@context-vault/core 2.10.2 → 2.11.0

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/package.json CHANGED
@@ -1,31 +1,22 @@
1
1
  {
2
2
  "name": "@context-vault/core",
3
- "version": "2.10.2",
3
+ "version": "2.11.0",
4
4
  "type": "module",
5
5
  "description": "Shared core: capture, index, retrieve, tools, and utilities for context-vault",
6
6
  "main": "src/index.js",
7
7
  "exports": {
8
8
  ".": "./src/index.js",
9
+ "./constants": "./src/constants.js",
9
10
  "./capture": "./src/capture/index.js",
10
- "./capture/formatters": "./src/capture/formatters.js",
11
- "./capture/file-ops": "./src/capture/file-ops.js",
12
- "./index/db": "./src/index/db.js",
13
- "./index/embed": "./src/index/embed.js",
11
+ "./capture/*": "./src/capture/*.js",
14
12
  "./index": "./src/index/index.js",
13
+ "./index/*": "./src/index/*.js",
15
14
  "./retrieve": "./src/retrieve/index.js",
16
- "./server/tools": "./src/server/tools.js",
17
- "./server/helpers": "./src/server/helpers.js",
18
- "./core/categories": "./src/core/categories.js",
19
- "./core/config": "./src/core/config.js",
20
- "./core/files": "./src/core/files.js",
21
- "./core/frontmatter": "./src/core/frontmatter.js",
22
- "./core/error-log": "./src/core/error-log.js",
23
- "./core/status": "./src/core/status.js",
24
- "./capture/importers": "./src/capture/importers.js",
25
- "./capture/import-pipeline": "./src/capture/import-pipeline.js",
26
- "./capture/ingest-url": "./src/capture/ingest-url.js",
15
+ "./retrieve/*": "./src/retrieve/*.js",
27
16
  "./sync": "./src/sync/sync.js",
28
- "./constants": "./src/constants.js"
17
+ "./sync/*": "./src/sync/*.js",
18
+ "./core/*": "./src/core/*.js",
19
+ "./server/*": "./src/server/*.js"
29
20
  },
30
21
  "files": [
31
22
  "src/"
@@ -32,10 +32,12 @@ export function writeEntryFile(
32
32
  tags,
33
33
  source,
34
34
  createdAt,
35
+ updatedAt,
35
36
  folder,
36
37
  category,
37
38
  identity_key,
38
39
  expires_at,
40
+ supersedes,
39
41
  },
40
42
  ) {
41
43
  // P5: folder is now a top-level param; also accept from meta for backward compat
@@ -61,9 +63,11 @@ export function writeEntryFile(
61
63
 
62
64
  if (identity_key) fmFields.identity_key = identity_key;
63
65
  if (expires_at) fmFields.expires_at = expires_at;
66
+ if (supersedes?.length) fmFields.supersedes = supersedes;
64
67
  fmFields.tags = tags || [];
65
68
  fmFields.source = source || "claude-code";
66
69
  fmFields.created = created;
70
+ if (updatedAt && updatedAt !== created) fmFields.updated = updatedAt;
67
71
 
68
72
  const mdBody = formatBody(kind, { title, body, meta });
69
73
 
@@ -26,6 +26,7 @@ export function writeEntry(
26
26
  folder,
27
27
  identity_key,
28
28
  expires_at,
29
+ supersedes,
29
30
  userId,
30
31
  },
31
32
  ) {
@@ -47,6 +48,7 @@ export function writeEntry(
47
48
  // Entity upsert: check for existing file at deterministic path
48
49
  let id;
49
50
  let createdAt;
51
+ let updatedAt;
50
52
  if (category === "entity" && identity_key) {
51
53
  const identitySlug = slugify(identity_key);
52
54
  const dir = resolve(ctx.config.vaultDir, kindToPath(kind));
@@ -58,13 +60,16 @@ export function writeEntry(
58
60
  const { meta: fmMeta } = parseFrontmatter(raw);
59
61
  id = fmMeta.id || ulid();
60
62
  createdAt = fmMeta.created || new Date().toISOString();
63
+ updatedAt = new Date().toISOString();
61
64
  } else {
62
65
  id = ulid();
63
66
  createdAt = new Date().toISOString();
67
+ updatedAt = createdAt;
64
68
  }
65
69
  } else {
66
70
  id = ulid();
67
71
  createdAt = new Date().toISOString();
72
+ updatedAt = createdAt;
68
73
  }
69
74
 
70
75
  const filePath = writeEntryFile(ctx.config.vaultDir, kind, {
@@ -75,10 +80,12 @@ export function writeEntry(
75
80
  tags,
76
81
  source,
77
82
  createdAt,
83
+ updatedAt,
78
84
  folder,
79
85
  category,
80
86
  identity_key,
81
87
  expires_at,
88
+ supersedes,
82
89
  });
83
90
 
84
91
  return {
@@ -92,8 +99,10 @@ export function writeEntry(
92
99
  tags,
93
100
  source,
94
101
  createdAt,
102
+ updatedAt,
95
103
  identity_key,
96
104
  expires_at,
105
+ supersedes,
97
106
  userId: userId || null,
98
107
  };
99
108
  }
@@ -121,6 +130,10 @@ export function updateEntryFile(ctx, existing, updates) {
121
130
  updates.source !== undefined ? updates.source : existing.source;
122
131
  const expires_at =
123
132
  updates.expires_at !== undefined ? updates.expires_at : existing.expires_at;
133
+ const supersedes =
134
+ updates.supersedes !== undefined
135
+ ? updates.supersedes
136
+ : fmMeta.supersedes || null;
124
137
 
125
138
  let mergedMeta;
126
139
  if (updates.meta !== undefined) {
@@ -130,6 +143,7 @@ export function updateEntryFile(ctx, existing, updates) {
130
143
  }
131
144
 
132
145
  // Build frontmatter
146
+ const now = new Date().toISOString();
133
147
  const fmFields = { id: existing.id };
134
148
  for (const [k, v] of Object.entries(mergedMeta)) {
135
149
  if (k === "folder") continue;
@@ -137,9 +151,11 @@ export function updateEntryFile(ctx, existing, updates) {
137
151
  }
138
152
  if (existing.identity_key) fmFields.identity_key = existing.identity_key;
139
153
  if (expires_at) fmFields.expires_at = expires_at;
154
+ if (supersedes?.length) fmFields.supersedes = supersedes;
140
155
  fmFields.tags = tags;
141
156
  fmFields.source = source || "claude-code";
142
157
  fmFields.created = fmMeta.created || existing.created_at;
158
+ if (now !== fmFields.created) fmFields.updated = now;
143
159
 
144
160
  const mdBody = formatBody(existing.kind, { title, body, meta: mergedMeta });
145
161
  const md = formatFrontmatter(fmFields) + mdBody;
@@ -159,8 +175,10 @@ export function updateEntryFile(ctx, existing, updates) {
159
175
  tags,
160
176
  source,
161
177
  createdAt: fmMeta.created || existing.created_at,
178
+ updatedAt: now,
162
179
  identity_key: existing.identity_key,
163
180
  expires_at,
181
+ supersedes,
164
182
  userId: existing.user_id || null,
165
183
  };
166
184
  }
@@ -180,6 +198,14 @@ export async function captureAndIndex(ctx, data) {
180
198
  const entry = writeEntry(ctx, data);
181
199
  try {
182
200
  await indexEntry(ctx, entry);
201
+ // Apply supersedes: mark referenced entries as superseded by this entry
202
+ if (entry.supersedes?.length && ctx.stmts.updateSupersededBy) {
203
+ for (const supersededId of entry.supersedes) {
204
+ if (typeof supersededId === "string" && supersededId.trim()) {
205
+ ctx.stmts.updateSupersededBy.run(entry.id, supersededId.trim());
206
+ }
207
+ }
208
+ }
183
209
  return entry;
184
210
  } catch (err) {
185
211
  // Rollback: restore previous content for entity upserts, delete for new entries
package/src/constants.js CHANGED
@@ -6,3 +6,10 @@ export const MAX_TAGS_COUNT = 20;
6
6
  export const MAX_META_LENGTH = 10 * 1024; // 10KB
7
7
  export const MAX_SOURCE_LENGTH = 200;
8
8
  export const MAX_IDENTITY_KEY_LENGTH = 200;
9
+
10
+ export const DEFAULT_GROWTH_THRESHOLDS = {
11
+ totalEntries: { warn: 1000, critical: 5000 },
12
+ eventEntries: { warn: 500, critical: 2000 },
13
+ vaultSizeBytes: { warn: 50 * 1024 * 1024, critical: 200 * 1024 * 1024 },
14
+ eventsWithoutTtl: { warn: 200 },
15
+ };
@@ -40,6 +40,17 @@ const CATEGORY_DIR_NAMES = {
40
40
  /** Set of valid category directory names (for reindex discovery) */
41
41
  export const CATEGORY_DIRS = new Set(Object.values(CATEGORY_DIR_NAMES));
42
42
 
