claude-mem-lite 2.12.3 → 2.14.0

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
   "plugins": [
     {
       "name": "claude-mem-lite",
-      "version": "2.12.3",
+      "version": "2.14.0",
       "source": "./",
       "description": "Lightweight persistent memory system for Claude Code — FTS5 search, episode batching, error-triggered recall"
     }

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-mem-lite",
-  "version": "2.12.3",
+  "version": "2.14.0",
   "description": "Lightweight persistent memory system for Claude Code — FTS5 search, episode batching, error-triggered recall",
   "author": {
     "name": "sdsrss"

package/README.md

@@ -52,7 +52,7 @@ The original sends **everything to the LLM and hopes it filters well**. claude-m
 
 ## Features
 
-- **Automatic capture** -- Hooks into Claude Code lifecycle (PostToolUse, PreToolUse, SessionStart, Stop, UserPromptSubmit) to record observations without manual effort
+- **Automatic capture** -- Hooks into Claude Code lifecycle (PostToolUse, SessionStart, Stop, UserPromptSubmit) to record observations without manual effort
 - **FTS5 search** -- BM25-ranked full-text search across observations, session summaries, and user prompts with importance weighting
 - **Timeline browsing** -- Navigate observations chronologically with anchor-based context windows
 - **Episode batching** -- Groups related file operations into coherent episodes before LLM encoding
@@ -79,20 +79,19 @@ The original sends **everything to the LLM and hopes it filters well**. claude-m
 - **Atomic writes** -- All file writes (episodes, CLAUDE.md) use write-to-tmp + rename to prevent corruption on crash
 - **Robust locking** -- PID-aware lock files with automatic stale/orphan cleanup (>30s timeout or dead PID)
 - **Stale session cleanup** -- Sessions active for >24h are automatically marked as abandoned on next start
-- **Intelligent dispatch** -- 3-tier progressive dispatch system automatically recommends the right skill or agent for the current task, triggered on SessionStart, UserPromptSubmit, and PreToolUse
-- **Resource registry** -- Indexes installed skills and agents with FTS5 search, composite scoring, and invocation tracking
+- **Resource registry** -- Indexes installed skills and agents with FTS5 search, composite scoring, and invocation tracking; searchable via `mem_registry` MCP tool
 - **Unified resource discovery** -- Shared filesystem traversal layer (`resource-discovery.mjs`) used by both runtime scanner and offline indexer, supporting flat directories, plugin nesting, and loose `.md` files
-- **Closed-loop feedback** -- Tracks whether recommendations were adopted and whether sessions succeeded, improving future dispatch quality
-- **Bilingual intent recognition** -- Understands user intent in both English and Chinese (15+ EN + 12+ CN intent categories)
-- **Domain synonym expansion** -- Dispatch queries expand to domain synonyms (e.g., "fix" → debug, bugfix, troubleshoot, diagnose, repair)
-- **DB-persisted cooldown** -- 5-minute cross-session cooldown and per-session dedup prevent repeated recommendations
+- **Domain synonym expansion** -- Registry search queries expand to domain synonyms (e.g., "fix" → debug, bugfix, troubleshoot, diagnose, repair)
 - **Dual LLM mode** -- Auto-detects `ANTHROPIC_API_KEY` for direct API calls; falls back to `claude -p` CLI when no key is available
-- **Haiku circuit breaker** -- After 3 consecutive LLM failures, disables Haiku dispatch for 5 minutes to prevent cascading latency
-- **Negation-aware intent** -- Handles complex prompts like "don't test, just fix the bug" — correctly excludes negated intents even in mixed English/Chinese input
+- **Lesson-learned indexing** -- `lesson_learned` field indexed in FTS5 with weight 8, making past debugging insights directly searchable
+- **Cross-source normalization** -- `mem_search` normalizes scores across observations, sessions, and prompts before merging, preventing any source from dominating results
+- **Exponential recency decay** -- Type-differentiated half-lives (decisions: 90d, discoveries: 60d, bugfixes: 14d, changes: 7d) consistently applied in all ranking paths
+- **Prompt-time memory injection** -- UserPromptSubmit hook automatically searches and injects relevant past observations with recency and importance weighting
+- **Dual injection dedup** -- `user-prompt-search.js` and `handleUserPrompt` coordinate via temp file to prevent duplicate memory injection
 - **Configurable LLM model** -- Switch between Haiku (fast/cheap) and Sonnet (deeper analysis) via `CLAUDE_MEM_MODEL` env var
 - **DB auto-recovery** -- Detects and cleans corrupted WAL/SHM files on startup; periodic WAL checkpoints prevent unbounded growth
 - **Schema auto-migration** -- Idempotent `ALTER TABLE` migrations run on every startup, safely adding new columns and indexes without data loss
-- **Exploration bonus** -- New resources in the registry get a fair chance in composite ranking; zombie resources (high recommend, zero adopt) are penalized
+- **Exploration bonus** -- New resources in the registry get a fair chance in composite ranking; zombie resources (high recommend, zero adopt) are penalized in scoring
 - **LLM concurrency control** -- File-based semaphore limits background workers to 2 concurrent LLM calls, preventing resource contention
 - **stdin overflow protection** -- Hook input truncated at 256KB with regex-based action salvage for oversized tool outputs
 - **Cross-session handoff** -- Captures session state (request, completed work, next steps, key files) on `/clear` or `/exit`, then injects context when the next session detects continuation intent via explicit keywords or FTS5 term overlap
@@ -144,8 +143,8 @@ Source files stay in the cloned repo. Update via `git pull && node install.mjs i
 ### What happens during installation
 
