enrichChunks(chunks, fullText, filePath) {
244
+ if (chunks.length === 0)
245
+ return chunks;
246
+ const toggles = fenceTogglesInline(fullText);
247
+ const hierarchy = buildHeadingHierarchy(fullText, toggles);
248
+ const title = extractDocTitle(fullText, filePath);
249
+ const summary = extractDocSummary(fullText);
250
+ return chunks.map((c) => {
251
+ let text = c.text;
252
+ // First chunk starting with its own heading — the heading IS the context
253
+ if (c.chunkIndex === 0 && HEADING_LINE_RE.test(text.trimStart()))
254
+ return c;
255
+ // Strip the old § leaf-heading prefix; we replace it with a richer one
256
+ text = text.replace(SECTION_PREFIX_RE, '');
257
+ const hpath = headingPathAt(hierarchy, c.startChar + 1);
258
+ const parts = [];
259
+ // Title + full heading path
260
+ if (hpath.length > 0 && hpath[0] !== title) {
261
+ parts.push(`§ ${title} · ${hpath.join(' > ')}`);
262
+ }
263
+ else if (hpath.length > 1) {
264
+ parts.push(`§ ${hpath.join(' > ')}`);
265
+ }
266
+ else {
267
+ parts.push(`§ ${title}`);
268
+ }
269
+ // Summary for non-first chunks — they are far from the doc intro
270
+ if (c.chunkIndex > 0 && summary) {
271
+ const snip = summary.length > 120 ? `${summary.slice(0, 117)}...` : summary;
272
+ parts.push(snip);
273
+ }
274
+ return { ...c, text: `${parts.join('\n')}\n${text}` };
275
+ });
276
+ }
277
+ /** Chunk spans for a text, using the SAME chunker the ingest used — this is
278
+ * what makes a chunk index resolvable back to a character range without
279
+ * storing the text twice. */
280
+ export async function chunkSpans(text) {
281
+ const chunks = await chunkDocument('cite', text);
282
+ return chunks.map((c) => ({
283
+ chunkIndex: c.chunkIndex,
284
+ startChar: c.startChar,
285
+ endChar: c.endChar,
286
+ }));
287
+ }
288
+ //# sourceMappingURL=document-chunking.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"document-chunking.js","sourceRoot":"","sources":["../../../src/knowledge/document-chunking.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAEH,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAYlC,MAAM,kBAAkB,GAAG,IAAI,CAAC;AAChC,MAAM,eAAe,GAAG,GAAG,CAAC;AAE5B,gFAAgF;AAChF,6EAA6E;AAC7E,wEAAwE;AACxE,MAAM,eAAe,GAAG,UAAU,CAAC;AACnC,MAAM,aAAa,GAAG,uBAAuB,CAAC;AAC9C,SAAS,kBAAkB,CAAC,IAAY;IACtC,MAAM,OAAO,GAAa,EAAE,CAAC;IAC7B,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,OAAO,SAAS,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;QAChC,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,SAAS,CAAC,CAAC;QAC1C,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QACjE,IAAI,aAAa,CAAC,IAAI,CAAC,IAAI,CAAC;YAAE,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QACtD,IAAI,GAAG,KAAK,CAAC,CAAC;YAAE,MAAM;QACtB,SAAS,GAAG,GAAG,GAAG,CAAC,CAAC;IACtB,CAAC;IACD,OAAO,OAAO,CAAC;AACjB,CAAC;AACD,SAAS,aAAa,CAAC,OAAiB,EAAE,GAAW;IACnD,IAAI,EAAE,GAAG,CAAC,EACR,EAAE,GAAG,OAAO,CAAC,MAAM,CAAC;IACtB,OAAO,EAAE,GAAG,EAAE,EAAE,CAAC;QACf,MAAM,GAAG,GAAG,CAAC,EAAE,GAAG,EAAE,CAAC,IAAI,CAAC,CAAC;QAC3B,IAAI,OAAO,CAAC,GAAG,CAAC,IAAI,GAAG;YAAE,EAAE,GAAG,GAAG,GAAG,CAAC,CAAC;;YACjC,EAAE,GAAG,GAAG,CAAC;IAChB,CAAC;IACD,