sensemaking 0.1.0 → 0.2.0

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/dist/esm/db.js CHANGED
@@ -3,21 +3,14 @@ import { join } from 'node:path';
3
3
  import { DatabaseSync } from 'node:sqlite';
4
4
  import { STATE_DIR } from './config.js';
5
5
  import { listFiles, parseFile } from './scan.js';
6
- // Rows -> SQLite only: schema/ALTER, the reconcile diff + transaction,
7
- // pragmas, the meta table, has() registration. Filesystem reading and
8
- // frontmatter parsing live in scan.ts; this module never touches
9
- // node:fs/gray-matter directly and never prints (throws are for callers).
6
+ // Rows -> SQLite: schema, reconcile, has(). Filesystem/frontmatter parsing lives in scan.ts.
10
7
  export const DB_FILENAME = 'cache.db';
11
- export const SCHEMA_VERSION = '1';
12
- // Quote a SQL identifier, escaping embedded double quotes so an unusual
13
- // frontmatter key (however unlikely) can't break out of the identifier.
8
+ // Cache shape version, independent of the config's own `version`.
9
+ export const SCHEMA_VERSION = '2';
14
10
  function quoteIdent(name) {
15
11
  return `"${name.split('"').join('""')}"`;
16
12
  }
17
- // The one custom SQL function in the whole tool: has(field, value).
18
- // - JSON-array field (stored as JSON text, e.g. `["a","b"]`) -> membership
19
- // - string field -> substring
20
- // - NULL -> false
13
+ // has(field, value): JSON-array field -> membership, string field -> substring, NULL -> false.
21
14
  function registerFunctions(db) {
22
15
  db.function('has', {
23
16
  deterministic: true,
@@ -26,30 +19,27 @@ function registerFunctions(db) {
26
19
  if (field === null || field === undefined) return 0;
27
20
  const needle = String(value);
28
21
  if (typeof field === 'string') {
29
- // Try JSON array first (arrays are stored as JSON text).
30
22
  if (field.startsWith('[')) {
31
23
  try {
32
24
  const parsed = JSON.parse(field);
33
25
  if (Array.isArray(parsed)) {
34
26
  return parsed.some((item)=>String(item) === needle) ? 1 : 0;
35
27
  }
36
- } catch {
37
- // fall through to substring match
38
- }
28
+ } catch {}
39
29
  }
40
30
  return field.includes(needle) ? 1 : 0;
41
31
  }
42
- // Numbers, etc: coerce to string and substring-match.
43
32
  return String(field).includes(needle) ? 1 : 0;
44
33
  });
45
34
  }
46
35
  function getColumns(db) {
47
- const rows = db.prepare('PRAGMA table_info(docs)').all();
36
+ const rows = db.prepare('PRAGMA table_info(frontmatter)').all();
48
37
  return new Set(rows.map((r)=>r.name));
49
38
  }
39
+ // Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter` can't dump file text into context.
50
40
  function ensureSchema(db) {
51
- db.exec(`CREATE TABLE IF NOT EXISTS docs ("path" TEXT PRIMARY KEY, "_mtime" REAL, "_size" INTEGER)`);
52
- db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');
41
+ db.exec(`CREATE TABLE IF NOT EXISTS frontmatter ("path" TEXT PRIMARY KEY, "_mtime" REAL, "_size" INTEGER)`);
42
+ db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, tokenize = 'porter unicode61')`);
53
43
  if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);
54
44
  }
55
45
  export function getMeta(db, key) {
@@ -64,18 +54,13 @@ export function setMeta(db, key, value) {
64
54
  db.prepare('INSERT INTO meta (key, value) VALUES (?, ?) ON CONFLICT(key) DO UPDATE SET value = excluded.value').run(key, value);
65
55
  }
66
56
  export function docCount(db) {
67
- const row = db.prepare('SELECT COUNT(*) AS n FROM docs').get();
57
+ const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get();
68
58
  return row.n;
69
59
  }
70
- // Reconcile the `docs` table against the filesystem: glob current files via
71
- // scan.ts, reparse stale (mtime/size changed) + new files in one
72
- // transaction, DELETE vanished paths, ALTER TABLE ADD COLUMN for newly
73
- // discovered frontmatter keys. Returns the number of files (re)parsed and
74
- // any warnings scan.ts collected (e.g. reserved-key collisions).
75
60
  export function reconcile(db, cfg, baseDir) {
76
61
  const files = listFiles(cfg, baseDir);
77
62
  const currentSet = new Set(files.map((f)=>f.relPath));
78
- const existingRows = db.prepare(`SELECT "path", "_mtime", "_size" FROM docs`).all();
63
+ const existingRows = db.prepare(`SELECT "path", "_mtime", "_size" FROM frontmatter`).all();
79
64
  const existing = new Map(existingRows.map((r)=>[
80
65
  r.path,
81
66
  r
@@ -104,21 +89,25 @@ export function reconcile(db, cfg, baseDir) {
104
89
  }
105
90
  parsedDocs.push(doc);
106
91
  }
107
- // seenColumns already contains the reserved columns (they're real columns
108
- // of `docs`), so spreading it alone avoids duplicate names in the INSERT.
109
92
  const allColumns = [
110
93
  ...seenColumns
111
94
  ];
112
- const insertSql = `INSERT OR REPLACE INTO docs (${allColumns.map(quoteIdent).join(', ')}) VALUES (${allColumns.map(()=>'?').join(', ')})`;
95
+ const insertSql = `INSERT OR REPLACE INTO frontmatter (${allColumns.map(quoteIdent).join(', ')}) VALUES (${allColumns.map(()=>'?').join(', ')})`;
113
96
  db.exec('BEGIN');
114
97
  try {
115
- for (const col of newColumns)db.exec(`ALTER TABLE docs ADD COLUMN ${quoteIdent(col)}`);
98
+ for (const col of newColumns)db.exec(`ALTER TABLE frontmatter ADD COLUMN ${quoteIdent(col)}`);
99
+ // FTS5 has no upsert, so delete-before-insert into `content`.
100
+ const delBody = db.prepare(`DELETE FROM content WHERE "path" = ?`);
116
101
  if (vanished.length > 0) {
117
- const del = db.prepare(`DELETE FROM docs WHERE "path" = ?`);
118
- for (const path of vanished)del.run(path);
102
+ const del = db.prepare(`DELETE FROM frontmatter WHERE "path" = ?`);
103
+ for (const path of vanished){
104
+ del.run(path);
105
+ delBody.run(path);
106
+ }
119
107
  }
120
108
  if (parsedDocs.length > 0) {
121
109
  const insert = db.prepare(insertSql);
110
+ const insertBody = db.prepare(`INSERT INTO content (title, summary, text, "path") VALUES (?, ?, ?, ?)`);
122
111
  for (const doc of parsedDocs){
123
112
  const values = allColumns.map((col)=>{
124
113
  var _doc_data_col;
@@ -128,6 +117,8 @@ export function reconcile(db, cfg, baseDir) {
128
117
  return (_doc_data_col = doc.data[col]) !== null && _doc_data_col !== void 0 ? _doc_data_col : null;
129
118
  });
130
119
  insert.run(...values);
120
+ delBody.run(doc.relPath);
121
+ insertBody.run(doc.search.title, doc.search.summary, doc.search.text, doc.relPath);
131
122
  }
132
123
  }
133
124
  db.exec('COMMIT');
@@ -140,11 +131,6 @@ export function reconcile(db, cfg, baseDir) {
140
131
  warnings
141
132
  };
142
133
  }
143
- // open(resolved config): open (or create) the on-disk SQLite cache at
144
- // `<baseDir>/.sense/cache.db`, reconcile against the filesystem, and return
145
- // a live handle. The DB is a warm start, never a source of truth -- the
146
- // `.md` files remain truth. Takes only an already-resolved config; discovery
147
- // and version-gating are config.ts's job, not this one's.
148
134
  export function open(cfg) {
149
135
  const stateDir = join(cfg.baseDir, STATE_DIR);
150
136
  mkdirSync(stateDir, {
@@ -155,6 +141,17 @@ export function open(cfg) {
155
141
  db.exec('PRAGMA journal_mode = WAL');
156
142
  db.exec('PRAGMA busy_timeout = 5000');
157
143
  registerFunctions(db);
144
+ db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');
145
+ // Schema version mismatch: reconcile only reparses changed files, so an old cache can't be patched up incrementally -- rebuild instead.
146
+ const version = getMeta(db, 'schema_version');
147
+ if (version !== null && version !== SCHEMA_VERSION) {
148
+ db.close();
149
+ rmSync(stateDir, {
150
+ recursive: true,
151
+ force: true
152
+ });
153
+ return open(cfg);
154
+ }
158
155
  ensureSchema(db);
159
156
  const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);
160
157
  return {
@@ -165,8 +162,7 @@ export function open(cfg) {
165
162
  warnings
166
163
  };
167
164
  }
168
- // Delete the `.sense/` state dir entirely and reconcile fresh -- the manual
169
- // reset for lingering columns or a doubted cache.
165
+ // Manual reset for a doubted cache.
170
166
  export function rebuild(cfg) {
171
167
  rmSync(join(cfg.baseDir, STATE_DIR), {
172
168
  recursive: true,