43
+ /**
44
+ * Staleness thresholds (in days) per knowledge kind.
45
+ * Kinds not listed here are considered enduring (no staleness threshold).
46
+ * Based on updated_at; falls back to created_at if updated_at is null.
47
+ */
48
+ export const KIND_STALENESS_DAYS = {
49
+ pattern: 180,
50
+ decision: 365,
51
+ reference: 90,
52
+ };
53
+
43
54
  export function categoryFor(kind) {
44
55
  return KIND_CATEGORY[kind] || "knowledge";
45
56
  }
@@ -1,6 +1,7 @@
1
1
  import { existsSync, readFileSync } from "node:fs";
2
2
  import { join, resolve } from "node:path";
3
3
  import { homedir } from "node:os";
4
+ import { DEFAULT_GROWTH_THRESHOLDS } from "../constants.js";
4
5
 
5
6
  export function parseArgs(argv) {
6
7
  const args = {};
@@ -31,6 +32,8 @@ export function resolveConfig() {
31
32
  dbPath: join(dataDir, "vault.db"),
32
33
  devDir: join(HOME, "dev"),
33
34
  eventDecayDays: 30,
35
+ thresholds: { ...DEFAULT_GROWTH_THRESHOLDS },
36
+ telemetry: false,
34
37
  resolvedFrom: "defaults",
35
38
  };
36
39
 
@@ -46,6 +49,30 @@ export function resolveConfig() {
46
49
  if (fc.dbPath) config.dbPath = fc.dbPath;
47
50
  if (fc.devDir) config.devDir = fc.devDir;
48
51
  if (fc.eventDecayDays != null) config.eventDecayDays = fc.eventDecayDays;
52
+ if (fc.thresholds) {
53
+ const t = fc.thresholds;
54
+ if (t.totalEntries)
55
+ config.thresholds.totalEntries = {
56
+ ...config.thresholds.totalEntries,
57
+ ...t.totalEntries,
58
+ };
59
+ if (t.eventEntries)
60
+ config.thresholds.eventEntries = {
61
+ ...config.thresholds.eventEntries,
62
+ ...t.eventEntries,
63
+ };
64
+ if (t.vaultSizeBytes)
65
+ config.thresholds.vaultSizeBytes = {
66
+ ...config.thresholds.vaultSizeBytes,
67
+ ...t.vaultSizeBytes,
68
+ };
69
+ if (t.eventsWithoutTtl)
70
+ config.thresholds.eventsWithoutTtl = {
71
+ ...config.thresholds.eventsWithoutTtl,
72
+ ...t.eventsWithoutTtl,
73
+ };
74
+ }
75
+ if (fc.telemetry != null) config.telemetry = fc.telemetry === true;
49
76
  // Hosted account linking (Phase 4)
50
77
  if (fc.hostedUrl) config.hostedUrl = fc.hostedUrl;
51
78
  if (fc.apiKey) config.apiKey = fc.apiKey;
@@ -96,6 +123,11 @@ export function resolveConfig() {
96
123
  if (process.env.CONTEXT_VAULT_HOSTED_URL) {
97
124
  config.hostedUrl = process.env.CONTEXT_VAULT_HOSTED_URL;
98
125
  }
126
+ if (process.env.CONTEXT_VAULT_TELEMETRY !== undefined) {
127
+ config.telemetry =
128
+ process.env.CONTEXT_VAULT_TELEMETRY === "1" ||
129
+ process.env.CONTEXT_VAULT_TELEMETRY === "true";
130
+ }
99
131
 
100
132
  if (cliArgs.vaultDir) {
101
133
  config.vaultDir = cliArgs.vaultDir;
@@ -65,8 +65,10 @@ const RESERVED_FM_KEYS = new Set([
65
65
  "tags",
66
66
  "source",
67
67
  "created",
68
+ "updated",
68
69
  "identity_key",
69
70
  "expires_at",
71
+ "supersedes",
70
72
  ]);
71
73
 
72
74
  export function extractCustomMeta(fmMeta) {
@@ -6,6 +6,7 @@ import { existsSync, readdirSync, statSync } from "node:fs";
6
6
  import { join } from "node:path";
7
7
  import { walkDir } from "./files.js";
8
8
  import { isEmbedAvailable } from "../index/embed.js";
9
+ import { KIND_STALENESS_DAYS } from "./categories.js";
9
10
 
10
11
  /**
11
12
  * Gather raw vault status data for formatting by consumers.
@@ -109,6 +110,30 @@ export function gatherVaultStatus(ctx, opts = {}) {
109
110
  errors.push(`Expired count failed: ${e.message}`);
110
111
  }
111
112
 
113
+ // Count event-category entries
114
+ let eventCount = 0;
115
+ try {
116
+ eventCount = db
117
+ .prepare(
118
+ `SELECT COUNT(*) as c FROM vault WHERE category = 'event' ${userAnd}`,
119
+ )
120
+ .get(...userParams).c;
121
+ } catch (e) {
122
+ errors.push(`Event count failed: ${e.message}`);
123
+ }
124
+
125
+ // Count event entries without expires_at
126
+ let eventsWithoutTtlCount = 0;
127
+ try {
128
+ eventsWithoutTtlCount = db
129
+ .prepare(
130
+ `SELECT COUNT(*) as c FROM vault WHERE category = 'event' AND expires_at IS NULL ${userAnd}`,
131
+ )
132
+ .get(...userParams).c;
133
+ } catch (e) {
134
+ errors.push(`Events without TTL count failed: ${e.message}`);
135
+ }
136
+
112
137
  // Embedding/vector status
113
138
  let embeddingStatus = null;
114
139
  try {
@@ -140,6 +165,27 @@ export function gatherVaultStatus(ctx, opts = {}) {
140
165
  errors.push(`Auto-captured feedback count failed: ${e.message}`);
141
166
  }
142
167
 
