claude-mem-lite 2.5.4 → 2.9.2

package/tool-schemas.mjs

@@ -5,6 +5,29 @@ import { z } from 'zod';
 
 export const OBS_TYPE_ENUM = z.enum(['decision', 'bugfix', 'feature', 'refactor', 'discovery', 'change']);
 
+// LLM-friendly coercion: accept string numbers and normalize to proper types
+const coerceInt = z.preprocess(
+  (v) => (typeof v === 'string' && /^-?\d+$/.test(v.trim())) ? parseInt(v.trim(), 10) : v,
+  z.number().int()
+);
+
+// LLM-friendly coercion: accept "true"/"false"/"True"/"TRUE" strings as boolean
+const coerceBool = z.preprocess(
+  (v) => typeof v === 'string' ? ({ true: true, false: false })[v.toLowerCase()] ?? v : v,
+  z.boolean()
+);
+
+// Coerce ids: accept single number, string "123", comma-separated "1,2,3", or array
+const coerceIntArray = z.preprocess(
+  (v) => {
+    if (Array.isArray(v)) return v.map(x => typeof x === 'string' ? parseInt(x, 10) : x);
+    if (typeof v === 'number') return [v];
+    if (typeof v === 'string') return v.split(',').map(s => parseInt(s.trim(), 10)).filter(n => !isNaN(n));
+    return v;
+  },
+  z.array(z.number().int())
+);
+
 export const memSearchSchema = {
   query: z.string().optional().describe('Search query (FTS5 syntax supported)'),
   type: z.enum(['observations', 'sessions', 'prompts']).optional().describe('Limit to one table'),
@@ -12,28 +35,28 @@ export const memSearchSchema = {
   project: z.string().optional().describe('Filter by project name'),
   date_from: z.string().optional().describe('Start date (ISO 8601 or YYYY-MM-DD)'),
   date_to: z.string().optional().describe('End date (ISO 8601 or YYYY-MM-DD). Date-only format is inclusive (covers full day)'),
-  importance: z.number().int().min(1).max(3).optional().describe('Minimum importance (1=routine, 2=notable, 3=critical)'),
-  limit: z.number().int().min(1).max(100).optional().describe('Max results (default 20)'),
-  offset: z.number().int().min(0).optional().describe('Offset for pagination'),
+  importance: coerceInt.pipe(z.number().int().min(1).max(3)).optional().describe('Minimum importance (1=routine, 2=notable, 3=critical)'),
+  limit: coerceInt.pipe(z.number().int().min(1).max(100)).optional().describe('Max results (default 20)'),
+  offset: coerceInt.pipe(z.number().int().min(0)).optional().describe('Offset for pagination'),
 };
 
 export const memTimelineSchema = {
-  anchor: z.number().int().optional().describe('Observation ID as center point'),
+  anchor: coerceInt.pipe(z.number().int()).optional().describe('Observation ID as center point'),
   query: z.string().optional().describe('FTS5 query to auto-find anchor'),
-  before: z.number().int().min(0).max(50).optional().describe('Items before anchor (default 5)'),
-  after: z.number().int().min(0).max(50).optional().describe('Items after anchor (default 5)'),
+  before: coerceInt.pipe(z.number().int().min(0).max(50)).optional().describe('Items before anchor (default 5)'),
+  after: coerceInt.pipe(z.number().int().min(0).max(50)).optional().describe('Items after anchor (default 5)'),
   project: z.string().optional().describe('Filter by project'),
 };
 
 export const memGetSchema = {
-  ids: z.array(z.number().int()).min(1).max(20).describe('Observation IDs to retrieve'),
+  ids: coerceIntArray.pipe(z.array(z.number().int()).min(1).max(20)).describe('Observation IDs to retrieve'),
   source: z.enum(['obs', 'session', 'prompt']).optional().describe('Record type: obs (default), session (S# from search), prompt (P# from search)'),
   fields: z.array(z.string()).optional().describe('Specific fields to return (default: all)'),
 };
 
 export const memDeleteSchema = {
-  ids: z.array(z.number().int()).min(1).max(50).describe('Observation IDs to delete'),
-  confirm: z.boolean().describe('false=preview what will be deleted, true=execute deletion'),
+  ids: coerceIntArray.pipe(z.array(z.number().int()).min(1).max(50)).describe('Observation IDs to delete'),
+  confirm: coerceBool.describe('false=preview what will be deleted, true=execute deletion'),
 };
 