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db.ts"],"sourcesContent":["import { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from './config.ts';\nimport { STATE_DIR } from './config.ts';\nimport type { ParsedDoc } from './scan.ts';\nimport { listFiles, parseFile } from './scan.ts';\n\n// Rows -> SQLite only: schema/ALTER, the reconcile diff + transaction,\n// pragmas, the meta table, has() registration. Filesystem reading and\n// frontmatter parsing live in scan.ts; this module never touches\n// node:fs/gray-matter directly and never prints (throws are for callers).\n\nexport const DB_FILENAME = 'cache.db';\nexport const SCHEMA_VERSION = '1';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n // Number of files (re)parsed by this open's reconcile -- 0 on a fully\n // warm cache. Tests use this as reconcile instrumentation.\n parsed: number;\n warnings: string[];\n}\n\n// Quote a SQL identifier, escaping embedded double quotes so an unusual\n// frontmatter key (however unlikely) can't break out of the identifier.\nfunction quoteIdent(name: string): string {\n return `\"${name.split('\"').join('\"\"')}\"`;\n}\n\n// The one custom SQL function in the whole tool: has(field, value).\n// - JSON-array field (stored as JSON text, e.g. `[\"a\",\"b\"]`) -> membership\n// - string field -> substring\n// - NULL -> false\nfunction registerFunctions(db: DatabaseSync): void {\n db.function('has', { deterministic: true, varargs: false }, (field: unknown, value: unknown): number => {\n if (field === null || field === undefined) return 0;\n\n const needle = String(value);\n\n if (typeof field === 'string') {\n // Try JSON array first (arrays are stored as JSON text).\n if (field.startsWith('[')) {\n try {\n const parsed = JSON.parse(field);\n if (Array.isArray(parsed)) {\n return parsed.some((item) => String(item) === needle) ? 1 : 0;\n }\n } catch {\n // fall through to substring match\n }\n }\n return field.includes(needle) ? 1 : 0;\n }\n\n // Numbers, etc: coerce to string and substring-match.\n return String(field).includes(needle) ? 1 : 0;\n });\n}\n\nfunction getColumns(db: DatabaseSync): Set<string> {\n const rows = db.prepare('PRAGMA table_info(docs)').all() as Array<{ name: string }>;\n return new Set(rows.map((r) => r.name));\n}\n\nfunction ensureSchema(db: DatabaseSync): void {\n db.exec(`CREATE TABLE IF NOT EXISTS docs (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_size\" INTEGER)`);\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n}\n\nexport function getMeta(db: DatabaseSync, key: string): string | null {\n const row = db.prepare('SELECT value FROM meta WHERE key = ?').get(key) as { value: string } | undefined;\n return row ? row.value : null;\n}\n\nexport function setMeta(db: DatabaseSync, key: string, value: string | null): void {\n if (value === null) {\n db.prepare('DELETE FROM meta WHERE key = ?').run(key);\n return;\n }\n db.prepare('INSERT INTO meta (key, value) VALUES (?, ?) ON CONFLICT(key) DO UPDATE SET value = excluded.value').run(key, value);\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM docs').get() as { n: number };\n return row.n;\n}\n\n// Reconcile the `docs` table against the filesystem: glob current files via\n// scan.ts, reparse stale (mtime/size changed) + new files in one\n// transaction, DELETE vanished paths, ALTER TABLE ADD COLUMN for newly\n// discovered frontmatter keys. Returns the number of files (re)parsed and\n// any warnings scan.ts collected (e.g. reserved-key collisions).\nexport function reconcile(db: DatabaseSync, cfg: Config, baseDir: string): { parsed: number; warnings: string[] } {\n const files = listFiles(cfg, baseDir);\n const currentSet = new Set(files.map((f) => f.relPath));\n\n const existingRows = db.prepare(`SELECT \"path\", \"_mtime\", \"_size\" FROM docs`).all() as Array<{\n path: string;\n _mtime: number;\n _size: number;\n }>;\n const existing = new Map(existingRows.map((r) => [r.path, r]));\n const vanished = existingRows.filter((r) => !currentSet.has(r.path)).map((r) => r.path);\n\n const toReparse = files.filter((f) => {\n const row = existing.get(f.relPath);\n return !row || row._mtime !== f.mtimeMs || row._size !== f.size;\n });\n\n if (vanished.length === 0 && toReparse.length === 0) return { parsed: 0, warnings: [] };\n\n const seenColumns = getColumns(db);\n const newColumns: string[] = [];\n const parsedDocs: ParsedDoc[] = [];\n const warnings: string[] = [];\n\n for (const file of toReparse) {\n const { doc, warnings: fileWarnings } = parseFile(file);\n warnings.push(...fileWarnings);\n for (const key of Object.keys(doc.data)) {\n if (!seenColumns.has(key)) {\n seenColumns.add(key);\n newColumns.push(key);\n }\n }\n parsedDocs.push(doc);\n }\n\n // seenColumns already contains the reserved columns (they're real columns\n // of `docs`), so spreading it alone avoids duplicate names in the INSERT.\n const allColumns = [...seenColumns];\n const insertSql = `INSERT OR REPLACE INTO docs (${allColumns.map(quoteIdent).join(', ')}) VALUES (${allColumns.map(() => '?').join(', ')})`;\n\n db.exec('BEGIN');\n try {\n for (const col of newColumns) db.exec(`ALTER TABLE docs ADD COLUMN ${quoteIdent(col)}`);\n if (vanished.length > 0) {\n const del = db.prepare(`DELETE FROM docs WHERE \"path\" = ?`);\n for (const path of vanished) del.run(path);\n }\n if (parsedDocs.length > 0) {\n const insert = db.prepare(insertSql);\n for (const doc of parsedDocs) {\n const values = allColumns.map((col) => {\n if (col === 'path') return doc.relPath;\n if (col === '_mtime') return doc.mtimeMs;\n if (col === '_size') return doc.size;\n return doc.data[col] ?? null;\n });\n insert.run(...values);\n }\n }\n db.exec('COMMIT');\n } catch (err) {\n db.exec('ROLLBACK');\n throw err;\n }\n\n return { parsed: parsedDocs.length, warnings };\n}\n\n// open(resolved config): open (or create) the on-disk SQLite cache at\n// `<baseDir>/.sense/cache.db`, reconcile against the filesystem, and return\n// a live handle. The DB is a warm start, never a source of truth -- the\n// `.md` files remain truth. Takes only an already-resolved config; discovery\n// and version-gating are config.ts's job, not this one's.\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n db.exec('PRAGMA busy_timeout = 5000');\n registerFunctions(db);\n\n ensureSchema(db);\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings };\n}\n\n// Delete the `.sense/` state dir entirely and reconcile fresh -- the manual\n// reset for lingering columns or a doubted cache.\nexport function rebuild(cfg: ResolvedConfig): OpenResult {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n return open(cfg);\n}\n"],"names":["mkdirSync","rmSync","join","DatabaseSync","STATE_DIR","listFiles","parseFile","DB_FILENAME","SCHEMA_VERSION","quoteIdent","name","split","registerFunctions","db","function","deterministic","varargs","field","value","undefined","needle","String","startsWith","parsed","JSON","parse","Array","isArray","some","item","includes","getColumns","rows","prepare","all","Set","map","r","ensureSchema","exec","getMeta","setMeta","key","row","get","run","docCount","n","reconcile","cfg","baseDir","files","currentSet","f","relPath","existingRows","existing","Map","path","vanished","filter","has","toReparse","_mtime","mtimeMs","_size","size","length","warnings","seenColumns","newColumns","parsedDocs","file","doc","fileWarnings","push","Object","keys","data","add","allColumns","insertSql","col","del","insert","values","err","open","stateDir","recursive","dbPath","rebuild","force"],"mappings":"AAAA,SAASA,SAAS,EAAEC,MAAM,QAAQ,UAAU;AAC5C,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,YAAY,QAAQ,cAAc;AAE3C,SAASC,SAAS,QAAQ,cAAc;AAExC,SAASC,SAAS,EAAEC,SAAS,QAAQ,YAAY;AAEjD,uEAAuE;AACvE,sEAAsE;AACtE,iEAAiE;AACjE,0EAA0E;AAE1E,OAAO,MAAMC,cAAc,WAAW;AACtC,OAAO,MAAMC,iBAAiB,IAAI;AAYlC,wEAAwE;AACxE,wEAAwE;AACxE,SAASC,WAAWC,IAAY;IAC9B,OAAO,CAAC,CAAC,EAAEA,KAAKC,KAAK,CAAC,KAAKT,IAAI,CAAC,MAAM,CAAC,CAAC;AAC1C;AAEA,oEAAoE;AACpE,6EAA6E;AAC7E,gCAAgC;AAChC,oBAAoB;AACpB,SAASU,kBAAkBC,EAAgB;IACzCA,GAAGC,QAAQ,CAAC,OAAO;QAAEC,eAAe;QAAMC,SAAS;IAAM,GAAG,CAACC,OAAgBC;QAC3E,IAAID,UAAU,QAAQA,UAAUE,WAAW,OAAO;QAElD,MAAMC,SAASC,OAAOH;QAEtB,IAAI,OAAOD,UAAU,UAAU;YAC7B,yDAAyD;YACzD,IAAIA,MAAMK,UAAU,CAAC,MAAM;gBACzB,IAAI;oBACF,MAAMC,SAASC,KAAKC,KAAK,CAACR;oBAC1B,IAAIS,MAAMC,OAAO,CAACJ,SAAS;wBACzB,OAAOA,OAAOK,IAAI,CAAC,CAACC,OAASR,OAAOQ,UAAUT,UAAU,IAAI;oBAC9D;gBACF,EAAE,OAAM;gBACN,kCAAkC;gBACpC;YACF;YACA,OAAOH,MAAMa,QAAQ,CAACV,UAAU,IAAI;QACtC;QAEA,sDAAsD;QACtD,OAAOC,OAAOJ,OAAOa,QAAQ,CAACV,UAAU,IAAI;IAC9C;AACF;AAEA,SAASW,WAAWlB,EAAgB;IAClC,MAAMmB,OAAOnB,GAAGoB,OAAO,CAAC,2BAA2BC,GAAG;IACtD,OAAO,IAAIC,IAAIH,KAAKI,GAAG,CAAC,CAACC,IAAMA,EAAE3B,IAAI;AACvC;AAEA,SAAS4B,aAAazB,EAAgB;IACpCA,GAAG0B,IAAI,CAAC,CAAC,yFAAyF,CAAC;IACnG1B,GAAG0B,IAAI,CAAC;IACR,IAAIC,QAAQ3B,IAAI,sBAAsB,MAAM4B,QAAQ5B,IAAI,kBAAkBL;AAC5E;AAEA,OAAO,SAASgC,QAAQ3B,EAAgB,EAAE6B,GAAW;IACnD,MAAMC,MAAM9B,GAAGoB,OAAO,CAAC,wCAAwCW,GAAG,CAACF;IACnE,OAAOC,MAAMA,IAAIzB,KAAK,GAAG;AAC3B;AAEA,OAAO,SAASuB,QAAQ5B,EAAgB,EAAE6B,GAAW,EAAExB,KAAoB;IACzE,IAAIA,UAAU,MAAM;QAClBL,GAAGoB,OAAO,CAAC,kCAAkCY,GAAG,CAACH;QACjD;IACF;IACA7B,GAAGoB,OAAO,CAAC,qGAAqGY,GAAG,CAACH,KAAKxB;AAC3H;AAEA,OAAO,SAAS4B,SAASjC,EAAgB;IACvC,MAAM8B,MAAM9B,GAAGoB,OAAO,CAAC,kCAAkCW,GAAG;IAC5D,OAAOD,IAAII,CAAC;AACd;AAEA,4EAA4E;AAC5E,iEAAiE;AACjE,uEAAuE;AACvE,0EAA0E;AAC1E,iEAAiE;AACjE,OAAO,SAASC,UAAUnC,EAAgB,EAAEoC,GAAW,EAAEC,OAAe;IACtE,MAAMC,QAAQ9C,UAAU4C,KAAKC;IAC7B,MAAME,aAAa,IAAIjB,IAAIgB,MAAMf,GAAG,CAAC,CAACiB,IAAMA,EAAEC,OAAO;IAErD,MAAMC,eAAe1C,GAAGoB,OAAO,CAAC,CAAC,0CAA0C,CAAC,EAAEC,GAAG;IAKjF,MAAMsB,WAAW,IAAIC,IAAIF,aAAanB,GAAG,CAAC,CAACC,IAAM;YAACA,EAAEqB,IAAI;YAAErB;SAAE;IAC5D,MAAMsB,WAAWJ,aAAaK,MAAM,CAAC,CAACvB,IAAM,CAACe,WAAWS,GAAG,CAACxB,EAAEqB,IAAI,GAAGtB,GAAG,CAAC,CAACC,IAAMA,EAAEqB,IAAI;IAEtF,MAAMI,YAAYX,MAAMS,MAAM,CAAC,CAACP;QAC9B,MAAMV,MAAMa,SAASZ,GAAG,CAACS,EAAEC,OAAO;QAClC,OAAO,CAACX,OAAOA,IAAIoB,MAAM,KAAKV,EAAEW,OAAO,IAAIrB,IAAIsB,KAAK,KAAKZ,EAAEa,IAAI;IACjE;IAEA,IAAIP,SAASQ,MAAM,KAAK,KAAKL,UAAUK,MAAM,KAAK,GAAG,OAAO;QAAE5C,QAAQ;QAAG6C,UAAU,EAAE;IAAC;IAEtF,MAAMC,cAActC,WAAWlB;IAC/B,MAAMyD,aAAuB,EAAE;IAC/B,MAAMC,aAA0B,EAAE;IAClC,MAAMH,WAAqB,EAAE;IAE7B,KAAK,MAAMI,QAAQV,UAAW;QAC5B,MAAM,EAAEW,GAAG,EAAEL,UAAUM,YAAY,EAAE,GAAGpE,UAAUkE;QAClDJ,SAASO,IAAI,IAAID;QACjB,KAAK,MAAMhC,OAAOkC,OAAOC,IAAI,CAACJ,IAAIK,IAAI,EAAG;YACvC,IAAI,CAACT,YAAYR,GAAG,CAACnB,MAAM;gBACzB2B,YAAYU,GAAG,CAACrC;gBAChB4B,WAAWK,IAAI,CAACjC;YAClB;QACF;QACA6B,WAAWI,IAAI,CAACF;IAClB;IAEA,0EAA0E;IAC1E,0EAA0E;IAC1E,MAAMO,aAAa;WAAIX;KAAY;IACnC,MAAMY,YAAY,CAAC,6BAA6B,EAAED,WAAW5C,GAAG,CAAC3B,YAAYP,IAAI,CAAC,MAAM,UAAU,EAAE8E,WAAW5C,GAAG,CAAC,IAAM,KAAKlC,IAAI,CAAC,MAAM,CAAC,CAAC;IAE3IW,GAAG0B,IAAI,CAAC;IACR,IAAI;QACF,KAAK,MAAM2C,OAAOZ,WAAYzD,GAAG0B,IAAI,CAAC,CAAC,4BAA4B,EAAE9B,WAAWyE,MAAM;QACtF,IAAIvB,SAASQ,MAAM,GAAG,GAAG;YACvB,MAAMgB,MAAMtE,GAAGoB,OAAO,CAAC,CAAC,iCAAiC,CAAC;YAC1D,KAAK,MAAMyB,QAAQC,SAAUwB,IAAItC,GAAG,CAACa;QACvC;QACA,IAAIa,WAAWJ,MAAM,GAAG,GAAG;YACzB,MAAMiB,SAASvE,GAAGoB,OAAO,CAACgD;YAC1B,KAAK,MAAMR,OAAOF,WAAY;gBAC5B,MAAMc,SAASL,WAAW5C,GAAG,CAAC,CAAC8C;wBAItBT;oBAHP,IAAIS,QAAQ,QAAQ,OAAOT,IAAInB,OAAO;oBACtC,IAAI4B,QAAQ,UAAU,OAAOT,IAAIT,OAAO;oBACxC,IAAIkB,QAAQ,SAAS,OAAOT,IAAIP,IAAI;oBACpC,QAAOO,gBAAAA,IAAIK,IAAI,CAACI,IAAI,cAAbT,2BAAAA,gBAAiB;gBAC1B;gBACAW,OAAOvC,GAAG,IAAIwC;YAChB;QACF;QACAxE,GAAG0B,IAAI,CAAC;IACV,EAAE,OAAO+C,KAAK;QACZzE,GAAG0B,IAAI,CAAC;QACR,MAAM+C;IACR;IAEA,OAAO;QAAE/D,QAAQgD,WAAWJ,MAAM;QAAEC;IAAS;AAC/C;AAEA,sEAAsE;AACtE,4EAA4E;AAC5E,wEAAwE;AACxE,6EAA6E;AAC7E,0DAA0D;AAC1D,OAAO,SAASmB,KAAKtC,GAAmB;IACtC,MAAMuC,WAAWtF,KAAK+C,IAAIC,OAAO,EAAE9C;IACnCJ,UAAUwF,UAAU;QAAEC,WAAW;IAAK;IACtC,MAAMC,SAASxF,KAAKsF,UAAUjF;IAE9B,MAAMM,KAAK,IAAIV,aAAauF;IAC5B7E,GAAG0B,IAAI,CAAC;IACR1B,GAAG0B,IAAI,CAAC;IACR3B,kBAAkBC;IAElByB,aAAazB;IACb,MAAM,EAAEU,MAAM,EAAE6C,QAAQ,EAAE,GAAGpB,UAAUnC,IAAIoC,KAAKA,IAAIC,OAAO;IAE3D,OAAO;QAAErC;QAAIoC;QAAKyC;QAAQnE;QAAQ6C;IAAS;AAC7C;AAEA,4EAA4E;AAC5E,kDAAkD;AAClD,OAAO,SAASuB,QAAQ1C,GAAmB;IACzChD,OAAOC,KAAK+C,IAAIC,OAAO,EAAE9C,YAAY;QAAEqF,WAAW;QAAMG,OAAO;IAAK;IACpE,OAAOL,KAAKtC;AACd"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/db.ts"],"sourcesContent":["import { mkdirSync, rmSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from './config.ts';\nimport { STATE_DIR } from './config.ts';\nimport type { ParsedDoc } from './scan.ts';\nimport { listFiles, parseFile } from './scan.ts';\n\n// Rows -> SQLite: schema, reconcile, has(). Filesystem/frontmatter parsing lives in scan.ts.\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`.\nexport const SCHEMA_VERSION = '2';\n\nexport interface OpenResult {\n db: DatabaseSync;\n cfg: ResolvedConfig;\n dbPath: string;\n parsed: number;\n warnings: string[];\n}\n\nfunction quoteIdent(name: string): string {\n return `\"${name.split('\"').join('\"\"')}\"`;\n}\n\n// has(field, value): JSON-array field -> membership, string field -> substring, NULL -> false.\nfunction registerFunctions(db: DatabaseSync): void {\n db.function('has', { deterministic: true, varargs: false }, (field: unknown, value: unknown): number => {\n if (field === null || field === undefined) return 0;\n\n const needle = String(value);\n\n if (typeof field === 'string') {\n if (field.startsWith('[')) {\n try {\n const parsed = JSON.parse(field);\n if (Array.isArray(parsed)) {\n return parsed.some((item) => String(item) === needle) ? 1 : 0;\n }\n } catch {}\n }\n return field.includes(needle) ? 1 : 0;\n }\n\n return String(field).includes(needle) ? 1 : 0;\n });\n}\n\nfunction getColumns(db: DatabaseSync): Set<string> {\n const rows = db.prepare('PRAGMA table_info(frontmatter)').all() as Array<{ name: string }>;\n return new Set(rows.map((r) => r.name));\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter` can't dump file text into context.\nfunction ensureSchema(db: DatabaseSync): void {\n db.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_size\" INTEGER)`);\n db.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, tokenize = 'porter unicode61')`);\n if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);\n}\n\nexport function getMeta(db: DatabaseSync, key: string): string | null {\n const row = db.prepare('SELECT value FROM meta WHERE key = ?').get(key) as { value: string } | undefined;\n return row ? row.value : null;\n}\n\nexport function setMeta(db: DatabaseSync, key: string, value: string | null): void {\n if (value === null) {\n db.prepare('DELETE FROM meta WHERE key = ?').run(key);\n return;\n }\n db.prepare('INSERT INTO meta (key, value) VALUES (?, ?) ON CONFLICT(key) DO UPDATE SET value = excluded.value').run(key, value);\n}\n\nexport function docCount(db: DatabaseSync): number {\n const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get() as { n: number };\n return row.n;\n}\n\nexport function reconcile(db: DatabaseSync, cfg: Config, baseDir: string): { parsed: number; warnings: string[] } {\n const files = listFiles(cfg, baseDir);\n const currentSet = new Set(files.map((f) => f.relPath));\n\n const existingRows = db.prepare(`SELECT \"path\", \"_mtime\", \"_size\" FROM frontmatter`).all() as Array<{\n path: string;\n _mtime: number;\n _size: number;\n }>;\n const existing = new Map(existingRows.map((r) => [r.path, r]));\n const vanished = existingRows.filter((r) => !currentSet.has(r.path)).map((r) => r.path);\n\n const toReparse = files.filter((f) => {\n const row = existing.get(f.relPath);\n return !row || row._mtime !== f.mtimeMs || row._size !== f.size;\n });\n\n if (vanished.length === 0 && toReparse.length === 0) return { parsed: 0, warnings: [] };\n\n const seenColumns = getColumns(db);\n const newColumns: string[] = [];\n const parsedDocs: ParsedDoc[] = [];\n const warnings: string[] = [];\n\n for (const file of toReparse) {\n const { doc, warnings: fileWarnings } = parseFile(file);\n warnings.push(...fileWarnings);\n for (const key of Object.keys(doc.data)) {\n if (!seenColumns.has(key)) {\n seenColumns.add(key);\n newColumns.push(key);\n }\n }\n parsedDocs.push(doc);\n }\n\n const allColumns = [...seenColumns];\n const insertSql = `INSERT OR REPLACE INTO frontmatter (${allColumns.map(quoteIdent).join(', ')}) VALUES (${allColumns.map(() => '?').join(', ')})`;\n\n db.exec('BEGIN');\n try {\n for (const col of newColumns) db.exec(`ALTER TABLE frontmatter ADD COLUMN ${quoteIdent(col)}`);\n // FTS5 has no upsert, so delete-before-insert into `content`.\n const delBody = db.prepare(`DELETE FROM content WHERE \"path\" = ?`);\n if (vanished.length > 0) {\n const del = db.prepare(`DELETE FROM frontmatter WHERE \"path\" = ?`);\n for (const path of vanished) {\n del.run(path);\n delBody.run(path);\n }\n }\n if (parsedDocs.length > 0) {\n const insert = db.prepare(insertSql);\n const insertBody = db.prepare(`INSERT INTO content (title, summary, text, \"path\") VALUES (?, ?, ?, ?)`);\n for (const doc of parsedDocs) {\n const values = allColumns.map((col) => {\n if (col === 'path') return doc.relPath;\n if (col === '_mtime') return doc.mtimeMs;\n if (col === '_size') return doc.size;\n return doc.data[col] ?? null;\n });\n insert.run(...values);\n delBody.run(doc.relPath);\n insertBody.run(doc.search.title, doc.search.summary, doc.search.text, doc.relPath);\n }\n }\n db.exec('COMMIT');\n } catch (err) {\n db.exec('ROLLBACK');\n throw err;\n }\n\n return { parsed: parsedDocs.length, warnings };\n}\n\nexport function open(cfg: ResolvedConfig): OpenResult {\n const stateDir = join(cfg.baseDir, STATE_DIR);\n mkdirSync(stateDir, { recursive: true });\n const dbPath = join(stateDir, DB_FILENAME);\n\n const db = new DatabaseSync(dbPath);\n db.exec('PRAGMA journal_mode = WAL');\n db.exec('PRAGMA busy_timeout = 5000');\n registerFunctions(db);\n\n db.exec('CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT)');\n\n // Schema version mismatch: reconcile only reparses changed files, so an old cache can't be patched up incrementally -- rebuild instead.\n const version = getMeta(db, 'schema_version');\n if (version !== null && version !== SCHEMA_VERSION) {\n db.close();\n rmSync(stateDir, { recursive: true, force: true });\n return open(cfg);\n }\n\n ensureSchema(db);\n const { parsed, warnings } = reconcile(db, cfg, cfg.baseDir);\n\n return { db, cfg, dbPath, parsed, warnings };\n}\n\n// Manual reset for a doubted cache.\nexport function rebuild(cfg: ResolvedConfig): OpenResult {\n rmSync(join(cfg.baseDir, STATE_DIR), { recursive: true, force: true });\n return open(cfg);\n}\n"],"names":["mkdirSync","rmSync","join","DatabaseSync","STATE_DIR","listFiles","parseFile","DB_FILENAME","SCHEMA_VERSION","quoteIdent","name","split","registerFunctions","db","function","deterministic","varargs","field","value","undefined","needle","String","startsWith","parsed","JSON","parse","Array","isArray","some","item","includes","getColumns","rows","prepare","all","Set","map","r","ensureSchema","exec","getMeta","setMeta","key","row","get","run","docCount","n","reconcile","cfg","baseDir","files","currentSet","f","relPath","existingRows","existing","Map","path","vanished","filter","has","toReparse","_mtime","mtimeMs","_size","size","length","warnings","seenColumns","newColumns","parsedDocs","file","doc","fileWarnings","push","Object","keys","data","add","allColumns","insertSql","col","delBody","del","insert","insertBody","values","search","title","summary","text","err","open","stateDir","recursive","dbPath","version","close","force","rebuild"],"mappings":"AAAA,SAASA,SAAS,EAAEC,MAAM,QAAQ,UAAU;AAC5C,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,YAAY,QAAQ,cAAc;AAE3C,SAASC,SAAS,QAAQ,cAAc;AAExC,SAASC,SAAS,EAAEC,SAAS,QAAQ,YAAY;AAEjD,6FAA6F;AAE7F,OAAO,MAAMC,cAAc,WAAW;AACtC,kEAAkE;AAClE,OAAO,MAAMC,iBAAiB,IAAI;AAUlC,SAASC,WAAWC,IAAY;IAC9B,OAAO,CAAC,CAAC,EAAEA,KAAKC,KAAK,CAAC,KAAKT,IAAI,CAAC,MAAM,CAAC,CAAC;AAC1C;AAEA,+FAA+F;AAC/F,SAASU,kBAAkBC,EAAgB;IACzCA,GAAGC,QAAQ,CAAC,OAAO;QAAEC,eAAe;QAAMC,SAAS;IAAM,GAAG,CAACC,OAAgBC;QAC3E,IAAID,UAAU,QAAQA,UAAUE,WAAW,OAAO;QAElD,MAAMC,SAASC,OAAOH;QAEtB,IAAI,OAAOD,UAAU,UAAU;YAC7B,IAAIA,MAAMK,UAAU,CAAC,MAAM;gBACzB,IAAI;oBACF,MAAMC,SAASC,KAAKC,KAAK,CAACR;oBAC1B,IAAIS,MAAMC,OAAO,CAACJ,SAAS;wBACzB,OAAOA,OAAOK,IAAI,CAAC,CAACC,OAASR,OAAOQ,UAAUT,UAAU,IAAI;oBAC9D;gBACF,EAAE,OAAM,CAAC;YACX;YACA,OAAOH,MAAMa,QAAQ,CAACV,UAAU,IAAI;QACtC;QAEA,OAAOC,OAAOJ,OAAOa,QAAQ,CAACV,UAAU,IAAI;IAC9C;AACF;AAEA,SAASW,WAAWlB,EAAgB;IAClC,MAAMmB,OAAOnB,GAAGoB,OAAO,CAAC,kCAAkCC,GAAG;IAC7D,OAAO,IAAIC,IAAIH,KAAKI,GAAG,CAAC,CAACC,IAAMA,EAAE3B,IAAI;AACvC;AAEA,8HAA8H;AAC9H,SAAS4B,aAAazB,EAAgB;IACpCA,GAAG0B,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAC1G1B,GAAG0B,IAAI,CAAC,CAAC,0HAA0H,CAAC;IACpI,IAAIC,QAAQ3B,IAAI,sBAAsB,MAAM4B,QAAQ5B,IAAI,kBAAkBL;AAC5E;AAEA,OAAO,SAASgC,QAAQ3B,EAAgB,EAAE6B,GAAW;IACnD,MAAMC,MAAM9B,GAAGoB,OAAO,CAAC,wCAAwCW,GAAG,CAACF;IACnE,OAAOC,MAAMA,IAAIzB,KAAK,GAAG;AAC3B;AAEA,OAAO,SAASuB,QAAQ5B,EAAgB,EAAE6B,GAAW,EAAExB,KAAoB;IACzE,IAAIA,UAAU,MAAM;QAClBL,GAAGoB,OAAO,CAAC,kCAAkCY,GAAG,CAACH;QACjD;IACF;IACA7B,GAAGoB,OAAO,CAAC,qGAAqGY,GAAG,CAACH,KAAKxB;AAC3H;AAEA,OAAO,SAAS4B,SAASjC,EAAgB;IACvC,MAAM8B,MAAM9B,GAAGoB,OAAO,CAAC,yCAAyCW,GAAG;IACnE,OAAOD,IAAII,CAAC;AACd;AAEA,OAAO,SAASC,UAAUnC,EAAgB,EAAEoC,GAAW,EAAEC,OAAe;IACtE,MAAMC,QAAQ9C,UAAU4C,KAAKC;IAC7B,MAAME,aAAa,IAAIjB,IAAIgB,MAAMf,GAAG,CAAC,CAACiB,IAAMA,EAAEC,OAAO;IAErD,MAAMC,eAAe1C,GAAGoB,OAAO,CAAC,CAAC,iDAAiD,CAAC,EAAEC,GAAG;IAKxF,MAAMsB,WAAW,IAAIC,IAAIF,aAAanB,GAAG,CAAC,CAACC,IAAM;YAACA,EAAEqB,IAAI;YAAErB;SAAE;IAC5D,MAAMsB,WAAWJ,aAAaK,MAAM,CAAC,CAACvB,IAAM,CAACe,WAAWS,GAAG,CAACxB,EAAEqB,IAAI,GAAGtB,GAAG,CAAC,CAACC,IAAMA,EAAEqB,IAAI;IAEtF,MAAMI,YAAYX,MAAMS,MAAM,CAAC,CAACP;QAC9B,MAAMV,MAAMa,SAASZ,GAAG,CAACS,EAAEC,OAAO;QAClC,OAAO,CAACX,OAAOA,IAAIoB,MAAM,KAAKV,EAAEW,OAAO,IAAIrB,IAAIsB,KAAK,KAAKZ,EAAEa,IAAI;IACjE;IAEA,IAAIP,SAASQ,MAAM,KAAK,KAAKL,UAAUK,MAAM,KAAK,GAAG,OAAO;QAAE5C,QAAQ;QAAG6C,UAAU,EAAE;IAAC;IAEtF,MAAMC,cAActC,WAAWlB;IAC/B,MAAMyD,aAAuB,EAAE;IAC/B,MAAMC,aAA0B,EAAE;IAClC,MAAMH,WAAqB,EAAE;IAE7B,KAAK,MAAMI,QAAQV,UAAW;QAC5B,MAAM,EAAEW,GAAG,EAAEL,UAAUM,YAAY,EAAE,GAAGpE,UAAUkE;QAClDJ,SAASO,IAAI,IAAID;QACjB,KAAK,MAAMhC,OAAOkC,OAAOC,IAAI,CAACJ,IAAIK,IAAI,EAAG;YACvC,IAAI,CAACT,YAAYR,GAAG,CAACnB,MAAM;gBACzB2B,YAAYU,GAAG,CAACrC;gBAChB4B,WAAWK,IAAI,CAACjC;YAClB;QACF;QACA6B,WAAWI,IAAI,CAACF;IAClB;IAEA,MAAMO,aAAa;WAAIX;KAAY;IACnC,MAAMY,YAAY,CAAC,oCAAoC,EAAED,WAAW5C,GAAG,CAAC3B,YAAYP,IAAI,CAAC,MAAM,UAAU,EAAE8E,WAAW5C,GAAG,CAAC,IAAM,KAAKlC,IAAI,CAAC,MAAM,CAAC,CAAC;IAElJW,GAAG0B,IAAI,CAAC;IACR,IAAI;QACF,KAAK,MAAM2C,OAAOZ,WAAYzD,GAAG0B,IAAI,CAAC,CAAC,mCAAmC,EAAE9B,WAAWyE,MAAM;QAC7F,8DAA8D;QAC9D,MAAMC,UAAUtE,GAAGoB,OAAO,CAAC,CAAC,oCAAoC,CAAC;QACjE,IAAI0B,SAASQ,MAAM,GAAG,GAAG;YACvB,MAAMiB,MAAMvE,GAAGoB,OAAO,CAAC,CAAC,wCAAwC,CAAC;YACjE,KAAK,MAAMyB,QAAQC,SAAU;gBAC3ByB,IAAIvC,GAAG,CAACa;gBACRyB,QAAQtC,GAAG,CAACa;YACd;QACF;QACA,IAAIa,WAAWJ,MAAM,GAAG,GAAG;YACzB,MAAMkB,SAASxE,GAAGoB,OAAO,CAACgD;YAC1B,MAAMK,aAAazE,GAAGoB,OAAO,CAAC,CAAC,sEAAsE,CAAC;YACtG,KAAK,MAAMwC,OAAOF,WAAY;gBAC5B,MAAMgB,SAASP,WAAW5C,GAAG,CAAC,CAAC8C;wBAItBT;oBAHP,IAAIS,QAAQ,QAAQ,OAAOT,IAAInB,OAAO;oBACtC,IAAI4B,QAAQ,UAAU,OAAOT,IAAIT,OAAO;oBACxC,IAAIkB,QAAQ,SAAS,OAAOT,IAAIP,IAAI;oBACpC,QAAOO,gBAAAA,IAAIK,IAAI,CAACI,IAAI,cAAbT,2BAAAA,gBAAiB;gBAC1B;gBACAY,OAAOxC,GAAG,IAAI0C;gBACdJ,QAAQtC,GAAG,CAAC4B,IAAInB,OAAO;gBACvBgC,WAAWzC,GAAG,CAAC4B,IAAIe,MAAM,CAACC,KAAK,EAAEhB,IAAIe,MAAM,CAACE,OAAO,EAAEjB,IAAIe,MAAM,CAACG,IAAI,EAAElB,IAAInB,OAAO;YACnF;QACF;QACAzC,GAAG0B,IAAI,CAAC;IACV,EAAE,OAAOqD,KAAK;QACZ/E,GAAG0B,IAAI,CAAC;QACR,MAAMqD;IACR;IAEA,OAAO;QAAErE,QAAQgD,WAAWJ,MAAM;QAAEC;IAAS;AAC/C;AAEA,OAAO,SAASyB,KAAK5C,GAAmB;IACtC,MAAM6C,WAAW5F,KAAK+C,IAAIC,OAAO,EAAE9C;IACnCJ,UAAU8F,UAAU;QAAEC,WAAW;IAAK;IACtC,MAAMC,SAAS9F,KAAK4F,UAAUvF;IAE9B,MAAMM,KAAK,IAAIV,aAAa6F;IAC5BnF,GAAG0B,IAAI,CAAC;IACR1B,GAAG0B,IAAI,CAAC;IACR3B,kBAAkBC;IAElBA,GAAG0B,IAAI,CAAC;IAER,wIAAwI;IACxI,MAAM0D,UAAUzD,QAAQ3B,IAAI;IAC5B,IAAIoF,YAAY,QAAQA,YAAYzF,gBAAgB;QAClDK,GAAGqF,KAAK;QACRjG,OAAO6F,UAAU;YAAEC,WAAW;YAAMI,OAAO;QAAK;QAChD,OAAON,KAAK5C;IACd;IAEAX,aAAazB;IACb,MAAM,EAAEU,MAAM,EAAE6C,QAAQ,EAAE,GAAGpB,UAAUnC,IAAIoC,KAAKA,IAAIC,OAAO;IAE3D,OAAO;QAAErC;QAAIoC;QAAK+C;QAAQzE;QAAQ6C;IAAS;AAC7C;AAEA,oCAAoC;AACpC,OAAO,SAASgC,QAAQnD,GAAmB;IACzChD,OAAOC,KAAK+C,IAAIC,OAAO,EAAE9C,YAAY;QAAE2F,WAAW;QAAMI,OAAO;IAAK;IACpE,OAAON,KAAK5C;AACd"}
@@ -1,6 +1,4 @@
1
- // Typed errors for library modules. Library code (config.ts, scan.ts,
2
- // db.ts, watch.ts) never calls process.exit and never prints -- it throws.
3
- // Only cli.ts catches, prints the message, and maps it to an exit code.
1
+ // Library code throws; only cli.ts prints and exits.
4
2
  export class SenseError extends Error {
5
3
  constructor(code, message){
6
4
  super(message);
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/errors.ts"],"sourcesContent":["// Typed errors for library modules. Library code (config.ts, scan.ts,\n// db.ts, watch.ts) never calls process.exit and never prints -- it throws.\n// Only cli.ts catches, prints the message, and maps it to an exit code.\nexport type SenseErrorCode = 'CONFIG_NOT_FOUND' | 'CONFIG_EXISTS' | 'CONFIG_VERSION_UNSUPPORTED' | 'WATCH_ACTIVE';\n\nexport class SenseError extends Error {\n code: SenseErrorCode;\n\n constructor(code: SenseErrorCode, message: string) {\n super(message);\n this.name = 'SenseError';\n this.code = code;\n }\n}\n"],"names":["SenseError","Error","code","message","name"],"mappings":"AAAA,sEAAsE;AACtE,2EAA2E;AAC3E,wEAAwE;AAGxE,OAAO,MAAMA,mBAAmBC;IAG9B,YAAYC,IAAoB,EAAEC,OAAe,CAAE;QACjD,KAAK,CAACA;QACN,IAAI,CAACC,IAAI,GAAG;QACZ,IAAI,CAACF,IAAI,GAAGA;IACd;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/errors.ts"],"sourcesContent":["// Library code throws; only cli.ts prints and exits.\nexport type SenseErrorCode = 'CONFIG_NOT_FOUND' | 'CONFIG_EXISTS' | 'CONFIG_VERSION_UNSUPPORTED' | 'WATCH_ACTIVE';\n\nexport class SenseError extends Error {\n code: SenseErrorCode;\n\n constructor(code: SenseErrorCode, message: string) {\n super(message);\n this.name = 'SenseError';\n this.code = code;\n }\n}\n"],"names":["SenseError","Error","code","message","name"],"mappings":"AAAA,qDAAqD;AAGrD,OAAO,MAAMA,mBAAmBC;IAG9B,YAAYC,IAAoB,EAAEC,OAAe,CAAE;QACjD,KAAK,CAACA;QACN,IAAI,CAACC,IAAI,GAAG;QACZ,IAAI,CAACF,IAAI,GAAGA;IACd;AACF"}
package/dist/esm/index.js CHANGED
@@ -1,6 +1,4 @@
1
- // Public library API. cli.ts is a consumer of this surface, not a source of
2
- // truth for it -- anything a test or another tool needs from sense should be
3
- // exported here rather than reached into src/ by relative path.
1
+ // Public library API.
4
2
  export { CONFIG_FILENAME, findConfigPath, initConfig, loadConfig, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './config.js';
5
3
  export { DB_FILENAME, docCount, getMeta, open, rebuild, reconcile, setMeta } from './db.js';
6
4
  export { SenseError } from './errors.js';
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/index.ts"],"sourcesContent":["// Public library API. cli.ts is a consumer of this surface, not a source of\n// truth for it -- anything a test or another tool needs from sense should be\n// exported here rather than reached into src/ by relative path.\n\nexport type { Config, ResolvedConfig } from './config.ts';\nexport { CONFIG_FILENAME, findConfigPath, initConfig, loadConfig, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './config.ts';\n\nexport type { OpenResult } from './db.ts';\nexport { DB_FILENAME, docCount, getMeta, open, rebuild, reconcile, setMeta } from './db.ts';\n\nexport type { SenseErrorCode } from './errors.ts';\nexport { SenseError } from './errors.ts';\n\nexport type { Row } from './output.ts';\nexport { printRows } from './output.ts';\n\nexport type { FileStat, ParsedDoc } from './scan.ts';\nexport { listFiles, parseFile } from './scan.ts';\n\nexport type { WatchEvent, WatchOptions } from './watch.ts';\nexport { runWatch } from './watch.ts';\n"],"names":["CONFIG_FILENAME","findConfigPath","initConfig","loadConfig","STATE_DIR","SUPPORTED_CONFIG_VERSION","DB_FILENAME","docCount","getMeta","open","rebuild","reconcile","setMeta","SenseError","printRows","listFiles","parseFile","runWatch"],"mappings":"AAAA,4EAA4E;AAC5E,6EAA6E;AAC7E,gEAAgE;AAGhE,SAASA,eAAe,EAAEC,cAAc,EAAEC,UAAU,EAAEC,UAAU,EAAEC,SAAS,EAAEC,wBAAwB,QAAQ,cAAc;AAG3H,SAASC,WAAW,EAAEC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,OAAO,EAAEC,SAAS,EAAEC,OAAO,QAAQ,UAAU;AAG5F,SAASC,UAAU,QAAQ,cAAc;AAGzC,SAASC,SAAS,QAAQ,cAAc;AAGxC,SAASC,SAAS,EAAEC,SAAS,QAAQ,YAAY;AAGjD,SAASC,QAAQ,QAAQ,aAAa"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/index.ts"],"sourcesContent":["// Public library API.\n\nexport type { Config, ResolvedConfig } from './config.ts';\nexport { CONFIG_FILENAME, findConfigPath, initConfig, loadConfig, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './config.ts';\n\nexport type { OpenResult } from './db.ts';\nexport { DB_FILENAME, docCount, getMeta, open, rebuild, reconcile, setMeta } from './db.ts';\n\nexport type { SenseErrorCode } from './errors.ts';\nexport { SenseError } from './errors.ts';\n\nexport type { Row } from './output.ts';\nexport { printRows } from './output.ts';\n\nexport type { FileStat, ParsedDoc } from './scan.ts';\nexport { listFiles, parseFile } from './scan.ts';\n\nexport type { WatchEvent, WatchOptions } from './watch.ts';\nexport { runWatch } from './watch.ts';\n"],"names":["CONFIG_FILENAME","findConfigPath","initConfig","loadConfig","STATE_DIR","SUPPORTED_CONFIG_VERSION","DB_FILENAME","docCount","getMeta","open","rebuild","reconcile","setMeta","SenseError","printRows","listFiles","parseFile","runWatch"],"mappings":"AAAA,sBAAsB;AAGtB,SAASA,eAAe,EAAEC,cAAc,EAAEC,UAAU,EAAEC,UAAU,EAAEC,SAAS,EAAEC,wBAAwB,QAAQ,cAAc;AAG3H,SAASC,WAAW,EAAEC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,OAAO,EAAEC,SAAS,EAAEC,OAAO,QAAQ,UAAU;AAG5F,SAASC,UAAU,QAAQ,cAAc;AAGzC,SAASC,SAAS,QAAQ,cAAc;AAGxC,SAASC,SAAS,EAAEC,SAAS,QAAQ,YAAY;AAGjD,SAASC,QAAQ,QAAQ,aAAa"}
@@ -1,25 +1,27 @@
1
1
  // rows -> table (default) | json
2
+ // Warned about on stderr (keeps --format json stdout machine-readable), never truncated.