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC;AACxB,CAAC;AACD,SAAS,iBAAiB,CAAC,IAAY,EAAE,GAAW,EAAE,OAAiB;IACrE,IAAI,CAAC,GAAG,IAAI,CAAC,WAAW,CAAC,KAAK,EAAE,GAAG,GAAG,CAAC,CAAC,CAAC;IACzC,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC;QAChB,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;QACtC,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,GAAG,CAAC,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QAC7D,IAAI,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,aAAa,CAAC,OAAO,EAAE,CAAC,GAAG,CAAC,CAAC;YAC9D,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QACzC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,CAAC,KAAK,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,wDAAwD;IAC3G,CAAC;IACD,MAAM,QAAQ,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IACpC,MAAM,SAAS,GAAG,QAAQ,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,QAAQ,CAAC,CAAC;IACnE,OAAO,eAAe,CAAC,IAAI,CAAC,SAAS,CAAC;QACpC,CAAC,aAAa,CAAC,OAAO,EAAE,CAAC,CAAC;QAC1B,QAAQ,KAAK,CAAC,CAAC;QACf,QAAQ,GAAG,GAAG;QACd,CAAC,CAAC,SAAS,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE;QACtC,CAAC,CAAC,IAAI,CAAC;AACX,CAAC;AACD,SAAS,mBAAmB,CAAC,KAAa,EAAE,IAAY;IACtD,IAAI,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC;QAAE,IAAI,GAAG,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;IAC9D,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,EAAE,CAAC;IACjC,MAAM,OAAO,GAAG,kBAAkB,CAAC,IAAI,CAAC,CAAC;IACzC,MAAM,MAAM,GAAgB,EAAE,CAAC;IAC/B,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,IAAI,UAAU,GAAG,CAAC,CAAC;IAEnB,OAAO,SAAS,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QAC/B,IAAI,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,GAAG,kBAAkB,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;QACpE,IAAI,cAAc,GAAG,KAAK,CAAC;QAC3B,IAAI,OAAO,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;YAC1B,MAAM,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC,kBAAkB,GAAG,GAAG,CAAC,CAAC,CAAC;YACxF,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,EAAE,OAAO,CAAC,CAAC;YAChD,IAAI,CAAC,GAAG,MAAM,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;YAClC,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC;gBAChB,MAAM,GAAG,GAAG,MAAM,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;gBACxC,MAAM,IAAI,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,GAAG,CAAC,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;gBAC/D,IACE,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC;oBAC1B,WAAW,GAAG,CAAC,GAAG,SAAS;oBAC3B,CAAC,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,CAAC,GAAG,CAAC,CAAC;oBAE5C,MAAM;gBACR,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,WAAW,CAAC,KAAK,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YACpD,CAAC;YACD,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,WAAW,GAAG,CAAC,GAAG,SAAS,EAAE,CAAC;gBAC5C,OAAO,GAAG,WAAW,GAAG,CAAC,GAAG,CAAC,CAAC;gBAC9B,cAAc,GAAG,IAAI,CAAC;YACxB,CAAC;iBAAM,CAAC;gBACN,IAAI,aAAa,GAAG,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC;gBAC/C,OAAO,aAAa,GAAG,CAAC,IAAI,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,aAAa,GAAG,CAAC,CAAC,EAAE,CAAC;oBACpF,aAAa,GAAG,MAAM,CAAC,WAAW,CAAC,MAAM,EAAE,aAAa,GAAG,CAAC,CAAC,CAAC;gBAChE,CAAC;gBACD,IAAI,aAAa,KAAK,CAAC,IAAI,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,CAAC,CAAC;oBAAE,aAAa,GAAG,CAAC,CAAC,CAAC;gBACvF,IAAI,aAAa,KAAK,CAAC,CAAC;oBAAE,OAAO,GAAG,WAAW,GAAG,aAAa,GAAG,CAAC,CAAC;YACtE,CAAC;QACH,CAAC;QACD,IAAI,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;QAC/C,MAAM,OAAO,GAAG,iBAAiB,CAAC,IAAI,EAAE,SAAS,GAAG,CAAC,EAAE,OAAO,CAAC,CAAC;QAChE,IAAI,OAAO,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,SAAS,CAAC,SAAS,EAAE,CAAC;YACzD,SAAS,GAAG,KAAK,OAAO,KAAK,SAAS,EAAE,CAAC;QAC3C,MAAM,CAAC,IAAI,CAAC;YACV,OAAO,EAAE,GAAG,KAAK,IAAI,UAAU,EAAE;YACjC,KAAK;YACL,IAAI,EAAE,SAAS;YACf,SAAS;YACT,OAAO;YACP,UAAU;SACX,CAAC,CAAC;QACH,UAAU,EAAE,CAAC;QACb,IAAI,OAAO,IAAI,IAAI,CAAC,MAAM;YAAE,MAAM;QAClC,SAAS,IAAI,cAAc;YACzB,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,GAAG,SAAS,CAAC;YAClC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,GAAG,SAAS,GAAG,eAAe,CAAC,CAAC;IACzD,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,aAAa,CAAC,KAAa,EAAE,IAAY;IAC7D,IAAI,CAAC;QACH,MAAM,GAAG,GAAG,MAAM,MAAM,CAAC,gBAA0B,CAAC,CAAC;QACrD,OAAO,GAAG,CAAC,aAAa,CAAC,KAAK,EAAE,IAAI,EAAE,kBAAkB,EAAE,eAAe,CAAC,CAAC;IAC7E,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,mBAAmB,CAAC,KAAK,EAAE,IAAI,CAAC,CAAC;IAC1C,CAAC;AACH,CAAC;AAED,qEAAqE;AACrE,qEAAqE;AACrE,sDAAsD;AACtD,EAAE;AACF,gFAAgF;AAChF,yEAAyE;AACzE,kEAAkE;AAClE,2CAA2C;AAC3C,EAAE;AACF,iEAAiE;AACjE,kFAAkF;AAClF,4EAA4E;AAC5E,qCAAqC;AAErC,MAAM,iBAAiB,GAAG,aAAa,CAAC;AACxC,+EAA+E;AAC/E,MAAM,iBAAiB,GAAG,GAAG,CAAC;AAE9B,SAAS,eAAe,CAAC,IAAY,EAAE,QAAgB;IACrD,MAAM,GAAG,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IAC/B,MAAM,KAAK,GAAG,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;IACrD,OAAO,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC;QAChC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE;QAClC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,QAAQ,EAAE,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;AAC5E,CAAC;AAED,SAAS,iBAAiB,CAAC,IAAY;IACrC,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IAC/B,IAAI,OAAO,GAAG,KAAK,CAAC;IACpB,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACzB,IAAI,aAAa,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;YAC7B,OAAO,GAAG,CAAC,OAAO,CAAC;YACnB,SAAS;QACX,CAAC;QACD,IAAI,OAAO;YAAE,SAAS;QACtB,IAAI,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;YAC/B,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;gBAAE,MAAM;YAC5B,SAAS;QACX,CAAC;QACD,MAAM,CAAC,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;QACtB,IAAI,CAAC,CAAC,IAAI,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC;YAC3B,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;gBAAE,MAAM;YAC5B,SAAS;QACX,CAAC;QACD,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAC7D,CAAC;IACD,MAAM,MAAM,GAAG,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;IAC/B,8EAA8E;IAC9E,2EAA2E;IAC3E,uEAAuE;IACvE,OAAO,MAAM,CAAC,MAAM,GAAG,iBAAiB;QACtC,CAAC,CAAC,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,iBAAiB,GAAG,CAAC,CAAC,CAAC,OAAO,EAAE,KAAK;QAC1D,CAAC,CAAC,MAAM,CAAC;AACb,CAAC;AAED,SAAS,qBAAqB,CAC5B,IAAY,EACZ,OAAiB;IAEjB,MAAM,GAAG,GAA2D,EAAE,CAAC;IACvE,MAAM,IAAI,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IAChC,MAAM,KAAK,GAAG,IAAI,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,CAAC;IACvD,MAAM,UAAU,GAAG,KAAK,CAAC,KAAK,CAAC,YAAY,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC;IAC1D,IAAI,UAAU,KAAK,SAAS,IAAI,CAAC,aAAa,CAAC,OAAO,EAAE,CAAC,CAAC,EAAE,CAAC;QAC3D,GAAG,CAAC,IAAI,CAAC;YACP,KAAK,EAAE,UAAU;YACjB,IAAI,EAAE,KAAK,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE;YACtC,MAAM,EAAE,CAAC;SACV,CAAC,CAAC;IACL,CAAC;IACD,IAAI,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC;IAC/B,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC;QAChB,MAAM,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QACjB,MAAM,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACjC,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QACtD,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC;QACpD,IAAI,KAAK,KAAK,SAAS,IAAI,CAAC,aAAa,CAAC,OAAO,EAAE,EAAE,CAAC,EAAE,CAAC;YACvD,GAAG,CAAC,IAAI,CAAC;gBACP,KAAK;gBACL,IAAI,EAAE,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE;gBACrC,MAAM,EAAE,EAAE;aACX,CAAC,CAAC;QACL,CAAC;QACD,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;IAC9B,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED,SAAS,aAAa,CACpB,SAAiE,EACjE,GAAW;IAEX,MAAM,KAAK,GAA2C,EAAE,CAAC;IACzD,KAAK,MAAM,CAAC,IAAI,SAAS,EAAE,CAAC;QAC1B,IAAI,CAAC,CAAC,MAAM,IAAI,GAAG;YAAE,MAAM;QAC3B,OAAO,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,KAAK;YAAE,KAAK,CAAC,GAAG,EAAE,CAAC;QACjF,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAChB,CAAC;IACD,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;AAClC,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,UAAU,YAAY,CAAC,MAAmB,EAAE,QAAgB,EAAE,QAAgB;IAClF,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,MAAM,CAAC;IACvC,MAAM,OAAO,GAAG,kBAAkB,CAAC,QAAQ,CAAC,CAAC;IAC7C,MAAM,SAAS,GAAG,qBAAqB,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAC;IAC3D,MAAM,KAAK,GAAG,eAAe,CAAC,QAAQ,EAAE,QAAQ,CAAC,CAAC;IAClD,MAAM,OAAO,GAAG,iBAAiB,CAAC,QAAQ,CAAC,CAAC;IAE5C,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE;QACtB,IAAI,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC;QAElB,yEAAyE;QACzE,IAAI,CAAC,CAAC,UAAU,KAAK,CAAC,IAAI,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;YAAE,OAAO,CAAC,CAAC;QAE3E,uEAAuE;QACvE,IAAI,GAAG,IAAI,CAAC,OAAO,CAAC,iBAAiB,EAAE,EAAE,CAAC,CAAC;QAE3C,MAAM,KAAK,GAAG,aAAa,CAAC,SAAS,EAAE,CAAC,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;QACxD,MAAM,KAAK,GAAa,EAAE,CAAC;QAE3B,4BAA4B;QAC5B,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,KAAK,CAAC,CAAC,CAAC,KAAK,KAAK,EAAE,CAAC;YAC3C,KAAK,CAAC,IAAI,CAAC,KAAK,KAAK,MAAM,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;QAClD,CAAC;aAAM,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC5B,KAAK,CAAC,IAAI,CAAC,KAAK,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;QACvC,CAAC;aAAM,CAAC;YACN,KAAK,CAAC,IAAI,CAAC,KAAK,KAAK,EAAE,CAAC,CAAC;QAC3B,CAAC;QAED,iEAAiE;QACjE,IAAI,CAAC,CAAC,UAAU,GAAG,CAAC,IAAI,OAAO,EAAE,CAAC;YAChC,MAAM,IAAI,GAAG,OAAO,CAAC,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,OAAO,CAAC;YAC5E,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACnB,CAAC;QAED,OAAO,EAAE,GAAG,CAAC,EAAE,IAAI,EAAE,GAAG,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,IAAI,EAAE,EAAE,CAAC;IACxD,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;8BAE8B;AAC9B,MAAM,CAAC,KAAK,UAAU,UAAU,CAAC,IAAY;IAC3C,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;IACjD,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QACxB,UAAU,EAAE,CAAC,CAAC,UAAU;QACxB,SAAS,EAAE,CAAC,CAAC,SAAS;QACtB,OAAO,EAAE,CAAC,CAAC,OAAO;KACnB,CAAC,CAAC,CAAC;AACN,CAAC"}
@@ -0,0 +1,124 @@
1
+ /**
2
+ * The document index — the append-only metadata log that records which
3
+ * documents are indexed, at which version, and whether that is still true.
4
+ *
5
+ * The log (`.monomind/knowledge/doc-metadata.jsonl`) is the pipeline's source
6
+ * of truth. Everything here reads or writes it:
7
+ *
8
+ * - WRITE: append a version record, tombstone a removed one.
9
+ * - READ: the last-wins live view, the full history, and lookups by path,
10
+ * canonical URL or `<scope>:<path>` docId.
11
+ * - JUDGE: which content hashes are current, which chunk keys are therefore
12
+ * superseded, and which indexed files no longer exist on disk.
13
+ *
14
+ * Nothing here touches the vector store. Removal tombstones metadata and lets
15
+ * the superseded filter hide the chunks — mark, don't destroy.
16
+ *
17
+ * Split out of document-pipeline.ts.
18
+ *
19
+ * @module v1/cli/knowledge/document-index
20
+ */
21
+ import type { DocumentMeta, ReconcileReport } from './document-types.js';
22
+ export declare const METADATA_FILE = "doc-metadata.jsonl";
23
+ export declare function metadataPath(rootDir: string): string;
24
+ /** Every record ever appended, in order, including superseded versions and
25
+ * removal tombstones. Reads the path directly instead of via `metadataPath`,
26
+ * which mkdir's. */
27
+ export declare function readMetadataLog(rootDir: string): DocumentMeta[];
28
+ export declare function readMetadata(rootDir: string): DocumentMeta[];
29
+ export declare function appendMetadata(rootDir: string, meta: DocumentMeta): void;
30
+ export declare function removeMetadataEntry(rootDir: string, filePath: string, scope: string): void;
31
+ /** Content hashes of the documents currently indexed under `rootDir`. */
32
+ export declare function liveContentHashes(rootDir: string): Set<string>;
33
+ /** True when a metadata log exists under `rootDir`.
34
+ *
35
+ * An empty live-hash set has two very different causes: the log is missing (we
36
+ * cannot judge what is current) or the log exists and every document has been
37
+ * removed (nothing is current). Collapsing them made `doc remove` of the LAST
38
+ * document a no-op — the tombstoned chunks came straight back in search.
39
+ *
40
+ * Reads the path directly instead of via `metadataPath`, which mkdir's. */
41
+ export declare function hasKnowledgeMetadata(rootDir: string): boolean;
42
+ /**
43
+ * True when `key` is a document chunk whose version is no longer current.
44
+ * Non-`doc:` keys are never superseded. When no metadata is available nothing
45
+ * is filtered, because "no metadata" must not read as "everything is stale".