168
+ // Stale knowledge entries — kinds with a threshold, not updated within N days
169
+ let staleKnowledge = [];
170
+ try {
171
+ const stalenessKinds = Object.entries(KIND_STALENESS_DAYS);
172
+ if (stalenessKinds.length > 0) {
173
+ const kindClauses = stalenessKinds
174
+ .map(
175
+ ([kind, days]) =>
176
+ `(kind = '${kind}' AND COALESCE(updated_at, created_at) <= datetime('now', '-${days} days'))`,
177
+ )
178
+ .join(" OR ");
179
+ staleKnowledge = db
180
+ .prepare(
181
+ `SELECT kind, title, COALESCE(updated_at, created_at) as last_updated FROM vault WHERE category = 'knowledge' AND (${kindClauses}) AND (expires_at IS NULL OR expires_at > datetime('now')) ${userAnd} ORDER BY last_updated ASC LIMIT 10`,
182
+ )
183
+ .all(...userParams);
184
+ }
185
+ } catch (e) {
186
+ errors.push(`Stale knowledge check failed: ${e.message}`);
187
+ }
188
+
143
189
  return {
144
190
  fileCount,
145
191
  subdirs,
@@ -150,10 +196,129 @@ export function gatherVaultStatus(ctx, opts = {}) {
150
196
  stalePaths,
151
197
  staleCount,
152
198
  expiredCount,
199
+ eventCount,
200
+ eventsWithoutTtlCount,
153
201
  embeddingStatus,
154
202
  embedModelAvailable,
155
203
  autoCapturedFeedbackCount,
204
+ staleKnowledge,
156
205
  resolvedFrom: config.resolvedFrom,
157
206
  errors,
158
207
  };
159
208
  }
209
+
210
+ /**
211
+ * Compute growth warnings based on vault status and configured thresholds.
212
+ *
213
+ * @param {object} status — result of gatherVaultStatus()
214
+ * @param {object} thresholds — from config.thresholds
215
+ * @returns {{ warnings: Array, hasCritical: boolean, hasWarnings: boolean, actions: string[] }}
216
+ */
217
+ export function computeGrowthWarnings(status, thresholds) {
218
+ if (!thresholds)
219
+ return {
220
+ warnings: [],
221
+ hasCritical: false,
222
+ hasWarnings: false,
223
+ actions: [],
224
+ };
225
+
226
+ const t = thresholds;
227
+ const warnings = [];
228
+ const actions = [];
229
+
230
+ const total = status.embeddingStatus?.total ?? 0;
231
+ const {
232
+ eventCount = 0,
233
+ eventsWithoutTtlCount = 0,
234
+ expiredCount = 0,
235
+ dbSizeBytes = 0,
236
+ } = status;
237
+
238
+ if (t.totalEntries?.critical != null && total >= t.totalEntries.critical) {
239
+ warnings.push({
240
+ level: "critical",
241
+ message: `Total entries: ${total.toLocaleString()} (exceeds critical limit of ${t.totalEntries.critical.toLocaleString()})`,
242
+ });
243
+ } else if (t.totalEntries?.warn != null && total >= t.totalEntries.warn) {
244
+ warnings.push({
245
+ level: "warn",
246
+ message: `Total entries: ${total.toLocaleString()} (exceeds recommended ${t.totalEntries.warn.toLocaleString()})`,
247
+ });
248
+ }
249
+
250
+ if (
251
+ t.eventEntries?.critical != null &&
252
+ eventCount >= t.eventEntries.critical
253
+ ) {
254
+ warnings.push({
255
+ level: "critical",
256
+ message: `Event entries: ${eventCount.toLocaleString()} (exceeds critical limit of ${t.eventEntries.critical.toLocaleString()})`,
257
+ });
258
+ } else if (
259
+ t.eventEntries?.warn != null &&
260
+ eventCount >= t.eventEntries.warn
261
+ ) {
262
+ const ttlNote =
263
+ eventsWithoutTtlCount > 0
264
+ ? ` (${eventsWithoutTtlCount.toLocaleString()} without TTL)`
265
+ : "";
266
+ warnings.push({
267
+ level: "warn",
268
+ message: `Event entries: ${eventCount.toLocaleString()}${ttlNote} (exceeds recommended ${t.eventEntries.warn.toLocaleString()})`,
269
+ });
270
+ }
271
+
272
+ if (
273
+ t.vaultSizeBytes?.critical != null &&
274
+ dbSizeBytes >= t.vaultSizeBytes.critical
275
+ ) {
276
+ warnings.push({
277
+ level: "critical",
278
+ message: `Database size: ${(dbSizeBytes / 1024 / 1024).toFixed(1)}MB (exceeds critical limit of ${(t.vaultSizeBytes.critical / 1024 / 1024).toFixed(0)}MB)`,
279
+ });
280
+ } else if (
281
+ t.vaultSizeBytes?.warn != null &&
282
+ dbSizeBytes >= t.vaultSizeBytes.warn
283
+ ) {
284
+ warnings.push({
285
+ level: "warn",
286
+ message: `Database size: ${(dbSizeBytes / 1024 / 1024).toFixed(1)}MB (exceeds recommended ${(t.vaultSizeBytes.warn / 1024 / 1024).toFixed(0)}MB)`,
287
+ });
288
+ }
289
+
290
+ if (
291
+ t.eventsWithoutTtl?.warn != null &&
292
+ eventsWithoutTtlCount >= t.eventsWithoutTtl.warn
293
+ ) {
294
+ warnings.push({
295
+ level: "warn",
296
+ message: `Event entries without expires_at: ${eventsWithoutTtlCount.toLocaleString()} (exceeds recommended ${t.eventsWithoutTtl.warn.toLocaleString()})`,
297
+ });
298
+ }
299
+
300
+ const hasCritical = warnings.some((w) => w.level === "critical");
301
+
302
+ if (expiredCount > 0) {
303
+ actions.push(
304
+ `Run \`context-vault prune\` to remove ${expiredCount} expired event entr${expiredCount === 1 ? "y" : "ies"}`,
305
+ );
306
+ }
307
+ const eventThresholdExceeded =
308
+ eventCount >= (t.eventEntries?.warn ?? Infinity);
309
+ const ttlThresholdExceeded =
310
+ eventsWithoutTtlCount >= (t.eventsWithoutTtl?.warn ?? Infinity);
311
+ if (
312
+ eventsWithoutTtlCount > 0 &&
313
+ (eventThresholdExceeded || ttlThresholdExceeded)
314
+ ) {
315
+ actions.push(
316
+ "Add `expires_at` to event/session entries to enable automatic cleanup",
317
+ );
318
+ }
319
+ if (total >= (t.totalEntries?.warn ?? Infinity)) {
320
+ actions.push("Consider archiving events older than 90 days");
321
+ }
322
+
323
+ return { warnings, hasCritical, hasWarnings: warnings.length > 0, actions };
324
+ }
@@ -0,0 +1,64 @@
1
+ import { existsSync, writeFileSync } from "node:fs";
2
+ import { join } from "node:path";
3
+
4
+ const TELEMETRY_ENDPOINT = "https://api.context-vault.com/telemetry";
5
+ const NOTICE_MARKER = ".telemetry-notice-shown";
6
+
7
+ export function isTelemetryEnabled(config) {
8
+ const envVal = process.env.CONTEXT_VAULT_TELEMETRY;
9
+ if (envVal !== undefined) return envVal === "1" || envVal === "true";
10
+ return config?.telemetry === true;
11
+ }
12
+
13
+ /**
14
+ * Fire-and-forget telemetry event. Never throws, never blocks.
15
+ * Payload contains only: event, code, tool, cv_version, node_version, platform, arch, ts.
16
+ * No message text, stack traces, vault content, file paths, or user identifiers.
17
+ */
18
+ export function sendTelemetryEvent(config, payload) {
19
+ if (!isTelemetryEnabled(config)) return;
20
+
21
+ const event = {
22
+ event: payload.event,
23
+ code: payload.code || null,
24
+ tool: payload.tool || null,
25
+ cv_version: payload.cv_version,
26
+ node_version: process.version,
27
+ platform: process.platform,
28
+ arch: process.arch,
29
+ ts: new Date().toISOString(),
30
+ };
31
+
32
+ fetch(TELEMETRY_ENDPOINT, {
33
+ method: "POST",
34
+ headers: { "Content-Type": "application/json" },
35
+ body: JSON.stringify(event),
36
+ signal: AbortSignal.timeout(5000),
37
+ }).catch(() => {});
38
+ }
39
+
40
+ /**
41
+ * Print the one-time telemetry notice to stderr.
42
+ * Uses a marker file in dataDir to ensure it's only shown once.
43
+ */
44
+ export function maybeShowTelemetryNotice(dataDir) {
45
+ try {
46
+ const markerPath = join(dataDir, NOTICE_MARKER);
47
+ if (existsSync(markerPath)) return;
48
+ writeFileSync(markerPath, new Date().toISOString() + "\n");
49
+ } catch {
50
+ return;
51
+ }
52
+
53
+ const lines = [
54
+ "[context-vault] Telemetry: disabled by default.",
55
+ "[context-vault] To help improve context-vault, you can opt in to anonymous error reporting.",
56
+ "[context-vault] Reports contain only: event type, error code, tool name, version, node version, platform, arch, timestamp.",
57
+ "[context-vault] No vault content, file paths, or personal data is ever sent.",
58
+ '[context-vault] Opt in: set "telemetry": true in ~/.context-mcp/config.json or set CONTEXT_VAULT_TELEMETRY=1.',
59
+ "[context-vault] Full payload schema: https://contextvault.dev/telemetry",
60
+ ];
61
+ for (const line of lines) {
62
+ process.stderr.write(line + "\n");
63
+ }
64
+ }
package/src/index/db.js CHANGED
@@ -55,7 +55,9 @@ export const SCHEMA_DDL = `
55
55
  file_path TEXT UNIQUE,
56
56
  identity_key TEXT,
57
57
  expires_at TEXT,
58
+ superseded_by TEXT,
58
59
  created_at TEXT DEFAULT (datetime('now')),
60
+ updated_at TEXT,
59
61
  user_id TEXT,
60
62
  team_id TEXT,
61
63
  body_encrypted BLOB,
@@ -67,9 +69,11 @@ export const SCHEMA_DDL = `
67
69
  CREATE INDEX IF NOT EXISTS idx_vault_kind ON vault(kind);
68
70
  CREATE INDEX IF NOT EXISTS idx_vault_category ON vault(category);
69
71
  CREATE INDEX IF NOT EXISTS idx_vault_category_created ON vault(category, created_at DESC);
72
+ CREATE INDEX IF NOT EXISTS idx_vault_updated ON vault(updated_at DESC);
70
73
  CREATE INDEX IF NOT EXISTS idx_vault_user ON vault(user_id);
71
74
  CREATE INDEX IF NOT EXISTS idx_vault_team ON vault(team_id);
72
75
  CREATE UNIQUE INDEX IF NOT EXISTS idx_vault_identity ON vault(user_id, kind, identity_key) WHERE identity_key IS NOT NULL;
76
+ CREATE INDEX IF NOT EXISTS idx_vault_superseded ON vault(superseded_by) WHERE superseded_by IS NOT NULL;
73
77
 