3
+ const OVERSIZE_BYTES = 50000;
4
+ // Flatten embedded newlines so they can't break table column alignment; JSON output is left alone.
5
+ function cell(value) {
6
+ return String(value !== null && value !== void 0 ? value : '').replace(/\s*\n\s*/g, ' ');
7
+ }
2
8
  export function printRows(rows, format) {
3
- if (format === 'json') {
4
- console.log(JSON.stringify(rows, null, 2));
5
- return;
6
- }
7
- if (rows.length === 0) {
8
- console.log('(0 rows)');
9
- return;
9
+ const rendered = render(rows, format);
10
+ console.log(rendered);
11
+ if (rendered.length > OVERSIZE_BYTES) {
12
+ console.warn(`warning: result is ${Math.round(rendered.length / 1000)} KB across ${rows.length} row(s); consider a LIMIT, fewer columns, or snippet() instead of whole values`);
10
13
  }
14
+ }
15
+ function render(rows, format) {
16
+ if (format === 'json') return JSON.stringify(rows, null, 2);
17
+ if (rows.length === 0) return '(0 rows)';
11
18
  const columns = Object.keys(rows[0]);
12
- const widths = columns.map((col)=>Math.max(col.length, ...rows.map((row)=>{
13
- var _row_col;
14
- return String((_row_col = row[col]) !== null && _row_col !== void 0 ? _row_col : '').length;
15
- })));
19
+ const cells = rows.map((row)=>columns.map((col)=>cell(row[col])));
20
+ const widths = columns.map((col, i)=>Math.max(col.length, ...cells.map((row)=>row[i].length)));
16
21
  const formatRow = (values)=>values.map((value, i)=>value.padEnd(widths[i])).join(' ');
17
- console.log(formatRow(columns));
18
- console.log(widths.map((w)=>'-'.repeat(w)).join(' '));
19
- for (const row of rows){
20
- console.log(formatRow(columns.map((col)=>{
21
- var _row_col;
22
- return String((_row_col = row[col]) !== null && _row_col !== void 0 ? _row_col : '');
23
- })));
24
- }
22
+ return [
23
+ formatRow(columns),
24
+ widths.map((w)=>'-'.repeat(w)).join(' '),
25
+ ...cells.map(formatRow)
26
+ ].join('\n');
25
27
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/output.ts"],"sourcesContent":["// rows -> table (default) | json\n\nexport type Row = Record<string, unknown>;\n\nexport function printRows(rows: Row[], format: 'table' | 'json'): void {\n if (format === 'json') {\n console.log(JSON.stringify(rows, null, 2));\n return;\n }\n\n if (rows.length === 0) {\n console.log('(0 rows)');\n return;\n }\n\n const columns = Object.keys(rows[0]);\n const widths = columns.map((col) => Math.max(col.length, ...rows.map((row) => String(row[col] ?? '').length)));\n\n const formatRow = (values: string[]) => values.map((value, i) => value.padEnd(widths[i])).join(' ');\n\n console.log(formatRow(columns));\n console.log(widths.map((w) => '-'.repeat(w)).join(' '));\n for (const row of rows) {\n console.log(formatRow(columns.map((col) => String(row[col] ?? ''))));\n }\n}\n"],"names":["printRows","rows","format","console","log","JSON","stringify","length","columns","Object","keys","widths","map","col","Math","max","row","String","formatRow","values","value","i","padEnd","join","w","repeat"],"mappings":"AAAA,iCAAiC;AAIjC,OAAO,SAASA,UAAUC,IAAW,EAAEC,MAAwB;IAC7D,IAAIA,WAAW,QAAQ;QACrBC,QAAQC,GAAG,CAACC,KAAKC,SAAS,CAACL,MAAM,MAAM;QACvC;IACF;IAEA,IAAIA,KAAKM,MAAM,KAAK,GAAG;QACrBJ,QAAQC,GAAG,CAAC;QACZ;IACF;IAEA,MAAMI,UAAUC,OAAOC,IAAI,CAACT,IAAI,CAAC,EAAE;IACnC,MAAMU,SAASH,QAAQI,GAAG,CAAC,CAACC,MAAQC,KAAKC,GAAG,CAACF,IAAIN,MAAM,KAAKN,KAAKW,GAAG,CAAC,CAACI;gBAAeA;mBAAPC,QAAOD,WAAAA,GAAG,CAACH,IAAI,cAARG,sBAAAA,WAAY,IAAIT,MAAM;;IAE3G,MAAMW,YAAY,CAACC,SAAqBA,OAAOP,GAAG,CAAC,CAACQ,OAAOC,IAAMD,MAAME,MAAM,CAACX,MAAM,CAACU,EAAE,GAAGE,IAAI,CAAC;IAE/FpB,QAAQC,GAAG,CAACc,UAAUV;IACtBL,QAAQC,GAAG,CAACO,OAAOC,GAAG,CAAC,CAACY,IAAM,IAAIC,MAAM,CAACD,IAAID,IAAI,CAAC;IAClD,KAAK,MAAMP,OAAOf,KAAM;QACtBE,QAAQC,GAAG,CAACc,UAAUV,QAAQI,GAAG,CAAC,CAACC;gBAAeG;mBAAPC,QAAOD,WAAAA,GAAG,CAACH,IAAI,cAARG,sBAAAA,WAAY;;IAChE;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/output.ts"],"sourcesContent":["// rows -> table (default) | json\n\nexport type Row = Record<string, unknown>;\n\n// Warned about on stderr (keeps --format json stdout machine-readable), never truncated.\nconst OVERSIZE_BYTES = 50_000;\n\n// Flatten embedded newlines so they can't break table column alignment; JSON output is left alone.\nfunction cell(value: unknown): string {\n return String(value ?? '').replace(/\\s*\\n\\s*/g, ' ');\n}\n\nexport function printRows(rows: Row[], format: 'table' | 'json'): void {\n const rendered = render(rows, format);\n console.log(rendered);\n if (rendered.length > OVERSIZE_BYTES) {\n console.warn(`warning: result is ${Math.round(rendered.length / 1000)} KB across ${rows.length} row(s); consider a LIMIT, fewer columns, or snippet() instead of whole values`);\n }\n}\n\nfunction render(rows: Row[], format: 'table' | 'json'): string {\n if (format === 'json') return JSON.stringify(rows, null, 2);\n if (rows.length === 0) return '(0 rows)';\n\n const columns = Object.keys(rows[0]);\n const cells = rows.map((row) => columns.map((col) => cell(row[col])));\n const widths = columns.map((col, i) => Math.max(col.length, ...cells.map((row) => row[i].length)));\n\n const formatRow = (values: string[]) => values.map((value, i) => value.padEnd(widths[i])).join(' ');\n\n return [formatRow(columns), widths.map((w) => '-'.repeat(w)).join(' '), ...cells.map(formatRow)].join('\\n');\n}\n"],"names":["OVERSIZE_BYTES","cell","value","String","replace","printRows","rows","format","rendered","render","console","log","length","warn","Math","round","JSON","stringify","columns","Object","keys","cells","map","row","col","widths","i","max","formatRow","values","padEnd","join","w","repeat"],"mappings":"AAAA,iCAAiC;AAIjC,yFAAyF;AACzF,MAAMA,iBAAiB;AAEvB,mGAAmG;AACnG,SAASC,KAAKC,KAAc;IAC1B,OAAOC,OAAOD,kBAAAA,mBAAAA,QAAS,IAAIE,OAAO,CAAC,aAAa;AAClD;AAEA,OAAO,SAASC,UAAUC,IAAW,EAAEC,MAAwB;IAC7D,MAAMC,WAAWC,OAAOH,MAAMC;IAC9BG,QAAQC,GAAG,CAACH;IACZ,IAAIA,SAASI,MAAM,GAAGZ,gBAAgB;QACpCU,QAAQG,IAAI,CAAC,CAAC,mBAAmB,EAAEC,KAAKC,KAAK,CAACP,SAASI,MAAM,GAAG,MAAM,WAAW,EAAEN,KAAKM,MAAM,CAAC,8EAA8E,CAAC;IAChL;AACF;AAEA,SAASH,OAAOH,IAAW,EAAEC,MAAwB;IACnD,IAAIA,WAAW,QAAQ,OAAOS,KAAKC,SAAS,CAACX,MAAM,MAAM;IACzD,IAAIA,KAAKM,MAAM,KAAK,GAAG,OAAO;IAE9B,MAAMM,UAAUC,OAAOC,IAAI,CAACd,IAAI,CAAC,EAAE;IACnC,MAAMe,QAAQf,KAAKgB,GAAG,CAAC,CAACC,MAAQL,QAAQI,GAAG,CAAC,CAACE,MAAQvB,KAAKsB,GAAG,CAACC,IAAI;IAClE,MAAMC,SAASP,QAAQI,GAAG,CAAC,CAACE,KAAKE,IAAMZ,KAAKa,GAAG,CAACH,IAAIZ,MAAM,KAAKS,MAAMC,GAAG,CAAC,CAACC,MAAQA,GAAG,CAACG,EAAE,CAACd,MAAM;IAE/F,MAAMgB,YAAY,CAACC,SAAqBA,OAAOP,GAAG,CAAC,CAACpB,OAAOwB,IAAMxB,MAAM4B,MAAM,CAACL,MAAM,CAACC,EAAE,GAAGK,IAAI,CAAC;IAE/F,OAAO;QAACH,UAAUV;QAAUO,OAAOH,GAAG,CAAC,CAACU,IAAM,IAAIC,MAAM,CAACD,IAAID,IAAI,CAAC;WAAUV,MAAMC,GAAG,CAACM;KAAW,CAACG,IAAI,CAAC;AACzG"}
@@ -11,6 +11,11 @@ export interface ParsedDoc {
11
11
  mtimeMs: number;
12
12
  size: number;
13
13
  data: Record<string, string | number | null>;
14
+ search: {
15
+ title: string;
16
+ summary: string;
17
+ text: string;
18
+ };
14
19
  }
15
20
  export declare function parseFile(file: FileStat): {
16
21
  doc: ParsedDoc;
package/dist/esm/scan.js CHANGED
@@ -2,20 +2,27 @@ import { readFileSync, statSync } from 'node:fs';
2
2
  import { join } from 'node:path';
3
3
  import fastGlob from 'fast-glob';
4
4
  import matter from 'gray-matter';
5
- // Filesystem -> rows. Pure data in, data + warnings out -- no node:sqlite,
6
- // no printing. db.ts is the only thing that knows what to do with the
7
- // result (diffing against `docs`, writing SQL).
8
- // Reserved `docs` columns: the real file path plus the mtime/size pair used
9
- // to detect staleness. A frontmatter key literally named one of these would
10
- // collide, so it's dropped (with a warning returned to the caller) rather
11
- // than clobbering the column.
5
+ import removeMarkdown from 'remove-markdown';
6
+ // Filesystem -> rows. Pure data in, data + warnings out; db.ts does the SQL.
7
+ // Reserved: colliding with these would clash with the `frontmatter` table's own columns or the `content` FTS5 table.
12
8
  const RESERVED_COLUMNS = new Set([
13
9
  'path',
14
10
  '_mtime',
15
- '_size'
11
+ '_size',
12
+ 'content'
16
13
  ]);
17
- // Glob the files `cfg.scan.include` matches under `baseDir`, stat each.
18
- // Sorted for deterministic output.
14
+ function normalizeText(value) {
15
+ if (value === null || value === undefined) return '';
16
+ return String(value).replace(/\s+/g, ' ').trim();
17
+ }
18
+ // Markdown is stripped at index time so snippets read as clean prose (also improves matching, e.g. `**bold**` indexes as `bold`).
19
+ // remove-markdown doesn't cover Obsidian wikilinks or table scaffolding, so those are handled first/after.
20
+ function stripText(value) {
21
+ const withoutWikilinks = value.replace(/\[\[([^\]|]+)\|([^\]]+)\]\]/g, '$2').replace(/\[\[([^\]]+)\]\]/g, '$1');
22
+ const withoutMarkdown = removeMarkdown(withoutWikilinks);
23
+ const withoutTables = withoutMarkdown.replace(/^\s*\|?[-\s|:]+\|\s*$/gm, '').replace(/\|/g, ' ');
24
+ return normalizeText(withoutTables);
25
+ }
19
26
  export function listFiles(cfg, baseDir) {
20
27
  const relPaths = fastGlob.sync(cfg.scan.include, {
21
28
  cwd: baseDir
@@ -31,23 +38,16 @@ export function listFiles(cfg, baseDir) {
31
38
  };
32
39
  });
33
40
  }
34
- // Value mapping: strings/numbers as-is; booleans -> 0/1; dates -> ISO
35
- // strings (lexicographic = chronological, no date lib); arrays/objects ->
36
- // JSON text. `null`/`undefined` map to SQL NULL.
37
41
  function mapValue(value) {
38
42
  if (value === null || value === undefined) return null;
39
43
  if (value instanceof Date) return value.toISOString();
40
44
  if (typeof value === 'boolean') return value ? 1 : 0;
41
45
  if (typeof value === 'string' || typeof value === 'number') return value;
42
- // arrays and plain objects
43
46
  return JSON.stringify(value);
44
47
  }
45
- // Read + parse one file's frontmatter, mapping values and dropping reserved
46
- // keys. Returns the parsed doc plus any warnings (e.g. a reserved key
47
- // collision) for the caller to surface.