 export const memSaveSchema = {
@@ -41,17 +64,17 @@ export const memSaveSchema = {
   title: z.string().optional().describe('Short title'),
   type: OBS_TYPE_ENUM.optional().describe('Observation type (default: discovery)'),
   project: z.string().optional().describe('Project name (default: inferred from CWD)'),
-  importance: z.number().int().min(1).max(3).optional().describe('Importance level: 1=routine, 2=notable, 3=critical (default: 1)'),
+  importance: coerceInt.pipe(z.number().int().min(1).max(3)).optional().describe('Importance level: 1=routine, 2=notable, 3=critical (default: 1)'),
 };
 
 export const memStatsSchema = {
   project: z.string().optional().describe('Filter by project'),
-  days: z.number().int().min(1).max(365).optional().describe('Look back N days (default 30)'),
+  days: coerceInt.pipe(z.number().int().min(1).max(365)).optional().describe('Look back N days (default 30)'),
 };
 
 export const memCompressSchema = {
-  preview: z.boolean().optional().describe('true=count candidates, false=execute compression (default: true)'),
-  age_days: z.number().int().min(30).max(365).optional().describe('Min age in days (default: 60)'),
+  preview: coerceBool.optional().describe('true=count candidates, false=execute compression (default: true)'),
+  age_days: coerceInt.pipe(z.number().int().min(30).max(365)).optional().describe('Min age in days (default: 60)'),
   project: z.string().optional().describe('Filter by project'),
 };
 
@@ -59,9 +82,11 @@ export const memMaintainSchema = {
   action: z.enum(['scan', 'execute']).describe('scan=analyze candidates, execute=apply changes'),
   operations: z.array(z.enum(['dedup', 'decay', 'cleanup', 'boost', 'purge_stale'])).optional()
     .describe('Operations to execute (for action=execute). purge_stale deletes idle-marked observations after user confirmation.'),
-  merge_ids: z.array(z.array(z.number().int()).min(2)).optional()
-    .describe('For dedup: [[keepId, removeId1, removeId2], ...] — first ID in each group is kept'),
-  retain_days: z.number().int().min(7).max(365).optional()
+  merge_ids: z.preprocess(
+    (v) => Array.isArray(v) ? v.map(g => Array.isArray(g) ? g.map(x => typeof x === 'string' ? parseInt(x, 10) : x) : g) : v,
+    z.array(z.array(z.number().int()).min(2))
+  ).optional().describe('For dedup: [[keepId, removeId1, removeId2], ...] — first ID in each group is kept'),
+  retain_days: coerceInt.pipe(z.number().int().min(7).max(365)).optional()
     .describe('For purge_stale: keep observations newer than N days (default 30)'),
   project: z.string().optional().describe('Filter by project'),
 };

package/utils.mjs

@@ -33,8 +33,11 @@ export const DEFAULT_DECAY_HALF_LIFE_MS = 14 * 86400000;
  */
 export function jaccardSimilarity(a, b) {
   if (!a || !b) return 0;
-  const setA = new Set(a.toLowerCase().split(/\s+/));
-  const setB = new Set(b.toLowerCase().split(/\s+/));
+  // Strip trailing punctuation from tokens to match MinHash normalization
+  // (prevents "server.rs," ≠ "server.rs" dedup failures)
+  const norm = s => s.toLowerCase().split(/\s+/).map(t => t.replace(/[,;:!?]+$/, ''));
+  const setA = new Set(norm(a));
+  const setB = new Set(norm(b));
   let intersection = 0;
   for (const w of setA) { if (setB.has(w)) intersection++; }
   const union = setA.size + setB.size - intersection;
@@ -109,12 +112,26 @@ export function scrubSecrets(text) {
 // ─── Token Estimation ─────────────────────────────────────────────────────
 
 /**
- * Estimate token count for a string using the ~4 chars/token heuristic.
+ * Estimate token count for a string.
+ * Uses ~4 chars/token for ASCII, ~1.5 chars/token for CJK characters.
  * @param {string} text Input text
  * @returns {number} Estimated token count (minimum 1)
  */
 export function estimateTokens(text) {
-  return Math.ceil(((text || '').length || 1) / 4);
+  const s = text || '';
+  if (!s) return 1;
+  // Count CJK characters (each ~1 token) vs ASCII (~4 chars/token)
+  let cjkCount = 0;
+  for (let i = 0; i < s.length; i++) {
+    const c = s.charCodeAt(i);
+    if ((c >= 0x4e00 && c <= 0x9fff) || (c >= 0x3400 && c <= 0x4dbf) ||
+        (c >= 0x3000 && c <= 0x303f) || (c >= 0xff00 && c <= 0xffef) ||
+        (c >= 0xac00 && c <= 0xd7af)) {
+      cjkCount++;
+    }
+  }
+  const asciiLen = s.length - cjkCount;
+  return Math.max(1, Math.ceil(asciiLen / 4) + Math.ceil(cjkCount / 1.5));
 }
 