74
78
  -- Single FTS5 table
75
79
  CREATE VIRTUAL TABLE IF NOT EXISTS vault_fts USING fts5(
@@ -147,13 +151,13 @@ export async function initDatabase(dbPath) {
147
151
 
148
152
  const freshDb = createDb(dbPath);
149
153
  freshDb.exec(SCHEMA_DDL);
150
- freshDb.exec("PRAGMA user_version = 7");
154
+ freshDb.exec("PRAGMA user_version = 9");
151
155
  return freshDb;
152
156
  }
153
157
 
154
158
  if (version < 5) {
155
159
  db.exec(SCHEMA_DDL);
156
- db.exec("PRAGMA user_version = 7");
160
+ db.exec("PRAGMA user_version = 9");
157
161
  } else if (version === 5) {
158
162
  // v5 -> v6 migration: add multi-tenancy + encryption columns
159
163
  // Wrapped in transaction with duplicate-column guards for idempotent retry
@@ -171,24 +175,91 @@ export async function initDatabase(dbPath) {
171
175
  addColumnSafe(`ALTER TABLE vault ADD COLUMN meta_encrypted BLOB`);
172
176
  addColumnSafe(`ALTER TABLE vault ADD COLUMN iv BLOB`);
173
177
  addColumnSafe(`ALTER TABLE vault ADD COLUMN team_id TEXT`);
178
+ addColumnSafe(`ALTER TABLE vault ADD COLUMN updated_at TEXT`);
179
+ addColumnSafe(`ALTER TABLE vault ADD COLUMN superseded_by TEXT`);
174
180
  db.exec(`CREATE INDEX IF NOT EXISTS idx_vault_user ON vault(user_id)`);
175
181
  db.exec(`CREATE INDEX IF NOT EXISTS idx_vault_team ON vault(team_id)`);
176
182
  db.exec(`DROP INDEX IF EXISTS idx_vault_identity`);
177
183
  db.exec(
178
184
  `CREATE UNIQUE INDEX IF NOT EXISTS idx_vault_identity ON vault(user_id, kind, identity_key) WHERE identity_key IS NOT NULL`,
179
185
  );
180
- db.exec("PRAGMA user_version = 7");
186
+ db.exec(
187
+ `UPDATE vault SET updated_at = created_at WHERE updated_at IS NULL`,
188
+ );
189
+ db.exec(
190
+ `CREATE INDEX IF NOT EXISTS idx_vault_updated ON vault(updated_at DESC)`,
191
+ );
192
+ db.exec(
193
+ `CREATE INDEX IF NOT EXISTS idx_vault_superseded ON vault(superseded_by) WHERE superseded_by IS NOT NULL`,
194
+ );
195
+ db.exec("PRAGMA user_version = 9");
181
196
  });
182
197
  } else if (version === 6) {
183
- // v6 -> v7 migration: add team_id column
198
+ // v6 -> v7+v8+v9 migration: add team_id, updated_at, superseded_by columns
184
199
  runTransaction(db, () => {
185
200
  try {
186
201
  db.exec(`ALTER TABLE vault ADD COLUMN team_id TEXT`);
187
202
  } catch (e) {
188
203
  if (!e.message.includes("duplicate column")) throw e;
189
204
  }
205
+ try {
206
+ db.exec(`ALTER TABLE vault ADD COLUMN updated_at TEXT`);
207
+ } catch (e) {
208
+ if (!e.message.includes("duplicate column")) throw e;
209
+ }
210
+ try {
211
+ db.exec(`ALTER TABLE vault ADD COLUMN superseded_by TEXT`);
212
+ } catch (e) {
213
+ if (!e.message.includes("duplicate column")) throw e;
214
+ }
190
215
  db.exec(`CREATE INDEX IF NOT EXISTS idx_vault_team ON vault(team_id)`);
191
- db.exec("PRAGMA user_version = 7");
216
+ db.exec(
217
+ `UPDATE vault SET updated_at = created_at WHERE updated_at IS NULL`,
218
+ );
219
+ db.exec(
220
+ `CREATE INDEX IF NOT EXISTS idx_vault_updated ON vault(updated_at DESC)`,
221
+ );
222
+ db.exec(
223
+ `CREATE INDEX IF NOT EXISTS idx_vault_superseded ON vault(superseded_by) WHERE superseded_by IS NOT NULL`,
224
+ );
225
+ db.exec("PRAGMA user_version = 9");
226
+ });
227
+ } else if (version === 7) {
228
+ // v7 -> v8+v9 migration: add updated_at, superseded_by columns
229
+ runTransaction(db, () => {
230
+ try {
231
+ db.exec(`ALTER TABLE vault ADD COLUMN updated_at TEXT`);
232
+ } catch (e) {
233
+ if (!e.message.includes("duplicate column")) throw e;
234
+ }
235
+ try {
236
+ db.exec(`ALTER TABLE vault ADD COLUMN superseded_by TEXT`);
237
+ } catch (e) {
238
+ if (!e.message.includes("duplicate column")) throw e;
239
+ }
240
+ db.exec(
241
+ `UPDATE vault SET updated_at = created_at WHERE updated_at IS NULL`,
242
+ );
243
+ db.exec(
244
+ `CREATE INDEX IF NOT EXISTS idx_vault_updated ON vault(updated_at DESC)`,
245
+ );
246
+ db.exec(
247
+ `CREATE INDEX IF NOT EXISTS idx_vault_superseded ON vault(superseded_by) WHERE superseded_by IS NOT NULL`,
248
+ );
249
+ db.exec("PRAGMA user_version = 9");
250
+ });
251
+ } else if (version === 8) {
252
+ // v8 -> v9 migration: add superseded_by column
253
+ runTransaction(db, () => {
254
+ try {
255
+ db.exec(`ALTER TABLE vault ADD COLUMN superseded_by TEXT`);
256
+ } catch (e) {
257
+ if (!e.message.includes("duplicate column")) throw e;
258
+ }
259
+ db.exec(
260
+ `CREATE INDEX IF NOT EXISTS idx_vault_superseded ON vault(superseded_by) WHERE superseded_by IS NOT NULL`,
261
+ );
262
+ db.exec("PRAGMA user_version = 9");
192
263
  });
193
264
  }
194
265
 