48
48
  export function parseFile(file) {
49
49
  const raw = readFileSync(file.absPath, 'utf8');
50
- const { data } = matter(raw);
50
+ const { data, content } = matter(raw);
51
51
  const warnings = [];
52
52
  const mapped = {};
53
53
  for (const key of Object.keys(data)){
@@ -62,7 +62,14 @@ export function parseFile(file) {
62
62
  relPath: file.relPath,
63
63
  mtimeMs: file.mtimeMs,
64
64
  size: file.size,
65
- data: mapped
65
+ data: mapped,
66
+ search: {
67
+ // title/summary are plain YAML strings -- whitespace-collapse only;
68
+ // the prose gets the full markdown strip.
69
+ title: normalizeText(data.title),
70
+ summary: normalizeText(data.summary),
71
+ text: stripText(content)
72
+ }
66
73
  },
67
74
  warnings
68
75
  };
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan.ts"],"sourcesContent":["import { readFileSync, statSync } from 'node:fs';\nimport { join } from 'node:path';\nimport fastGlob from 'fast-glob';\nimport matter from 'gray-matter';\nimport type { Config } from './config.ts';\n\n// Filesystem -> rows. Pure data in, data + warnings out -- no node:sqlite,\n// no printing. db.ts is the only thing that knows what to do with the\n// result (diffing against `docs`, writing SQL).\n\n// Reserved `docs` columns: the real file path plus the mtime/size pair used\n// to detect staleness. A frontmatter key literally named one of these would\n// collide, so it's dropped (with a warning returned to the caller) rather\n// than clobbering the column.\nconst RESERVED_COLUMNS = new Set(['path', '_mtime', '_size']);\n\nexport interface FileStat {\n relPath: string;\n absPath: string;\n mtimeMs: number;\n size: number;\n}\n\n// Glob the files `cfg.scan.include` matches under `baseDir`, stat each.\n// Sorted for deterministic output.\nexport function listFiles(cfg: Config, baseDir: string): FileStat[] {\n const relPaths = fastGlob.sync(cfg.scan.include, { cwd: baseDir }).sort();\n return relPaths.map((relPath) => {\n const absPath = join(baseDir, relPath);\n const st = statSync(absPath);\n return { relPath, absPath, mtimeMs: st.mtimeMs, size: st.size };\n });\n}\n\nexport interface ParsedDoc {\n relPath: string;\n mtimeMs: number;\n size: number;\n // Already value-mapped (booleans -> 0/1, dates -> ISO, arrays/objects ->\n // JSON text), reserved keys already dropped -- ready to bind into SQL.\n data: Record<string, string | number | null>;\n}\n\n// Value mapping: strings/numbers as-is; booleans -> 0/1; dates -> ISO\n// strings (lexicographic = chronological, no date lib); arrays/objects ->\n// JSON text. `null`/`undefined` map to SQL NULL.\nfunction mapValue(value: unknown): string | number | null {\n if (value === null || value === undefined) return null;\n if (value instanceof Date) return value.toISOString();\n if (typeof value === 'boolean') return value ? 1 : 0;\n if (typeof value === 'string' || typeof value === 'number') return value;\n // arrays and plain objects\n return JSON.stringify(value);\n}\n\n// Read + parse one file's frontmatter, mapping values and dropping reserved\n// keys. Returns the parsed doc plus any warnings (e.g. a reserved key\n// collision) for the caller to surface.\nexport function parseFile(file: FileStat): { doc: ParsedDoc; warnings: string[] } {\n const raw = readFileSync(file.absPath, 'utf8');\n const { data } = matter(raw);\n\n const warnings: string[] = [];\n const mapped: Record<string, string | number | null> = {};\n\n for (const key of Object.keys(data)) {\n if (RESERVED_COLUMNS.has(key)) {\n warnings.push(`warning: ${file.relPath} has a frontmatter key named \"${key}\", which is reserved; ignoring it`);\n continue;\n }\n mapped[key] = mapValue(data[key]);\n }\n\n return {\n doc: { relPath: file.relPath, mtimeMs: file.mtimeMs, size: file.size, data: mapped },\n warnings,\n };\n}\n"],"names":["readFileSync","statSync","join","fastGlob","matter","RESERVED_COLUMNS","Set","listFiles","cfg","baseDir","relPaths","sync","scan","include","cwd","sort","map","relPath","absPath","st","mtimeMs","size","mapValue","value","undefined","Date","toISOString","JSON","stringify","parseFile","file","raw","data","warnings","mapped","key","Object","keys","has","push","doc"],"mappings":"AAAA,SAASA,YAAY,EAAEC,QAAQ,QAAQ,UAAU;AACjD,SAASC,IAAI,QAAQ,YAAY;AACjC,OAAOC,cAAc,YAAY;AACjC,OAAOC,YAAY,cAAc;AAGjC,2EAA2E;AAC3E,sEAAsE;AACtE,gDAAgD;AAEhD,4EAA4E;AAC5E,4EAA4E;AAC5E,0EAA0E;AAC1E,8BAA8B;AAC9B,MAAMC,mBAAmB,IAAIC,IAAI;IAAC;IAAQ;IAAU;CAAQ;AAS5D,wEAAwE;AACxE,mCAAmC;AACnC,OAAO,SAASC,UAAUC,GAAW,EAAEC,OAAe;IACpD,MAAMC,WAAWP,SAASQ,IAAI,CAACH,IAAII,IAAI,CAACC,OAAO,EAAE;QAAEC,KAAKL;IAAQ,GAAGM,IAAI;IACvE,OAAOL,SAASM,GAAG,CAAC,CAACC;QACnB,MAAMC,UAAUhB,KAAKO,SAASQ;QAC9B,MAAME,KAAKlB,SAASiB;QACpB,OAAO;YAAED;YAASC;YAASE,SAASD,GAAGC,OAAO;YAAEC,MAAMF,GAAGE,IAAI;QAAC;IAChE;AACF;AAWA,sEAAsE;AACtE,0EAA0E;AAC1E,iDAAiD;AACjD,SAASC,SAASC,KAAc;IAC9B,IAAIA,UAAU,QAAQA,UAAUC,WAAW,OAAO;IAClD,IAAID,iBAAiBE,MAAM,OAAOF,MAAMG,WAAW;IACnD,IAAI,OAAOH,UAAU,WAAW,OAAOA,QAAQ,IAAI;IACnD,IAAI,OAAOA,UAAU,YAAY,OAAOA,UAAU,UAAU,OAAOA;IACnE,2BAA2B;IAC3B,OAAOI,KAAKC,SAAS,CAACL;AACxB;AAEA,4EAA4E;AAC5E,sEAAsE;AACtE,wCAAwC;AACxC,OAAO,SAASM,UAAUC,IAAc;IACtC,MAAMC,MAAM/B,aAAa8B,KAAKZ,OAAO,EAAE;IACvC,MAAM,EAAEc,IAAI,EAAE,GAAG5B,OAAO2B;IAExB,MAAME,WAAqB,EAAE;IAC7B,MAAMC,SAAiD,CAAC;IAExD,KAAK,MAAMC,OAAOC,OAAOC,IAAI,CAACL,MAAO;QACnC,IAAI3B,iBAAiBiC,GAAG,CAACH,MAAM;YAC7BF,SAASM,IAAI,CAAC,CAAC,SAAS,EAAET,KAAKb,OAAO,CAAC,8BAA8B,EAAEkB,IAAI,iCAAiC,CAAC;YAC7G;QACF;QACAD,MAAM,CAACC,IAAI,GAAGb,SAASU,IAAI,CAACG,IAAI;IAClC;IAEA,OAAO;QACLK,KAAK;YAAEvB,SAASa,KAAKb,OAAO;YAAEG,SAASU,KAAKV,OAAO;YAAEC,MAAMS,KAAKT,IAAI;YAAEW,MAAME;QAAO;QACnFD;IACF;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan.ts"],"sourcesContent":["import { readFileSync, statSync } from 'node:fs';\nimport { join } from 'node:path';\nimport fastGlob from 'fast-glob';\nimport matter from 'gray-matter';\nimport removeMarkdown from 'remove-markdown';\nimport type { Config } from './config.ts';\n\n// Filesystem -> rows. Pure data in, data + warnings out; db.ts does the SQL.\n\n// Reserved: colliding with these would clash with the `frontmatter` table's own columns or the `content` FTS5 table.\nconst RESERVED_COLUMNS = new Set(['path', '_mtime', '_size', 'content']);\n\nfunction normalizeText(value: unknown): string {\n if (value === null || value === undefined) return '';\n return String(value).replace(/\\s+/g, ' ').trim();\n}\n\n// Markdown is stripped at index time so snippets read as clean prose (also improves matching, e.g. `**bold**` indexes as `bold`).\n// remove-markdown doesn't cover Obsidian wikilinks or table scaffolding, so those are handled first/after.\nfunction stripText(value: string): string {\n const withoutWikilinks = value.replace(/\\[\\[([^\\]|]+)\\|([^\\]]+)\\]\\]/g, '$2').replace(/\\[\\[([^\\]]+)\\]\\]/g, '$1');\n const withoutMarkdown = removeMarkdown(withoutWikilinks);\n const withoutTables = withoutMarkdown.replace(/^\\s*\\|?[-\\s|:]+\\|\\s*$/gm, '').replace(/\\|/g, ' ');\n return normalizeText(withoutTables);\n}\n\nexport interface FileStat {\n relPath: string;\n absPath: string;\n mtimeMs: number;\n size: number;\n}\n\nexport function listFiles(cfg: Config, baseDir: string): FileStat[] {\n const relPaths = fastGlob.sync(cfg.scan.include, { cwd: baseDir }).sort();\n return relPaths.map((relPath) => {\n const absPath = join(baseDir, relPath);\n const st = statSync(absPath);\n return { relPath, absPath, mtimeMs: st.mtimeMs, size: st.size };\n });\n}\n\nexport interface ParsedDoc {\n relPath: string;\n mtimeMs: number;\n size: number;\n data: Record<string, string | number | null>;\n // title/summary are duplicated from frontmatter so bm25() can weight them above the body text.\n search: { title: string; summary: string; text: string };\n}\n\nfunction mapValue(value: unknown): string | number | null {\n if (value === null || value === undefined) return null;\n if (value instanceof Date) return value.toISOString();\n if (typeof value === 'boolean') return value ? 1 : 0;\n if (typeof value === 'string' || typeof value === 'number') return value;\n return JSON.stringify(value);\n}\n\nexport function parseFile(file: FileStat): { doc: ParsedDoc; warnings: string[] } {\n const raw = readFileSync(file.absPath, 'utf8');\n const { data, content } = matter(raw);\n\n const warnings: string[] = [];\n const mapped: Record<string, string | number | null> = {};\n\n for (const key of Object.keys(data)) {\n if (RESERVED_COLUMNS.has(key)) {\n warnings.push(`warning: ${file.relPath} has a frontmatter key named \"${key}\", which is reserved; ignoring it`);\n continue;\n }\n mapped[key] = mapValue(data[key]);\n }\n\n return {\n doc: {\n relPath: file.relPath,\n mtimeMs: file.mtimeMs,\n size: file.size,\n data: mapped,\n search: {\n // title/summary are plain YAML strings -- whitespace-collapse only;\n // the prose gets the full markdown strip.\n title: normalizeText(data.title),\n summary: normalizeText(data.summary),\n text: stripText(content),\n },\n },\n warnings,\n };\n}\n"],"names":["readFileSync","statSync","join","fastGlob","matter","removeMarkdown","RESERVED_COLUMNS","Set","normalizeText","value","undefined","String","replace","trim","stripText","withoutWikilinks","withoutMarkdown","withoutTables","listFiles","cfg","baseDir","relPaths","sync","scan","include","cwd","sort","map","relPath","absPath","st","mtimeMs","size","mapValue","Date","toISOString","JSON","stringify","parseFile","file","raw","data","content","warnings","mapped","key","Object","keys","has","push","doc","search","title","summary","text"],"mappings":"AAAA,SAASA,YAAY,EAAEC,QAAQ,QAAQ,UAAU;AACjD,SAASC,IAAI,QAAQ,YAAY;AACjC,OAAOC,cAAc,YAAY;AACjC,OAAOC,YAAY,cAAc;AACjC,OAAOC,oBAAoB,kBAAkB;AAG7C,6EAA6E;AAE7E,qHAAqH;AACrH,MAAMC,mBAAmB,IAAIC,IAAI;IAAC;IAAQ;IAAU;IAAS;CAAU;AAEvE,SAASC,cAAcC,KAAc;IACnC,IAAIA,UAAU,QAAQA,UAAUC,WAAW,OAAO;IAClD,OAAOC,OAAOF,OAAOG,OAAO,CAAC,QAAQ,KAAKC,IAAI;AAChD;AAEA,kIAAkI;AAClI,2GAA2G;AAC3G,SAASC,UAAUL,KAAa;IAC9B,MAAMM,mBAAmBN,MAAMG,OAAO,CAAC,gCAAgC,MAAMA,OAAO,CAAC,qBAAqB;IAC1G,MAAMI,kBAAkBX,eAAeU;IACvC,MAAME,gBAAgBD,gBAAgBJ,OAAO,CAAC,2BAA2B,IAAIA,OAAO,CAAC,OAAO;IAC5F,OAAOJ,cAAcS;AACvB;AASA,OAAO,SAASC,UAAUC,GAAW,EAAEC,OAAe;IACpD,MAAMC,WAAWlB,SAASmB,IAAI,CAACH,IAAII,IAAI,CAACC,OAAO,EAAE;QAAEC,KAAKL;IAAQ,GAAGM,IAAI;IACvE,OAAOL,SAASM,GAAG,CAAC,CAACC;QACnB,MAAMC,UAAU3B,KAAKkB,SAASQ;QAC9B,MAAME,KAAK7B,SAAS4B;QACpB,OAAO;YAAED;YAASC;YAASE,SAASD,GAAGC,OAAO;YAAEC,MAAMF,GAAGE,IAAI;QAAC;IAChE;AACF;AAWA,SAASC,SAASxB,KAAc;IAC9B,IAAIA,UAAU,QAAQA,UAAUC,WAAW,OAAO;IAClD,IAAID,iBAAiByB,MAAM,OAAOzB,MAAM0B,WAAW;IACnD,IAAI,OAAO1B,UAAU,WAAW,OAAOA,QAAQ,IAAI;IACnD,IAAI,OAAOA,UAAU,YAAY,OAAOA,UAAU,UAAU,OAAOA;IACnE,OAAO2B,KAAKC,SAAS,CAAC5B;AACxB;AAEA,OAAO,SAAS6B,UAAUC,IAAc;IACtC,MAAMC,MAAMxC,aAAauC,KAAKV,OAAO,EAAE;IACvC,MAAM,EAAEY,IAAI,EAAEC,OAAO,EAAE,GAAGtC,OAAOoC;IAEjC,MAAMG,WAAqB,EAAE;IAC7B,MAAMC,SAAiD,CAAC;IAExD,KAAK,MAAMC,OAAOC,OAAOC,IAAI,CAACN,MAAO;QACnC,IAAInC,iBAAiB0C,GAAG,CAACH,MAAM;YAC7BF,SAASM,IAAI,CAAC,CAAC,SAAS,EAAEV,KAAKX,OAAO,CAAC,8BAA8B,EAAEiB,IAAI,iCAAiC,CAAC;YAC7G;QACF;QACAD,MAAM,CAACC,IAAI,GAAGZ,SAASQ,IAAI,CAACI,IAAI;IAClC;IAEA,OAAO;QACLK,KAAK;YACHtB,SAASW,KAAKX,OAAO;YACrBG,SAASQ,KAAKR,OAAO;YACrBC,MAAMO,KAAKP,IAAI;YACfS,MAAMG;YACNO,QAAQ;gBACN,oEAAoE;gBACpE,0CAA0C;gBAC1CC,OAAO5C,cAAciC,KAAKW,KAAK;gBAC/BC,SAAS7C,cAAciC,KAAKY,OAAO;gBACnCC,MAAMxC,UAAU4B;YAClB;QACF;QACAC;IACF;AACF"}
package/dist/esm/watch.js CHANGED
@@ -2,19 +2,11 @@ import { watch as fsWatch } from 'node:fs';
2
2
  import { STATE_DIR } from './config.js';
3
3
  import { docCount, getMeta, open, reconcile, setMeta } from './db.js';
4
4
  import { SenseError } from './errors.js';
5
- // Watch is a cache pre-warmer, not a correctness mechanism: every query
6
- // `open()` reconciles against the filesystem regardless, so a missed or
7
- // coalesced fs event can never change a query result -- it only moves one
8
- // file's parse from ahead-of-time to query-time. That's why any event, of
9
- // any kind, triggers the exact same response: a debounced full reconcile.