46
+ *
47
+ * `metadataPresent` defaults to the old `live.size > 0` heuristic so existing
48
+ * two-argument callers keep their exact behaviour; pass `hasKnowledgeMetadata`
49
+ * to also filter correctly once the last document has been removed.
50
+ */
51
+ export declare function isSupersededKey(key: string, live: Set<string>, metadataPresent?: boolean): boolean;
52
+ export declare function supersededOverfetchLimit(limit: number, live: Set<string>): number;
53
+ export declare function listDocuments(rootDir?: string, scope?: string): DocumentMeta[];
54
+ /**
55
+ * Every recorded version of one document, oldest first (RCL-06).
56
+ *
57
+ * `target` is either an indexed file path or a capture's `canonicalUrl`. Read
58
+ * from the append-only log rather than the last-wins view, which is how a
59
+ * superseded version stays addressable after its record has been replaced or
60
+ * tombstoned.
61
+ *
62
+ * Best-effort by design: `readMetadata` compacts the log once it passes 1MB
63
+ * and keeps only live records, so history older than a compaction is gone.
64
+ * The `supersedes` pointer on each record is the durable part.
65
+ */
66
+ export declare function listDocumentVersions(rootDir?: string, target?: string, scope?: string): DocumentMeta[];
67
+ /**
68
+ * The indexed record for whatever a caller wrote down: a file path, a
69
+ * capture's `canonicalUrl` (fragment ignored), or the `<scope>:<path>` docId.
70
+ *
71
+ * Live records win. A superseded version is only reached through the
72
+ * append-only log, and only when nothing live matches — otherwise citing a
73
+ * re-captured page would resolve against the version it replaced.
74
+ */
75
+ export declare function findDocumentRecord(rootDir?: string, target?: string, scope?: string): DocumentMeta | undefined;
76
+ export declare function removeDocument(filePath: string, scope?: string, rootDir?: string): Promise<void>;
77
+ /**
78
+ * True for macOS AppleDouble sidecars (`._name`).
79
+ *
80
+ * Matches on the BASENAME PREFIX only. A legitimate document may contain `._`
81
+ * elsewhere in its name (`v1._2-release.md`), or live under a dot-directory
82
+ * that is deliberately indexed (`.monodesign/` critique snapshots), and
83
+ * neither may be rejected.
84
+ */
85
+ export declare function isResourceFork(filePath: string): boolean;
86
+ /**
87
+ * Reconcile the document index against the filesystem: find index entries whose
88
+ * source file no longer exists and, only when explicitly asked, tombstone them.
89
+ *
90
+ * WHY — `removeDocument` only ever tombstoned metadata, and nothing has ever
91
+ * compared the index against the disk, so a deleted file stayed searchable
92
+ * forever. Measured 2026-07-28: 109 of 257 live entries (42.4%) had no file
93
+ * behind them, including `docs/concepts/memory.md`. The Second Brain was
94
+ * answering questions from documents the user had deleted.
95
+ *
96
+ * WHY IT IS THIS CAUTIOUS — "drop the index entry when the file is missing" is
97
+ * a rule with a known catastrophic reading. A missing file is also an unmounted
98
+ * volume, a checked-out branch, a partial clone, or a permissions failure. Two
99
+ * guards were tried against real data and REJECTED; they are recorded here so
100
+ * they are not re-proposed:
101
+ *
102
+ * - "abort if >50% of entries are missing" — the real, legitimate missing
103
+ * fraction was 42.4%, so the threshold never fires in the one case we have.
104
+ * Any threshold that would have blocked this reconcile is fitted to nothing.
105
+ * - "only reconcile when the parent directory still exists" — 26 of the 109
106
+ * missing files had no parent directory, because `docs/concepts`,
107
+ * `docs/adrs` and `docs/commands` were legitimately deleted wholesale. A
108
+ * deleted directory and an unmounted volume are indistinguishable there.