 1. **Install dependencies** -- `npm install --omit=dev` (compiles native `better-sqlite3`)
-2. **Register MCP server** -- `mem` server with 7 tools (search, timeline, get, save, stats, delete, compress)
-3. **Configure hooks** -- `PostToolUse`, `PreToolUse`, `SessionStart`, `Stop`, `UserPromptSubmit` lifecycle hooks
+2. **Register MCP server** -- `mem` server with 9 tools (search, timeline, get, save, stats, delete, compress, maintain, registry)
+3. **Configure hooks** -- `PostToolUse`, `SessionStart`, `Stop`, `UserPromptSubmit` lifecycle hooks
 4. **Create data directory** -- `~/.claude-mem-lite/` (hidden) for database, runtime, and managed resource files
 5. **Auto-migrate** -- If `~/.claude-mem/` (original claude-mem) or `~/claude-mem-lite/` (pre-v0.5 unhidden) exists, migrates database and runtime files to `~/.claude-mem-lite/`, preserving the original untouched
 6. **Initialize database** -- SQLite with WAL mode, FTS5 indexes created on first server start
@@ -203,6 +202,8 @@ rm -rf ~/claude-mem-lite/   # pre-v0.5 unhidden (if not auto-moved)
 | `mem_stats` | View statistics: counts, type distribution, top projects, daily activity. |
 | `mem_delete` | Delete observations by ID with preview/confirm workflow. FTS5 cleanup is automatic. |
 | `mem_compress` | Compress old low-value observations into weekly summaries to reduce noise. |
+| `mem_maintain` | Memory maintenance: scan for duplicates/stale/broken items, then execute cleanup/dedup operations. |
+| `mem_registry` | Manage resource registry: search for skills/agents by need, list resources, view stats, import/remove tools, reindex. |
 
 ### Skill Commands (in Claude Code chat)
 
@@ -231,13 +232,15 @@ Five core tables with FTS5 virtual tables for search:
 ```
 id, memory_session_id, project, type, title, subtitle,
 text, narrative, concepts, facts, files_read, files_modified,
-importance, related_ids, created_at, created_at_epoch
+importance, related_ids, created_at, created_at_epoch,
+lesson_learned, minhash_sig, access_count, compressed_into, search_aliases
 ```
 
 **session_summaries** -- LLM-generated session summaries
 ```
 id, memory_session_id, project, request, investigated,
-learned, completed, next_steps, files_read, files_edited, notes
+learned, completed, next_steps, files_read, files_edited, notes,
+remaining_items, lessons, key_decisions
 ```
 
 **sdk_sessions** -- Session tracking
@@ -257,7 +260,7 @@ project, type, session_id, working_on, completed, unfinished,
 key_files, key_decisions, match_keywords, created_at_epoch
 ```
 
-FTS5 indexes: `observations_fts`, `session_summaries_fts`, `user_prompts_fts`
+FTS5 indexes: `observations_fts` (title, subtitle, narrative, text, facts, concepts, lesson_learned), `session_summaries_fts`, `user_prompts_fts`
 
 ## How It Works
 
@@ -270,7 +273,6 @@ SessionStart
   -> Clean orphaned/stale lock files
   -> Query recent observations (24h)
   -> Inject context into CLAUDE.md + stdout
-  -> Dispatch: recommend skill/agent based on user prompt (Tier 0→1→2→3)
 
 PostToolUse (every tool execution)
   -> Bash pre-filter skips noise in ~5ms (Read paths tracked to reads file)
@@ -282,75 +284,36 @@ PostToolUse (every tool execution)
   -> Spawn LLM episode worker for significant episodes
   -> Error-triggered recall: search memory for related past fixes
 
-PreToolUse (before tool execution)
-  -> Dispatch: recommend skill/agent based on current action context (Tier 0→1→2)
-
-UserPromptSubmit
-  -> Capture user prompt text to user_prompts table
-  -> Increment session prompt counter
-  -> Handoff: detect continuation intent → inject previous session context
-  -> Dispatch: recommend skill/agent based on user's actual prompt (Tier 0→1→2)
-  -> Primary dispatch point — user intent is clearest here
+UserPromptSubmit (two parallel paths)
+  -> [user-prompt-search.js] Auto-search memory via FTS5 + active file context
+  -> [user-prompt-search.js] Inject relevant past observations with recency/importance weighting
+  -> [user-prompt-search.js] Write injected IDs to temp file for dedup
+  -> [hook.mjs handleUserPrompt] Capture user prompt text to user_prompts table
+  -> [hook.mjs handleUserPrompt] Increment session prompt counter
+  -> [hook.mjs handleUserPrompt] Handoff: detect continuation intent → inject previous session context
+  -> [hook.mjs handleUserPrompt] Semantic memory injection (hook-memory.mjs), deduped via temp file
 
 Stop
   -> Flush final episode buffer
   -> Save handoff snapshot (on /exit)
-  -> Collect dispatch feedback: adoption detection + outcome scoring
   -> Mark session completed
   -> Spawn LLM summary worker (poll-based wait)
 ```
 