10
- // Per-file event handling would be complexity with no correctness payoff.
5
+ // Watch is a cache pre-warmer, not a correctness mechanism: open() always reconciles anyway, so any fs event just triggers a debounced full reconcile.
11
6
  const DEBOUNCE_MS = 200;
12
7
  const HEARTBEAT_INTERVAL_MS = 5000;
13
8
  const STALE_HEARTBEAT_MS = 15000;
14
- // Runs in the foreground until SIGINT/SIGTERM; never forks or daemonizes --
15
- // process lifecycle belongs to the OS (launchd/systemd/terminal). Resolves
16
- // on clean shutdown; throws SenseError("WATCH_ACTIVE", ...) if another
17
- // watcher's heartbeat is still fresh and `--force` wasn't given.
9
+ // Runs in the foreground until SIGINT/SIGTERM. Throws WATCH_ACTIVE if another watcher's heartbeat is still fresh and --force wasn't given.
18
10
  export async function runWatch(cfg, opts = {}) {
19
11
  var _opts_onEvent;
20
12
  const onEvent = (_opts_onEvent = opts.onEvent) !== null && _opts_onEvent !== void 0 ? _opts_onEvent : ()=>{};
@@ -67,9 +59,7 @@ export async function runWatch(cfg, opts = {}) {
67
59
  }
68
60
  }, DEBOUNCE_MS);
69
61
  };
70
- // Ignore events from our own state dir: the heartbeat writes cache.db
71
- // every few seconds, and without this filter each write would schedule a
72
- // (harmless but pointless) reconcile forever — a perpetual self-trigger.
62
+ // Ignore our own state dir, or the heartbeat write would retrigger itself forever.
73
63
  const watcher = fsWatch(baseDir, {
74
64
  recursive: true
75
65
  }, (_event, filename)=>{
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/watch.ts"],"sourcesContent":["import { watch as fsWatch } from 'node:fs';\nimport type { ResolvedConfig } from './config.ts';\nimport { STATE_DIR } from './config.ts';\nimport { docCount, getMeta, open, reconcile, setMeta } from './db.ts';\nimport { SenseError } from './errors.ts';\n\n// Watch is a cache pre-warmer, not a correctness mechanism: every query\n// `open()` reconciles against the filesystem regardless, so a missed or\n// coalesced fs event can never change a query result -- it only moves one\n// file's parse from ahead-of-time to query-time. That's why any event, of\n// any kind, triggers the exact same response: a debounced full reconcile.\n// Per-file event handling would be complexity with no correctness payoff.\nconst DEBOUNCE_MS = 200;\nconst HEARTBEAT_INTERVAL_MS = 5000;\nconst STALE_HEARTBEAT_MS = 15000;\n\n// This module never prints -- it emits events via `onEvent` and throws on\n// the one exit-worthy condition (a live watcher already present). cli.ts is\n// the only place that touches console/process.exit.\nexport type WatchEvent = { type: 'started'; baseDir: string; dbPath: string } | { type: 'reconciled'; parsed: number; total: number; warnings: string[] } | { type: 'reconcile-error'; message: string };\n\nexport interface WatchOptions {\n force?: boolean;\n onEvent?: (event: WatchEvent) => void;\n}\n\n// Runs in the foreground until SIGINT/SIGTERM; never forks or daemonizes --\n// process lifecycle belongs to the OS (launchd/systemd/terminal). Resolves\n// on clean shutdown; throws SenseError(\"WATCH_ACTIVE\", ...) if another\n// watcher's heartbeat is still fresh and `--force` wasn't given.\nexport async function runWatch(cfg: ResolvedConfig, opts: WatchOptions = {}): Promise<void> {\n const onEvent = opts.onEvent ?? (() => {});\n const { db, dbPath, warnings: initialWarnings, parsed: initialParsed } = open(cfg);\n const baseDir = cfg.baseDir;\n\n const existingHeartbeat = getMeta(db, 'watch_heartbeat');\n if (existingHeartbeat && !opts.force) {\n const age = Date.now() - Date.parse(existingHeartbeat);\n if (age >= 0 && age < STALE_HEARTBEAT_MS) {\n db.close();\n throw new SenseError('WATCH_ACTIVE', `another watcher appears active (heartbeat ${Math.round(age / 1000)}s ago); use --force to override`);\n }\n }\n\n onEvent({ type: 'started', baseDir, dbPath });\n if (initialWarnings.length > 0 || initialParsed > 0) {\n onEvent({ type: 'reconciled', parsed: initialParsed, total: docCount(db), warnings: initialWarnings });\n }\n\n const touchHeartbeat = () => {\n setMeta(db, 'watch_heartbeat', new Date().toISOString());\n setMeta(db, 'watch_pid', String(process.pid));\n };\n touchHeartbeat();\n\n let debounceTimer: NodeJS.Timeout | null = null;\n const scheduleReconcile = () => {\n if (debounceTimer) clearTimeout(debounceTimer);\n debounceTimer = setTimeout(() => {\n debounceTimer = null;\n try {\n const { parsed, warnings } = reconcile(db, cfg, baseDir);\n onEvent({ type: 'reconciled', parsed, total: docCount(db), warnings });\n } catch (err) {\n onEvent({ type: 'reconcile-error', message: (err as Error).message });\n }\n }, DEBOUNCE_MS);\n };\n\n // Ignore events from our own state dir: the heartbeat writes cache.db\n // every few seconds, and without this filter each write would schedule a\n // (harmless but pointless) reconcile forever — a perpetual self-trigger.\n const watcher = fsWatch(baseDir, { recursive: true }, (_event, filename) => {\n if (typeof filename === 'string' && filename.startsWith(STATE_DIR)) return;\n scheduleReconcile();\n });\n const heartbeatTimer = setInterval(touchHeartbeat, HEARTBEAT_INTERVAL_MS);\n\n return new Promise<void>((resolveShutdown) => {\n const cleanExit = () => {\n clearInterval(heartbeatTimer);\n if (debounceTimer) clearTimeout(debounceTimer);\n watcher.close();\n setMeta(db, 'watch_heartbeat', null);\n setMeta(db, 'watch_pid', null);\n db.close();\n resolveShutdown();\n };\n process.once('SIGINT', cleanExit);\n process.once('SIGTERM', cleanExit);\n });\n}\n"],"names":["watch","fsWatch","STATE_DIR","docCount","getMeta","open","reconcile","setMeta","SenseError","DEBOUNCE_MS","HEARTBEAT_INTERVAL_MS","STALE_HEARTBEAT_MS","runWatch","cfg","opts","onEvent","db","dbPath","warnings","initialWarnings","parsed","initialParsed","baseDir","existingHeartbeat","force","age","Date","now","parse","close","Math","round","type","length","total","touchHeartbeat","toISOString","String","process","pid","debounceTimer","scheduleReconcile","clearTimeout","setTimeout","err","message","watcher","recursive","_event","filename","startsWith","heartbeatTimer","setInterval","Promise","resolveShutdown","cleanExit","clearInterval","once"],"mappings":"AAAA,SAASA,SAASC,OAAO,QAAQ,UAAU;AAE3C,SAASC,SAAS,QAAQ,cAAc;AACxC,SAASC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,SAAS,EAAEC,OAAO,QAAQ,UAAU;AACtE,SAASC,UAAU,QAAQ,cAAc;AAEzC,wEAAwE;AACxE,wEAAwE;AACxE,0EAA0E;AAC1E,0EAA0E;AAC1E,0EAA0E;AAC1E,0EAA0E;AAC1E,MAAMC,cAAc;AACpB,MAAMC,wBAAwB;AAC9B,MAAMC,qBAAqB;AAY3B,4EAA4E;AAC5E,2EAA2E;AAC3E,uEAAuE;AACvE,iEAAiE;AACjE,OAAO,eAAeC,SAASC,GAAmB,EAAEC,OAAqB,CAAC,CAAC;QACzDA;IAAhB,MAAMC,WAAUD,gBAAAA,KAAKC,OAAO,cAAZD,2BAAAA,gBAAiB,KAAO;IACxC,MAAM,EAAEE,EAAE,EAAEC,MAAM,EAAEC,UAAUC,eAAe,EAAEC,QAAQC,aAAa,EAAE,GAAGhB,KAAKQ;IAC9E,MAAMS,UAAUT,IAAIS,OAAO;IAE3B,MAAMC,oBAAoBnB,QAAQY,IAAI;IACtC,IAAIO,qBAAqB,CAACT,KAAKU,KAAK,EAAE;QACpC,MAAMC,MAAMC,KAAKC,GAAG,KAAKD,KAAKE,KAAK,CAACL;QACpC,IAAIE,OAAO,KAAKA,MAAMd,oBAAoB;YACxCK,GAAGa,KAAK;YACR,MAAM,IAAIrB,WAAW,gBAAgB,CAAC,0CAA0C,EAAEsB,KAAKC,KAAK,CAACN,MAAM,MAAM,+BAA+B,CAAC;QAC3I;IACF;IAEAV,QAAQ;QAAEiB,MAAM;QAAWV;QAASL;IAAO;IAC3C,IAAIE,gBAAgBc,MAAM,GAAG,KAAKZ,gBAAgB,GAAG;QACnDN,QAAQ;YAAEiB,MAAM;YAAcZ,QAAQC;YAAea,OAAO/B,SAASa;YAAKE,UAAUC;QAAgB;IACtG;IAEA,MAAMgB,iBAAiB;QACrB5B,QAAQS,IAAI,mBAAmB,IAAIU,OAAOU,WAAW;QACrD7B,QAAQS,IAAI,aAAaqB,OAAOC,QAAQC,GAAG;IAC7C;IACAJ;IAEA,IAAIK,gBAAuC;IAC3C,MAAMC,oBAAoB;QACxB,IAAID,eAAeE,aAAaF;QAChCA,gBAAgBG,WAAW;YACzBH,gBAAgB;YAChB,IAAI;gBACF,MAAM,EAAEpB,MAAM,EAAEF,QAAQ,EAAE,GAAGZ,UAAUU,IAAIH,KAAKS;gBAChDP,QAAQ;oBAAEiB,MAAM;oBAAcZ;oBAAQc,OAAO/B,SAASa;oBAAKE;gBAAS;YACtE,EAAE,OAAO0B,KAAK;gBACZ7B,QAAQ;oBAAEiB,MAAM;oBAAmBa,SAAS,AAACD,IAAcC,OAAO;gBAAC;YACrE;QACF,GAAGpC;IACL;IAEA,sEAAsE;IACtE,yEAAyE;IACzE,yEAAyE;IACzE,MAAMqC,UAAU7C,QAAQqB,SAAS;QAAEyB,WAAW;IAAK,GAAG,CAACC,QAAQC;QAC7D,IAAI,OAAOA,aAAa,YAAYA,SAASC,UAAU,CAAChD,YAAY;QACpEuC;IACF;IACA,MAAMU,iBAAiBC,YAAYjB,gBAAgBzB;IAEnD,OAAO,IAAI2C,QAAc,CAACC;QACxB,MAAMC,YAAY;YAChBC,cAAcL;YACd,IAAIX,eAAeE,aAAaF;YAChCM,QAAQjB,KAAK;YACbtB,QAAQS,IAAI,mBAAmB;YAC/BT,QAAQS,IAAI,aAAa;YACzBA,GAAGa,KAAK;YACRyB;QACF;QACAhB,QAAQmB,IAAI,CAAC,UAAUF;QACvBjB,QAAQmB,IAAI,CAAC,WAAWF;IAC1B;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/watch.ts"],"sourcesContent":["import { watch as fsWatch } from 'node:fs';\nimport type { ResolvedConfig } from './config.ts';\nimport { STATE_DIR } from './config.ts';\nimport { docCount, getMeta, open, reconcile, setMeta } from './db.ts';\nimport { SenseError } from './errors.ts';\n\n// Watch is a cache pre-warmer, not a correctness mechanism: open() always reconciles anyway, so any fs event just triggers a debounced full reconcile.