109
+ *
110
+ * What does discriminate is the ROOT. An intact, readable root carrying a
111
+ * metadata log means the tree is genuinely present, so a missing file is
112
+ * genuinely gone. A missing root means nothing beneath it is knowable and
113
+ * nothing may be removed — hence throw rather than reconcile.
114
+ *
115
+ * Removal tombstones metadata; it does not delete store rows. Chunks stay on
116
+ * disk and fall out of search through the existing superseded filter, which
117
+ * keeps this consistent with the mark-don't-destroy rule and leaves the whole
118
+ * operation reversible from the archive.
119
+ */
120
+ export declare function reconcileIndex(rootDir?: string, opts?: {
121
+ scope?: string;
122
+ apply?: boolean;
123
+ }): Promise<ReconcileReport>;
124
+ //# sourceMappingURL=document-index.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"document-index.d.ts","sourceRoot":"","sources":["../../../src/knowledge/document-index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AASH,OAAO,KAAK,EAAE,YAAY,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAC;AAEzE,eAAO,MAAM,aAAa,uBAAuB,CAAC;AAIlD,wBAAgB,YAAY,CAAC,OAAO,EAAE,MAAM,GAAG,MAAM,CAIpD;AAED;;qBAEqB;AACrB,wBAAgB,eAAe,CAAC,OAAO,EAAE,MAAM,GAAG,YAAY,EAAE,CAa/D;AAED,wBAAgB,YAAY,CAAC,OAAO,EAAE,MAAM,GAAG,YAAY,EAAE,CAoC5D;AAED,wBAAgB,cAAc,CAAC,OAAO,EAAE,MAAM,EAAE,IAAI,EAAE,YAAY,GAAG,IAAI,CAExE;AAED,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI,CAa1F;AAkBD,yEAAyE;AACzE,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,MAAM,GAAG,GAAG,CAAC,MAAM,CAAC,CAI9D;AAED;;;;;;;2EAO2E;AAC3E,wBAAgB,oBAAoB,CAAC,OAAO,EAAE,MAAM,GAAG,OAAO,CAE7D;AAED;;;;;;;;GAQG;AACH,wBAAgB,eAAe,CAC7B,GAAG,EAAE,MAAM,EACX,IAAI,EAAE,GAAG,CAAC,MAAM,CAAC,EACjB,eAAe,UAAgB,GAC9B,OAAO,CAIT;AAQD,wBAAgB,wBAAwB,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,MAAM,CAGjF;AAID,wBAAgB,aAAa,CAAC,OAAO,SAAmB,EAAE,KAAK,CAAC,EAAE,MAAM,GAAG,YAAY,EAAE,CAGxF;AAED;;;;;;;;;;;GAWG;AACH,wBAAgB,oBAAoB,CAClC,OAAO,SAAmB,EAC1B,MAAM,CAAC,EAAE,MAAM,EACf,KAAK,CAAC,EAAE,MAAM,GACb,YAAY,EAAE,CAUhB;AAED;;;;;;;GAOG;AACH,wBAAgB,kBAAkB,CAChC,OAAO,SAAmB,EAC1B,MAAM,SAAK,EACX,KAAK,CAAC,EAAE,MAAM,GACb,YAAY,GAAG,SAAS,CA0B1B;AAED,wBAAsB,cAAc,CAClC,QAAQ,EAAE,MAAM,EAChB,KAAK,SAAW,EAChB,OAAO,SAAmB,GACzB,OAAO,CAAC,IAAI,CAAC,CAIf;AAID;;;;;;;GAOG;AACH,wBAAgB,cAAc,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAExD;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiCG;AACH,wBAAsB,cAAc,CAClC,OAAO,SAAmB,EAC1B,IAAI,CAAC,EAAE;IAAE,KAAK,CAAC,EAAE,MAAM,CAAC;IAAC,KAAK,CAAC,EAAE,OAAO,CAAA;CAAE,GACzC,OAAO,CAAC,eAAe,CAAC,CAwC1B"}