@@ -199,13 +270,13 @@ export function prepareStatements(db) {
199
270
  try {
200
271
  return {
201
272
  insertEntry: db.prepare(
202
- `INSERT INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
273
+ `INSERT INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at, updated_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
203
274
  ),
204
275
  insertEntryEncrypted: db.prepare(
205
- `INSERT INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at, body_encrypted, title_encrypted, meta_encrypted, iv) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
276
+ `INSERT INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at, updated_at, body_encrypted, title_encrypted, meta_encrypted, iv) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
206
277
  ),
207
278
  updateEntry: db.prepare(
208
- `UPDATE vault SET title = ?, body = ?, meta = ?, tags = ?, source = ?, category = ?, identity_key = ?, expires_at = ? WHERE file_path = ?`,
279
+ `UPDATE vault SET title = ?, body = ?, meta = ?, tags = ?, source = ?, category = ?, identity_key = ?, expires_at = ?, updated_at = datetime('now') WHERE file_path = ?`,
209
280
  ),
210
281
  deleteEntry: db.prepare(`DELETE FROM vault WHERE id = ?`),
211
282
  getRowid: db.prepare(`SELECT rowid FROM vault WHERE id = ?`),
@@ -215,12 +286,18 @@ export function prepareStatements(db) {
215
286
  `SELECT * FROM vault WHERE kind = ? AND identity_key = ? AND user_id IS ?`,
216
287
  ),
217
288
  upsertByIdentityKey: db.prepare(
218
- `UPDATE vault SET title = ?, body = ?, meta = ?, tags = ?, source = ?, category = ?, file_path = ?, expires_at = ? WHERE kind = ? AND identity_key = ? AND user_id IS ?`,
289
+ `UPDATE vault SET title = ?, body = ?, meta = ?, tags = ?, source = ?, category = ?, file_path = ?, expires_at = ?, updated_at = datetime('now') WHERE kind = ? AND identity_key = ? AND user_id IS ?`,
219
290
  ),
220
291
  insertVecStmt: db.prepare(
221
292
  `INSERT INTO vault_vec (rowid, embedding) VALUES (?, ?)`,
222
293
  ),
223
294
  deleteVecStmt: db.prepare(`DELETE FROM vault_vec WHERE rowid = ?`),
295
+ updateSupersededBy: db.prepare(
296
+ `UPDATE vault SET superseded_by = ? WHERE id = ?`,
297
+ ),
298
+ clearSupersededByRef: db.prepare(
299
+ `UPDATE vault SET superseded_by = NULL WHERE superseded_by = ?`,
300
+ ),
224
301
  };
225
302
  } catch (e) {
226
303
  throw new Error(
@@ -111,6 +111,7 @@ export async function indexEntry(
111
111
  identity_key || null,
112
112
  expires_at || null,
113
113
  createdAt,
114
+ createdAt,
114
115
  encrypted.body_encrypted,
115
116
  encrypted.title_encrypted,
116
117
  encrypted.meta_encrypted,
@@ -131,6 +132,7 @@ export async function indexEntry(
131
132
  identity_key || null,
132
133
  expires_at || null,
133
134
  createdAt,
135
+ createdAt,
134
136
  );
135
137
  }
136
138
  } catch (e) {
@@ -186,6 +188,39 @@ export async function indexEntry(
186
188
  }
187
189
  }
188
190
 
191
+ /**
192
+ * Prune expired entries: delete files, vec rows, and DB rows for all entries
193
+ * where expires_at <= now(). Safe to call on startup or CLI — non-destructive
194
+ * to active data.
195
+ *
196
+ * @param {import('../server/types.js').BaseCtx} ctx
197
+ * @returns {Promise<number>} count of pruned entries
198
+ */
199
+ export async function pruneExpired(ctx) {
200
+ const expired = ctx.db
201
+ .prepare(
202
+ "SELECT id, file_path FROM vault WHERE expires_at IS NOT NULL AND expires_at <= datetime('now')",
203
+ )
204
+ .all();
205
+
206
+ for (const row of expired) {
207
+ if (row.file_path) {
208
+ try {
209
+ unlinkSync(row.file_path);
210
+ } catch {}
211
+ }
212
+ const vRowid = ctx.stmts.getRowid.get(row.id)?.rowid;
213
+ if (vRowid) {
214
+ try {
215
+ ctx.deleteVec(Number(vRowid));
216
+ } catch {}
217
+ }
218
+ ctx.stmts.deleteEntry.run(row.id);
219
+ }
220
+
221
+ return expired.length;
222
+ }
223
+
189
224
  /**
190
225
  * Bulk reindex: sync vault directory into the database.
191
226
  * P2: Wrapped in a transaction for atomicity.
@@ -205,7 +240,7 @@ export async function reindex(ctx, opts = {}) {
205
240
  // Use INSERT OR IGNORE for reindex — handles files with duplicate frontmatter IDs
206
241
  // user_id is NULL for reindex (always local mode)
207
242
  const upsertEntry = ctx.db.prepare(
208
- `INSERT OR IGNORE INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at) VALUES (?, NULL, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
243
+ `INSERT OR IGNORE INTO vault (id, user_id, kind, category, title, body, meta, tags, source, file_path, identity_key, expires_at, created_at, updated_at) VALUES (?, NULL, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
209
244
  );
210
245
 
211
246
  // Auto-discover kind directories, supporting both:
@@ -309,6 +344,7 @@ export async function reindex(ctx, opts = {}) {
309
344
  identity_key,
310
345
  expires_at,
311
346
  created,
347
+ fmMeta.updated || created,
312
348
  );
313
349
  if (result.changes > 0) {
314
350
  const rowidResult = ctx.stmts.getRowid.get(id);
package/src/index.js CHANGED
@@ -47,7 +47,7 @@ export {
47
47
  deleteVec,
48
48
  } from "./index/db.js";
49
49
  export { embed, embedBatch, resetEmbedPipeline } from "./index/embed.js";
50
- export { indexEntry, reindex } from "./index/index.js";
50
+ export { indexEntry, reindex, pruneExpired } from "./index/index.js";
51
51
 
52
52
  // Retrieve layer
53
53
  export { hybridSearch } from "./retrieve/index.js";
@@ -9,10 +9,23 @@
9
9
 
10
10
  const FTS_WEIGHT = 0.4;
11
11
  const VEC_WEIGHT = 0.6;
12
+ const NEAR_DUP_THRESHOLD = 0.92;
12
13
 