\nconst DEBOUNCE_MS = 200;\nconst HEARTBEAT_INTERVAL_MS = 5000;\nconst STALE_HEARTBEAT_MS = 15000;\n\nexport type WatchEvent = { type: 'started'; baseDir: string; dbPath: string } | { type: 'reconciled'; parsed: number; total: number; warnings: string[] } | { type: 'reconcile-error'; message: string };\n\nexport interface WatchOptions {\n force?: boolean;\n onEvent?: (event: WatchEvent) => void;\n}\n\n// Runs in the foreground until SIGINT/SIGTERM. Throws WATCH_ACTIVE if another watcher's heartbeat is still fresh and --force wasn't given.\nexport async function runWatch(cfg: ResolvedConfig, opts: WatchOptions = {}): Promise<void> {\n const onEvent = opts.onEvent ?? (() => {});\n const { db, dbPath, warnings: initialWarnings, parsed: initialParsed } = open(cfg);\n const baseDir = cfg.baseDir;\n\n const existingHeartbeat = getMeta(db, 'watch_heartbeat');\n if (existingHeartbeat && !opts.force) {\n const age = Date.now() - Date.parse(existingHeartbeat);\n if (age >= 0 && age < STALE_HEARTBEAT_MS) {\n db.close();\n throw new SenseError('WATCH_ACTIVE', `another watcher appears active (heartbeat ${Math.round(age / 1000)}s ago); use --force to override`);\n }\n }\n\n onEvent({ type: 'started', baseDir, dbPath });\n if (initialWarnings.length > 0 || initialParsed > 0) {\n onEvent({ type: 'reconciled', parsed: initialParsed, total: docCount(db), warnings: initialWarnings });\n }\n\n const touchHeartbeat = () => {\n setMeta(db, 'watch_heartbeat', new Date().toISOString());\n setMeta(db, 'watch_pid', String(process.pid));\n };\n touchHeartbeat();\n\n let debounceTimer: NodeJS.Timeout | null = null;\n const scheduleReconcile = () => {\n if (debounceTimer) clearTimeout(debounceTimer);\n debounceTimer = setTimeout(() => {\n debounceTimer = null;\n try {\n const { parsed, warnings } = reconcile(db, cfg, baseDir);\n onEvent({ type: 'reconciled', parsed, total: docCount(db), warnings });\n } catch (err) {\n onEvent({ type: 'reconcile-error', message: (err as Error).message });\n }\n }, DEBOUNCE_MS);\n };\n\n // Ignore our own state dir, or the heartbeat write would retrigger itself forever.\n const watcher = fsWatch(baseDir, { recursive: true }, (_event, filename) => {\n if (typeof filename === 'string' && filename.startsWith(STATE_DIR)) return;\n scheduleReconcile();\n });\n const heartbeatTimer = setInterval(touchHeartbeat, HEARTBEAT_INTERVAL_MS);\n\n return new Promise<void>((resolveShutdown) => {\n const cleanExit = () => {\n clearInterval(heartbeatTimer);\n if (debounceTimer) clearTimeout(debounceTimer);\n watcher.close();\n setMeta(db, 'watch_heartbeat', null);\n setMeta(db, 'watch_pid', null);\n db.close();\n resolveShutdown();\n };\n process.once('SIGINT', cleanExit);\n process.once('SIGTERM', cleanExit);\n });\n}\n"],"names":["watch","fsWatch","STATE_DIR","docCount","getMeta","open","reconcile","setMeta","SenseError","DEBOUNCE_MS","HEARTBEAT_INTERVAL_MS","STALE_HEARTBEAT_MS","runWatch","cfg","opts","onEvent","db","dbPath","warnings","initialWarnings","parsed","initialParsed","baseDir","existingHeartbeat","force","age","Date","now","parse","close","Math","round","type","length","total","touchHeartbeat","toISOString","String","process","pid","debounceTimer","scheduleReconcile","clearTimeout","setTimeout","err","message","watcher","recursive","_event","filename","startsWith","heartbeatTimer","setInterval","Promise","resolveShutdown","cleanExit","clearInterval","once"],"mappings":"AAAA,SAASA,SAASC,OAAO,QAAQ,UAAU;AAE3C,SAASC,SAAS,QAAQ,cAAc;AACxC,SAASC,QAAQ,EAAEC,OAAO,EAAEC,IAAI,EAAEC,SAAS,EAAEC,OAAO,QAAQ,UAAU;AACtE,SAASC,UAAU,QAAQ,cAAc;AAEzC,uJAAuJ;AACvJ,MAAMC,cAAc;AACpB,MAAMC,wBAAwB;AAC9B,MAAMC,qBAAqB;AAS3B,2IAA2I;AAC3I,OAAO,eAAeC,SAASC,GAAmB,EAAEC,OAAqB,CAAC,CAAC;QACzDA;IAAhB,MAAMC,WAAUD,gBAAAA,KAAKC,OAAO,cAAZD,2BAAAA,gBAAiB,KAAO;IACxC,MAAM,EAAEE,EAAE,EAAEC,MAAM,EAAEC,UAAUC,eAAe,EAAEC,QAAQC,aAAa,EAAE,GAAGhB,KAAKQ;IAC9E,MAAMS,UAAUT,IAAIS,OAAO;IAE3B,MAAMC,oBAAoBnB,QAAQY,IAAI;IACtC,IAAIO,qBAAqB,CAACT,KAAKU,KAAK,EAAE;QACpC,MAAMC,MAAMC,KAAKC,GAAG,KAAKD,KAAKE,KAAK,CAACL;QACpC,IAAIE,OAAO,KAAKA,MAAMd,oBAAoB;YACxCK,GAAGa,KAAK;YACR,MAAM,IAAIrB,WAAW,gBAAgB,CAAC,0CAA0C,EAAEsB,KAAKC,KAAK,CAACN,MAAM,MAAM,+BAA+B,CAAC;QAC3I;IACF;IAEAV,QAAQ;QAAEiB,MAAM;QAAWV;QAASL;IAAO;IAC3C,IAAIE,gBAAgBc,MAAM,GAAG,KAAKZ,gBAAgB,GAAG;QACnDN,QAAQ;YAAEiB,MAAM;YAAcZ,QAAQC;YAAea,OAAO/B,SAASa;YAAKE,UAAUC;QAAgB;IACtG;IAEA,MAAMgB,iBAAiB;QACrB5B,QAAQS,IAAI,mBAAmB,IAAIU,OAAOU,WAAW;QACrD7B,QAAQS,IAAI,aAAaqB,OAAOC,QAAQC,GAAG;IAC7C;IACAJ;IAEA,IAAIK,gBAAuC;IAC3C,MAAMC,oBAAoB;QACxB,IAAID,eAAeE,aAAaF;QAChCA,gBAAgBG,WAAW;YACzBH,gBAAgB;YAChB,IAAI;gBACF,MAAM,EAAEpB,MAAM,EAAEF,QAAQ,EAAE,GAAGZ,UAAUU,IAAIH,KAAKS;gBAChDP,QAAQ;oBAAEiB,MAAM;oBAAcZ;oBAAQc,OAAO/B,SAASa;oBAAKE;gBAAS;YACtE,EAAE,OAAO0B,KAAK;gBACZ7B,QAAQ;oBAAEiB,MAAM;oBAAmBa,SAAS,AAACD,IAAcC,OAAO;gBAAC;YACrE;QACF,GAAGpC;IACL;IAEA,mFAAmF;IACnF,MAAMqC,UAAU7C,QAAQqB,SAAS;QAAEyB,WAAW;IAAK,GAAG,CAACC,QAAQC;QAC7D,IAAI,OAAOA,aAAa,YAAYA,SAASC,UAAU,CAAChD,YAAY;QACpEuC;IACF;IACA,MAAMU,iBAAiBC,YAAYjB,gBAAgBzB;IAEnD,OAAO,IAAI2C,QAAc,CAACC;QACxB,MAAMC,YAAY;YAChBC,cAAcL;YACd,IAAIX,eAAeE,aAAaF;YAChCM,QAAQjB,KAAK;YACbtB,QAAQS,IAAI,mBAAmB;YAC/BT,QAAQS,IAAI,aAAa;YACzBA,GAAGa,KAAK;YACRyB;QACF;QACAhB,QAAQmB,IAAI,CAAC,UAAUF;QACvBjB,QAAQmB,IAAI,CAAC,WAAWF;IAC1B;AACF"}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "sensemaking",
3
- "version": "0.1.0",
3
+ "version": "0.2.0",
4
4
  "description": "Query a knowledge base you build with an agent: filter notes by frontmatter, then search inside them",
5
5
  "keywords": [
6
6
  "markdown",
@@ -48,7 +48,8 @@
48
48
  },
49
49
  "dependencies": {
50
50
  "fast-glob": "^3.3.3",
51
- "gray-matter": "^4.0.3"
51
+ "gray-matter": "^4.0.3",
52
+ "remove-markdown": "^0.6.4"
52
53
  },
53
54
  "devDependencies": {
54
55
  "@types/mocha": "*",
package/schema.json CHANGED
@@ -19,7 +19,7 @@
19
19
  "type": "object",
20
20
  "additionalProperties": false,
21
21
  "required": ["include"],
22
- "description": "Which files become rows in the `docs` table.",
22
+ "description": "Which files become rows in the `frontmatter` table.",
23
23
  "properties": {
24
24
  "include": {
25
25
  "type": "array",
@@ -31,7 +31,7 @@
31
31
  },
32
32
  "queries": {
33
33
  "type": "object",
34
- "description": "Named SQL queries runnable as `sense <name> [params...]`. Each value is a SQL SELECT against the `docs` table (one row per file, one column per discovered frontmatter key, plus reserved `path`/`_mtime`/`_size`). `?` placeholders bind to CLI positional arguments in order. The custom `has(field, value)` function does array-membership on a JSON-array field, substring match on a string field, and is always false on NULL. The names `init`, `query`, `watch`, `status`, and `rebuild` are reserved subcommands — a query with one of those names is unreachable from the CLI.",
34
+ "description": "Named SQL queries runnable as `sense <name> [params...]`. Tables: `frontmatter` (one row per file, one column per discovered frontmatter key, plus `path`/`_mtime`/`_size`) and `content`, an FTS5 index (`title`, `summary`, `text`, `path`) for content search. `?` placeholders bind to CLI positional args in order. `has(field, value)`: array membership on a JSON-array field, substring match on a string, false on NULL. Canonical query: `SELECT f.path, content.title, content.summary, snippet(content, -1, '«', '»', '…', 10) AS hit FROM frontmatter f JOIN content ON content.path = f.path WHERE content MATCH ? ORDER BY bm25(content, 10.0, 5.0, 1.0) LIMIT 10`. Reserved frontmatter keys: `path`, `_mtime`, `_size`, `content`. Reserved query names (unreachable as subcommands): `init`, `query`, `watch`, `status`, `rebuild`.",
35
35
  "additionalProperties": { "type": "string" }
36
36
  }
37
37
  }