 // ─── MinHash Signatures ──────────────────────────────────────────────────
@@ -236,32 +253,22 @@ const SYNONYM_PAIRS = [
   ['prod', 'production'],
   ['async', 'asynchronous'],
   ['sync', 'synchronous'],
-  // Semantic equivalents — bridges terms users type interchangeably
+  // Semantic equivalents — precise synonyms only (overly broad bridges removed)
   ['login', 'signin'],
-  ['login', 'auth'],
-  ['signin', 'auth'],
   ['bug', 'error'],
-  ['bug', 'issue'],
   ['bug', 'defect'],
   ['crash', 'panic'],
   ['crash', 'segfault'],
   ['slow', 'latency'],
-  ['slow', 'perf'],
   ['remove', 'delete'],
   ['setup', 'install'],
-  ['setup', 'config'],
   ['deploy', 'release'],
   ['deploy', 'publish'],
   ['refactor', 'restructure'],
-  ['refactor', 'cleanup'],
   ['test', 'spec'],
-  ['api', 'endpoint'],
-  ['api', 'route'],
   ['cache', 'caching'],
   ['cache', 'memoize'],
   ['optimize', 'optimization'],
-  ['optimize', 'performance'],
-  ['speed', 'performance'],
   ['fix', 'bugfix'],
   ['fix', 'patch'],
   ['debug', 'debugging'],
@@ -357,11 +364,11 @@ function expandToken(token) {
 export function sanitizeFtsQuery(query) {
   if (!query) return null;
   const cleaned = query
-    .replace(/[{}()[\]^~*:"]/g, ' ')
+    .replace(/[{}()[\]^~*:"\\]/g, ' ')
     .replace(/(^|\s)-/g, '$1')
     .trim();
   if (!cleaned) return null;
-  const tokens = cleaned.split(/\s+/).filter(t => t && !/^-+$/.test(t) && !FTS5_KEYWORDS.has(t.toUpperCase()));
+  const tokens = cleaned.split(/\s+/).filter(t => t && !/^-+$/.test(t) && !FTS5_KEYWORDS.has(t.toUpperCase()) && !/^NEAR\/\d+$/i.test(t));
   if (tokens.length === 0) return null;
   // Replace single CJK character tokens with bigrams for better phrase matching.
   // Individual CJK chars ("系","统") are too noisy; bigrams ("系统") capture compound words.
@@ -370,14 +377,15 @@ export function sanitizeFtsQuery(query) {
   const hasBigrams = bigramSet.size > 0;
   const finalTokens = [];
   const seen = new Set();
+  const rawTokensSeen = new Set(); // track raw tokens to prevent bigram duplicates
   for (const t of tokens) {
     // Skip single CJK characters when we have bigrams — they're subsumed by bigram tokens
     if (hasBigrams && /^[\u4e00-\u9fff\u3400-\u4dbf]$/.test(t)) continue;
     const expanded = expandToken(t);
-    if (!seen.has(expanded)) { seen.add(expanded); finalTokens.push(expanded); }
+    if (!seen.has(expanded)) { seen.add(expanded); rawTokensSeen.add(t); finalTokens.push(expanded); }
   }
   for (const bg of bigramSet) {
-    if (!seen.has(bg)) { seen.add(bg); finalTokens.push(bg); }
+    if (!seen.has(bg) && !rawTokensSeen.has(bg)) { seen.add(bg); finalTokens.push(bg); }
   }
   if (finalTokens.length === 0) return null;
   // FTS5 requires explicit AND after parenthesized OR groups
@@ -450,12 +458,10 @@ export function computeRuleImportance(episode) {
     if (files.some(f => /\.config\.|tsconfig|Dockerfile|docker-compose|package\.json|\.yml$|\.yaml$/i.test(basename(f))) && importance < 2) importance = 2;
   }
 
-  // Tool diversity: Edit + Bash + another tool = complete dev cycle
-  if (toolTypes.size >= 3 && toolTypes.has('Edit') && importance < 2) importance = 2;
   // Debug cycle: error followed by edit = active debugging
   if (hasErrorThenEdit && importance < 2) importance = 2;
-  // Broad change: many files touched
-  if ((episode.files || []).length >= 5 && importance < 2) importance = 2;
+  // Broad change: many files touched (8+ indicates significant scope)
+  if ((episode.files || []).length >= 8 && importance < 2) importance = 2;
 