13
14
  /**
14
- * Strip FTS5 metacharacters from query words and build an AND query.
15
- * Returns null if no valid words remain.
15
+ * Dot product of two Float32Array vectors (cosine similarity for unit vectors).
16
+ */
17
+ export function dotProduct(a, b) {
18
+ let sum = 0;
19
+ for (let i = 0; i < a.length; i++) sum += a[i] * b[i];
20
+ return sum;
21
+ }
22
+
23
+ /**
24
+ * Build a tiered FTS5 query that prioritises phrase match, then proximity,
25
+ * then AND. Multi-word queries become:
26
+ * "word1 word2" OR NEAR("word1" "word2", 10) OR "word1" AND "word2"
27
+ * Single-word queries remain a simple quoted term.
28
+ * Returns null if no valid words remain after stripping FTS5 metacharacters.
16
29
  */
17
30
  export function buildFtsQuery(query) {
18
31
  const words = query
@@ -20,7 +33,11 @@ export function buildFtsQuery(query) {
20
33
  .map((w) => w.replace(/[*"():^~{}]/g, ""))
21
34
  .filter((w) => w.length > 0);
22
35
  if (!words.length) return null;
23
- return words.map((w) => `"${w}"`).join(" AND ");
36
+ if (words.length === 1) return `"${words[0]}"`;
37
+ const phrase = `"${words.join(" ")}"`;
38
+ const near = `NEAR(${words.map((w) => `"${w}"`).join(" ")}, 10)`;
39
+ const and = words.map((w) => `"${w}"`).join(" AND ");
40
+ return `${phrase} OR ${near} OR ${and}`;
24
41
  }
25
42
 
26
43
  /**
@@ -44,6 +61,7 @@ export function buildFilterClauses({
44
61
  until,
45
62
  userIdFilter,
46
63
  teamIdFilter,
64
+ includeSuperseeded = false,
47
65
  }) {
48
66
  const clauses = [];
49
67
  const params = [];
@@ -68,6 +86,9 @@ export function buildFilterClauses({
68
86
  params.push(until);
69
87
  }
70
88
  clauses.push("(e.expires_at IS NULL OR e.expires_at > datetime('now'))");
89
+ if (!includeSuperseeded) {
90
+ clauses.push("e.superseded_by IS NULL");
91
+ }
71
92
  return { clauses, params };
72
93
  }
73
94
 
@@ -92,15 +113,19 @@ export async function hybridSearch(
92
113
  decayDays = 30,
93
114
  userIdFilter,
94
115
  teamIdFilter = null,
116
+ includeSuperseeded = false,
95
117
  } = {},
96
118
  ) {
97
119
  const results = new Map();
120
+ const idToRowid = new Map();
121
+ let queryVec = null;
98
122
  const extraFilters = buildFilterClauses({
99
123
  categoryFilter,
100
124
  since,
101
125
  until,
102
126
  userIdFilter,
103
127
  teamIdFilter,
128
+ includeSuperseeded,
104
129
  });
105
130
 
106
131
  // FTS5 search
@@ -144,7 +169,7 @@ export async function hybridSearch(
144
169
  .prepare("SELECT COUNT(*) as c FROM vault_vec")
145
170
  .get().c;
146
171
  if (vecCount > 0) {
147
- const queryVec = await ctx.embed(query);
172
+ queryVec = await ctx.embed(query);
148
173
  if (queryVec) {
149
174
  // Increase limits in hosted mode to compensate for post-filtering
150
175
  const hasPostFilter = userIdFilter !== undefined || teamIdFilter;
@@ -188,6 +213,7 @@ export async function hybridSearch(
188
213
  continue;
189
214
 
190
215
  const { rowid: _rowid, ...cleanRow } = row;
216
+ idToRowid.set(cleanRow.id, Number(row.rowid));
191
217
  // sqlite-vec returns L2 distance [0, 2] for normalized vectors.
192
218
  // Convert to similarity [1, 0] with: 1 - distance/2
193
219
  const vecScore = Math.max(0, 1 - vr.distance / 2) * VEC_WEIGHT;
@@ -215,5 +241,56 @@ export async function hybridSearch(
215
241
  }
216
242
 
217
243
  const sorted = [...results.values()].sort((a, b) => b.score - a.score);
244
+
245
+ // Near-duplicate suppression: when embeddings are available and we have more
246
+ // candidates than needed, skip results that are too similar to already-selected ones.
247
+ if (queryVec && idToRowid.size > 0 && sorted.length > limit) {
248
+ const rowidsToFetch = sorted
249
+ .filter((c) => idToRowid.has(c.id))
250
+ .map((c) => idToRowid.get(c.id));
251
+
252
+ const embeddingMap = new Map();
253
+ if (rowidsToFetch.length > 0) {
254
+ try {
255
+ const placeholders = rowidsToFetch.map(() => "?").join(",");
256
+ const vecData = ctx.db
257
+ .prepare(
258
+ `SELECT rowid, embedding FROM vault_vec WHERE rowid IN (${placeholders})`,
259
+ )
260
+ .all(...rowidsToFetch);
261
+ for (const row of vecData) {
262
+ const buf = row.embedding;
263
+ if (buf) {
264
+ embeddingMap.set(
265
+ Number(row.rowid),
266
+ new Float32Array(buf.buffer, buf.byteOffset, buf.byteLength / 4),
267
+ );
268
+ }
269
+ }
270
+ } catch (_) {
271
+ return sorted.slice(offset, offset + limit);
272
+ }
273
+ }
274
+
275
+ const selected = [];
276
+ const selectedVecs = [];
277
+ for (const candidate of sorted) {
278
+ if (selected.length >= offset + limit) break;
279
+ const rowid = idToRowid.get(candidate.id);
280
+ const vec = rowid !== undefined ? embeddingMap.get(rowid) : null;
281
+ if (vec && selectedVecs.length > 0) {
282
+ let maxSim = 0;
283
+ for (const sv of selectedVecs) {
284
+ const sim = dotProduct(sv, vec);
285
+ if (sim > maxSim) maxSim = sim;
286
+ }
287
+ if (maxSim > NEAR_DUP_THRESHOLD) continue;
288
+ }
289
+ selected.push(candidate);
290
+ if (vec) selectedVecs.push(vec);
291
+ }
292
+ return selected.slice(offset, offset + limit);
293
+ }
294
+
218
295
  return sorted.slice(offset, offset + limit);
219
296
  }
@@ -1,4 +1,4 @@
1
- import { gatherVaultStatus } from "../../core/status.js";
1
+ import { gatherVaultStatus, computeGrowthWarnings } from "../../core/status.js";
2
2
  import { errorLogPath, errorLogCount } from "../../core/error-log.js";
3
3
  import { ok } from "../helpers.js";
4
4
 
@@ -40,7 +40,7 @@ export function handler(_args, ctx) {
40
40
  `Data dir: ${config.dataDir}`,
41
41
  `Config: ${config.configPath}`,
42
42
  `Resolved via: ${status.resolvedFrom}`,
43
- `Schema: v7 (teams)`,
43
+ `Schema: v9 (updated_at, superseded_by)`,
44
44
  ];
45
45
 
46
46
  if (status.embeddingStatus) {
@@ -58,7 +58,7 @@ export function handler(_args, ctx) {
58
58
  lines.push(`Decay: ${config.eventDecayDays} days (event recency window)`);
59
59
  if (status.expiredCount > 0) {
60
60
  lines.push(
61
- `Expired: ${status.expiredCount} entries (pruned on next reindex)`,
61
+ `Expired: ${status.expiredCount} entries pending prune (run \`context-vault prune\` to remove now)`,
62
62
  );
63
63
  }
64
64
 
@@ -93,6 +93,25 @@ export function handler(_args, ctx) {
93
93
  lines.push(`Auto-reindex will fix this on next search or save.`);
94
94
  }
95
95
 
96
+ if (status.staleKnowledge?.length > 0) {
97
+ lines.push(``);
98
+ lines.push(`### ⚠ Potentially Stale Knowledge`);
99
+ lines.push(
100
+ `Not updated within kind staleness window (pattern: 180d, decision: 365d, reference: 90d):`,
101
+ );
102
+ for (const entry of status.staleKnowledge) {
103
+ const lastUpdated = entry.last_updated
104
+ ? entry.last_updated.split("T")[0]
105
+ : "unknown";
106
+ lines.push(
107
+ `- "${entry.title}" (${entry.kind}) — last updated ${lastUpdated}`,
108
+ );
109
+ }
110
+ lines.push(
111
+ `Use save_context to refresh or add expires_at to retire stale entries.`,
112
+ );
113
+ }
114
+
96
115
  // Error log
97
116
  const logPath = errorLogPath(config.dataDir);
98
117
  const logCount = errorLogCount(config.dataDir);
@@ -120,6 +139,21 @@ export function handler(_args, ctx) {
120
139
  }
121
140
  }
122
141
 
142
+ // Growth warnings
143
+ const growth = computeGrowthWarnings(status, config.thresholds);
144
+ if (growth.hasWarnings) {
145
+ lines.push("", "### ⚠ Vault Growth Warning");
146
+ for (const w of growth.warnings) {
147
+ lines.push(` ${w.message}`);
148
+ }
149
+ if (growth.actions.length) {
150
+ lines.push("", "Suggested growth actions:");
151
+ for (const a of growth.actions) {
152
+ lines.push(` • ${a}`);
153
+ }
154
+ }
155
+ }
156
+
123
157
  // Suggested actions
124
158
  const actions = [];
125
159
  if (status.stalePaths)
@@ -42,6 +42,12 @@ export const inputSchema = {
42
42
  .optional()
43
43
  .describe("ISO date, return entries created before this"),
44
44
  limit: z.number().optional().describe("Max results to return (default 10)"),
45
+ include_superseded: z
46
+ .boolean()
47
+ .optional()
48
+ .describe(
49
+ "If true, include entries that have been superseded by newer ones. Default: false.",
50
+ ),
45
51
  };
46
52
 
47
53
  /**
@@ -50,7 +56,17 @@ export const inputSchema = {
50
56
  * @param {import('../types.js').ToolShared} shared
51
57
  */
52
58
  export async function handler(
53
- { query, kind, category, identity_key, tags, since, until, limit },
59
+ {
60
+ query,
61
+ kind,
62
+ category,
63
+ identity_key,
64
+ tags,
65
+ since,
66
+ until,
67
+ limit,
68
+ include_superseded,
69
+ },
54
70
  ctx,
55
71
  { ensureIndexed, reindexFailed },
56
72
  ) {
@@ -126,6 +142,7 @@ export async function handler(
126
142
  limit: fetchLimit,
127
143
  decayDays: config.eventDecayDays || 30,
128
144
  userIdFilter: userId,
145
+ includeSuperseeded: include_superseded ?? false,
129
146
  });
130
147
 
131
148
  // Post-filter by tags if provided, then apply requested limit
@@ -238,8 +255,12 @@ export async function handler(
238
255
  lines.push(
239
256
  `### [${i + 1}/${filtered.length}] ${r.title || "(untitled)"} [${r.kind}/${r.category}]`,
240
257
  );
258
+ const dateStr =
259
+ r.updated_at && r.updated_at !== r.created_at
260
+ ? `${r.created_at} (updated ${r.updated_at})`
261
+ : r.created_at || "";
241
262
  lines.push(
242
- `${r.score.toFixed(3)} · ${tagStr} · ${relPath} · id: \`${r.id}\``,
263
+ `${r.score.toFixed(3)} · ${tagStr} · ${relPath} · ${dateStr} · id: \`${r.id}\``,
243
264
  );
244
265
  lines.push(r.body?.slice(0, 300) + (r.body?.length > 300 ? "..." : ""));
245
266
  lines.push("");
@@ -65,17 +65,6 @@ export async function handler(
65
65
 
66
66
  await ensureIndexed();
67
67
 
68
- // Hosted tier limit enforcement
69
- if (ctx.checkLimits) {
70
- const usage = ctx.checkLimits();
71
- if (usage.entryCount >= usage.maxEntries) {
72
- return err(
73
- `Entry limit reached (${usage.maxEntries}). Upgrade to Pro for unlimited entries.`,
74
- "LIMIT_EXCEEDED",
75
- );
76
- }
77
- }
78
-
79
68
  try {
80
69
  const { ingestUrl } = await import("../../capture/ingest-url.js");
81
70
  const entryData = await ingestUrl(targetUrl, { kind, tags });
@@ -6,7 +6,7 @@ import { ok } from "../helpers.js";
6
6
  export const name = "list_context";
7
7
 
8
8
  export const description =
9
- "Browse vault entries without a search query. Returns id, title, kind, category, tags, created_at. Use get_context with a query for semantic search. Use this to browse by tags or find recent entries.";
9
+ "Browse vault entries without a search query. Returns id, title, kind, category, tags, created_at, updated_at. Use get_context with a query for semantic search. Use this to browse by tags or find recent entries.";
10
10
 
11
11
  export const inputSchema = {
12
12
  kind: z
@@ -101,7 +101,7 @@ export async function handler(
101
101
  params.push(fetchLimit, effectiveOffset);
102
102
  const rows = ctx.db
103
103
  .prepare(
104
- `SELECT id, title, kind, category, tags, created_at, SUBSTR(body, 1, 120) as preview FROM vault ${where} ORDER BY created_at DESC LIMIT ? OFFSET ?`,
104
+ `SELECT id, title, kind, category, tags, created_at, updated_at, SUBSTR(body, 1, 120) as preview FROM vault ${where} ORDER BY created_at DESC LIMIT ? OFFSET ?`,
105
105
  )
106
106
  .all(...params);
107
107
 
@@ -140,8 +140,12 @@ export async function handler(
140
140
  for (const r of filtered) {
141
141
  const entryTags = r.tags ? JSON.parse(r.tags) : [];
142
142
  const tagStr = entryTags.length ? entryTags.join(", ") : "none";
143
+ const dateStr =
144
+ r.updated_at && r.updated_at !== r.created_at
145
+ ? `${r.created_at} (updated ${r.updated_at})`
146
+ : r.created_at;
143
147
  lines.push(
144
- `- **${r.title || "(untitled)"}** [${r.kind}/${r.category}] — ${tagStr} — ${r.created_at} — \`${r.id}\``,
148
+ `- **${r.title || "(untitled)"}** [${r.kind}/${r.category}] — ${tagStr} — ${dateStr} — \`${r.id}\``,
145
149
  );
146
150
  if (r.preview)
147
151
  lines.push(
@@ -15,6 +15,63 @@ import {
15
15
  MAX_IDENTITY_KEY_LENGTH,
16
16
  } from "../../constants.js";
17
17
 