   return importance;
 }
@@ -466,16 +472,65 @@ export function computeRuleImportance(episode) {
  * @param {string} text Input text containing CJK characters
  * @returns {string} Space-separated bigrams
  */
+// Common CJK compound words (2-4 chars) — dictionary-first tokenization.
+// When a compound word is found, it's emitted as a whole token instead of being
+// split into overlapping bigrams. This dramatically reduces noise:
+// "数据库" → "数据库" (1 token) instead of "数据 据库" (2 noisy tokens)
+const CJK_COMPOUNDS = new Set([
+  // tech/programming
+  '数据库', '数据', '接口', '函数', '变量', '组件', '模块', '配置', '框架', '部署',
+  '测试', '调试', '编译', '打包', '构建', '缓存', '索引', '迁移', '回滚', '权限',
+  '认证', '授权', '加密', '解密', '序列', '并发', '异步', '同步', '线程', '进程',
+  '容器', '集群', '服务器', '中间件', '网关', '负载', '监控', '日志', '告警',
+  '前端', '后端', '全栈', '响应式', '路由', '状态', '渲染', '样式', '布局',
+  // actions
+  '修复', '重构', '优化', '升级', '安装', '卸载', '导入', '导出', '上传', '下载',
+  '提交', '推送', '合并', '发布', '上线', '回退', '审查', '审核', '评审',
+  // errors/issues
+  '报错', '崩溃', '泄露', '溢出', '死锁', '超时', '中断', '异常', '故障',
+  // architecture
+  '架构', '设计', '方案', '规划', '文档', '注释', '版本', '分支', '依赖',
+  '性能', '安全', '漏洞', '补丁',
+]);
+
+// Sort by length descending for greedy matching
+const CJK_SORTED = [...CJK_COMPOUNDS].sort((a, b) => b.length - a.length);
+
+/**
+ * Generate search tokens from CJK text using dictionary-first tokenization.
+ * Compound words are emitted whole; remaining chars use bigram fallback.
+ * "修复了数据库崩溃" → "修复 数据库 崩溃" (3 clean tokens)
+ * vs old bigram: "修复 复了 了数 数据 据库 库崩 崩溃" (7 noisy tokens)
+ * @param {string} text Input text containing CJK characters
+ * @returns {string} Space-separated tokens
+ */
 export function cjkBigrams(text) {
   if (!text) return '';
   const runs = text.match(/[\u4e00-\u9fff\u3400-\u4dbf]{2,}/g) || [];
-  const bigrams = [];
+  const tokens = [];
   for (const run of runs) {
-    for (let i = 0; i < run.length - 1; i++) {
-      bigrams.push(run[i] + run[i + 1]);
+    let i = 0;
+    while (i < run.length) {
+      let matched = false;
+      // Greedy dictionary match (longest first)
+      for (const word of CJK_SORTED) {
+        if (i + word.length <= run.length && run.slice(i, i + word.length) === word) {
+          tokens.push(word);
+          i += word.length;
+          matched = true;
+          break;
+        }
+      }
+      if (!matched) {
+        // Fallback: bigram for unknown compound
+        if (i + 1 < run.length) {
+          tokens.push(run[i] + run[i + 1]);
+        }
+        i++;
+      }
     }
   }
-  return bigrams.join(' ');
+  return [...new Set(tokens)].join(' ');
 }
 
 // ─── Project Inference ───────────────────────────────────────────────────────
@@ -506,8 +561,8 @@ export function inferProject() {
  */
 export function detectBashSignificance(input, response) {
   const cmd = (input.command || '').toLowerCase();
-  const isError = /\berror\b|fail(ed|ure)?|exception|panic|traceback|errno|enoent|command not found/i.test(response)
-    && response.length > 30;
+  const isError = /\berror\b|\bERR!|fail(ed|ure)?|exception|panic|traceback|errno|enoent|command not found/i.test(response)
+    && response.length > 15;
   const isTest = /\b(test|jest|pytest|vitest|mocha|spec|cypress|playwright)\b/i.test(cmd);
   const isBuild = /\b(build|compile|tsc|webpack|vite|rollup|esbuild|make|cargo)\b/i.test(cmd);
   const isGit = /\bgit\s+(commit|merge|rebase|cherry-pick|push)\b/i.test(cmd);
@@ -573,7 +628,9 @@ export function extractFilePaths(input) {
     if (match) {
       for (const m of match) {
         const p = m.trim();
-        if (!p.startsWith('/dev/') && !p.startsWith('/proc/') && !p.startsWith('/tmp/')) {
+        if (!p.startsWith('/dev/') && !p.startsWith('/proc/') && !p.startsWith('/tmp/')
+          // Skip single-component paths like /exit, /clear — likely slash commands, not files
+          && (p.indexOf('/', 1) !== -1 || /\.\w+$/.test(p))) {
           paths.push(p);
         }
       }