18
+ const DEFAULT_SIMILARITY_THRESHOLD = 0.85;
19
+
20
+ async function findSimilar(ctx, embedding, threshold, userId) {
21
+ try {
22
+ const vecCount = ctx.db
23
+ .prepare("SELECT COUNT(*) as c FROM vault_vec")
24
+ .get().c;
25
+ if (vecCount === 0) return [];
26
+
27
+ const vecRows = ctx.db
28
+ .prepare(
29
+ `SELECT v.rowid, v.distance FROM vault_vec v WHERE embedding MATCH ? ORDER BY distance LIMIT ?`,
30
+ )
31
+ .all(embedding, 10);
32
+
33
+ if (!vecRows.length) return [];
34
+
35
+ const rowids = vecRows.map((vr) => vr.rowid);
36
+ const placeholders = rowids.map(() => "?").join(",");
37
+ const hydrated = ctx.db
38
+ .prepare(
39
+ `SELECT rowid, id, title, category, user_id FROM vault WHERE rowid IN (${placeholders})`,
40
+ )
41
+ .all(...rowids);
42
+
43
+ const byRowid = new Map();
44
+ for (const row of hydrated) byRowid.set(row.rowid, row);
45
+
46
+ const results = [];
47
+ for (const vr of vecRows) {
48
+ const similarity = Math.max(0, 1 - vr.distance / 2);
49
+ if (similarity < threshold) continue;
50
+ const row = byRowid.get(vr.rowid);
51
+ if (!row) continue;
52
+ if (userId !== undefined && row.user_id !== userId) continue;
53
+ if (row.category === "entity") continue;
54
+ results.push({ id: row.id, title: row.title, score: similarity });
55
+ }
56
+ return results;
57
+ } catch {
58
+ return [];
59
+ }
60
+ }
61
+
62
+ function formatSimilarWarning(similar) {
63
+ const lines = ["", "⚠ Similar entries already exist:"];
64
+ for (const e of similar) {
65
+ const score = e.score.toFixed(2);
66
+ const title = e.title ? `"${e.title}"` : "(no title)";
67
+ lines.push(` - ${title} (${score}) — id: ${e.id}`);
68
+ }
69
+ lines.push(
70
+ " Consider using `id: <existing>` in save_context to update instead.",
71
+ );
72
+ return lines.join("\n");
73
+ }
74
+
18
75
  /**
19
76
  * Validate input fields for save_context. Returns an error response or null.
20
77
  */
@@ -150,6 +207,26 @@ export const inputSchema = {
150
207
  "Required for entity kinds (contact, project, tool, source). The unique identifier for this entity.",
151
208
  ),
152
209
  expires_at: z.string().optional().describe("ISO date for TTL expiry"),
210
+ supersedes: z
211
+ .array(z.string())
212
+ .optional()
213
+ .describe(
214
+ "Array of entry IDs that this entry supersedes/replaces. Those entries will be marked with superseded_by pointing to this new entry and excluded from future search results by default.",
215
+ ),
216
+ dry_run: z
217
+ .boolean()
218
+ .optional()
219
+ .describe(
220
+ "If true, check for similar entries without saving. Returns similarity results without creating a new entry. Only applies to knowledge and event categories.",
221
+ ),
222
+ similarity_threshold: z
223
+ .number()
224
+ .min(0)
225
+ .max(1)
226
+ .optional()
227
+ .describe(
228
+ "Cosine similarity threshold for duplicate detection (0–1, default 0.85). Entries above this score are flagged as similar. Only applies to knowledge and event categories.",
229
+ ),
153
230
  };
154
231
 
155
232
  /**
@@ -169,6 +246,9 @@ export async function handler(
169
246
  source,
170
247
  identity_key,
171
248
  expires_at,
249
+ supersedes,
250
+ dry_run,
251
+ similarity_threshold,
172
252
  },
173
253
  ctx,
174
254
  { ensureIndexed },
@@ -231,6 +311,7 @@ export async function handler(
231
311
  meta,
232
312
  source,
233
313
  expires_at,
314
+ supersedes,
234
315
  });
235
316
  await indexEntry(ctx, entry);
236
317
  const relPath = entry.filePath
@@ -261,25 +342,45 @@ export async function handler(
261
342
  );
262
343
  }
263
344
 
264
- // Hosted tier limit enforcement (skipped in local mode — no checkLimits on ctx)
265
- if (ctx.checkLimits) {
266
- const usage = ctx.checkLimits();
267
- if (usage.entryCount >= usage.maxEntries) {
268
- return err(
269
- `Entry limit reached (${usage.maxEntries}). Upgrade to Pro for unlimited entries.`,
270
- "LIMIT_EXCEEDED",
345
+ await ensureIndexed();
346
+
347
+ // ── Similarity check (knowledge + event only) ────────────────────────────
348
+ const category = categoryFor(normalizedKind);
349
+ let similarEntries = [];
350
+
351
+ if (category === "knowledge" || category === "event") {
352
+ const threshold = similarity_threshold ?? DEFAULT_SIMILARITY_THRESHOLD;
353
+ const embeddingText = [title, body].filter(Boolean).join(" ");
354
+ const queryEmbedding = await ctx.embed(embeddingText);
355
+ if (queryEmbedding) {
356
+ similarEntries = await findSimilar(
357
+ ctx,
358
+ queryEmbedding,
359
+ threshold,
360
+ userId,
271
361
  );
272
362
  }
273
- if (usage.storageMb >= usage.maxStorageMb) {
274
- return err(
275
- `Storage limit reached (${usage.maxStorageMb} MB). Upgrade to Pro for more storage.`,
276
- "LIMIT_EXCEEDED",
363
+ }
364
+
365
+ if (dry_run) {
366
+ const parts = ["(dry run — nothing saved)"];
367
+ if (similarEntries.length) {
368
+ parts.push("", "⚠ Similar entries already exist:");
369
+ for (const e of similarEntries) {
370
+ const score = e.score.toFixed(2);
371
+ const titleDisplay = e.title ? `"${e.title}"` : "(no title)";
372
+ parts.push(` - ${titleDisplay} (${score}) — id: ${e.id}`);
373
+ }
374
+ parts.push(
375
+ "",
376
+ "Use save_context with `id: <existing>` to update one, or omit `dry_run` to save as new.",
277
377
  );
378
+ } else {
379
+ parts.push("", "No similar entries found. Safe to save.");
278
380
  }
381
+ return ok(parts.join("\n"));
279
382
  }
280
383
 
281
- await ensureIndexed();
282
-
283
384
  const mergedMeta = { ...(meta || {}) };
284
385
  if (folder) mergedMeta.folder = folder;
285
386
  const finalMeta = Object.keys(mergedMeta).length ? mergedMeta : undefined;
@@ -294,6 +395,7 @@ export async function handler(
294
395
  folder,
295
396
  identity_key,
296
397
  expires_at,
398
+ supersedes,
297
399
  userId,
298
400
  });
299
401
  const relPath = entry.filePath
@@ -303,5 +405,28 @@ export async function handler(
303
405
  if (title) parts.push(` title: ${title}`);
304
406
  if (tags?.length) parts.push(` tags: ${tags.join(", ")}`);
305
407
  parts.push("", "_Use this id to update or delete later._");
408
+ if (similarEntries.length) {
409
+ parts.push(formatSimilarWarning(similarEntries));
410
+ }
411
+
412
+ const criticalLimit = config.thresholds?.totalEntries?.critical;
413
+ if (criticalLimit != null) {
414
+ try {
415
+ const countRow = ctx.db
416
+ .prepare(
417
+ userId !== undefined
418
+ ? "SELECT COUNT(*) as c FROM vault WHERE user_id = ?"
419
+ : "SELECT COUNT(*) as c FROM vault",
420
+ )
421
+ .get(...(userId !== undefined ? [userId] : []));
422
+ if (countRow.c >= criticalLimit) {
423
+ parts.push(
424
+ ``,
425
+ `ℹ Vault has ${countRow.c.toLocaleString()} entries. Consider running \`context-vault reindex\` or reviewing old entries.`,
426
+ );
427
+ }
428
+ } catch {}
429
+ }
430
+
306
431
  return ok(parts.join("\n"));
307
432
  }
@@ -1,6 +1,7 @@
1
1
  import { reindex } from "../index/index.js";
2
2
  import { captureAndIndex } from "../capture/index.js";
3
3
  import { err } from "./helpers.js";
4
+ import { sendTelemetryEvent } from "../core/telemetry.js";
4
5
  import pkg from "../../package.json" with { type: "json" };
5
6
 
6
7
  import * as getContext from "./tools/get-context.js";
@@ -57,6 +58,12 @@ export function registerTools(server, ctx) {
57
58
  timestamp: Date.now(),
58
59
  };
59
60
  }
61
+ sendTelemetryEvent(ctx.config, {
62
+ event: "tool_error",
63
+ code: "TIMEOUT",
64
+ tool: toolName,
65
+ cv_version: pkg.version,
66
+ });
60
67
  return err(
61
68
  "Tool timed out after 60s. Try a simpler query or run `context-vault reindex` first.",
62
69
  "TIMEOUT",
@@ -70,6 +77,12 @@ export function registerTools(server, ctx) {
70
77
  timestamp: Date.now(),
71
78
  };
72
79
  }
80
+ sendTelemetryEvent(ctx.config, {
81
+ event: "tool_error",
82
+ code: "UNKNOWN",
83
+ tool: toolName,
84
+ cv_version: pkg.version,
85
+ });
73
86
  try {
74
87
  await captureAndIndex(ctx, {
75
88
  kind: "feedback",