-### Intelligent Dispatch
+### Resource Registry
 
-The dispatch system proactively recommends skills and agents during coding sessions via a 3-tier progressive architecture:
+The resource registry (`registry.mjs`, `registry-retriever.mjs`) indexes installed skills and agents into a searchable FTS5 database. Unlike the previous proactive dispatch system, the registry is now on-demand — Claude searches it via the `mem_registry` MCP tool when it needs to discover relevant skills or agents.
 
 ```
-Tier 0: Fast Filter (<1ms)
-  -> Skip read-only tools (Read, Glob, Grep, LSP...)
-  -> Skip simple Bash queries (ls, cat, git status...)
-  -> Skip when Claude already chose a Skill or Task agent
-  -> Skip MCP-internal tools
-
-Tier 1: Context Signal Extraction (<1ms)
-  -> Intent: extract from user prompt (test, fix, deploy, review...)
-  -> Tech stack: infer from recent file extensions (.ts → typescript)
-  -> Action: infer from tool name (Edit → edit, Bash+jest → test)
-  -> Error domain: classify errors (type-error, test-fail, build-fail...)
-
-Tier 2: FTS5 Retrieval (<5ms)
-  -> Expand signals with domain synonyms (15+ EN, 12+ CN categories)
-  -> BM25-ranked search over resource registry
-  -> Composite scoring: BM25 (40%) + repo stars (15%) + success rate (15%) + adoption rate (10%)
-
-Tier 3: Haiku Semantic Dispatch (~500ms, SessionStart only)
-  -> Activated when FTS5 confidence is low or top results are ambiguous
-  -> LLM generates semantic search query for refined retrieval
-  -> Disabled for PreToolUse (2s hook timeout insufficient)
+Registry pipeline:
+  -> registry-scanner.mjs discovers skills/agents on filesystem
+  -> resource-discovery.mjs handles flat dirs, plugin nesting, loose .md files
+  -> registry-indexer.mjs indexes content into FTS5 with metadata
+  -> registry-retriever.mjs provides BM25-ranked search with synonym expansion
+  -> mem_registry MCP tool exposes search/list/stats/import/remove/reindex actions
 ```
 
-**Dispatch triggers:**
-
-| Hook | Budget | Tiers | Use case |
-|------|--------|-------|----------|
-| SessionStart | 10s | 0→1→2→3 | Analyze previous session's next_steps, suggest skill/agent upfront |
-| UserPromptSubmit | 2s | 0→1→2 | Primary dispatch point — user's actual prompt has clearest intent |
-| PreToolUse | 2s | 0→1→2 | React to current action context in real-time |
-
-**Feedback loop (Stop hook):**
-
-At session end, the system reviews all recommendations made during the session:
-- **Adoption detection** -- Did Claude actually use the recommended skill (`Skill` tool) or agent (`Task` tool)?
-- **Outcome detection** -- Was the session successful (edits without errors), partial (errors then fixes), or failed?
-- **Score calculation** -- Adopted + success = 1.0, adopted + partial = 0.5, adopted + failure = 0.2
-- Stats feed back into composite scoring, improving future dispatch quality over time
-
-**Injection templates:**
-
-| Resource type | Location | Template |
-|---------------|----------|----------|
-| Skill | `~/.claude/skills/` (native) | Short hint: use `/skill <name>` |
-| Skill | managed directory | Full skill content injected (up to 3KB) |
-| Agent | any | Agent definition injected for `Task` tool delegation |
+Composite scoring for search results: BM25 relevance (40%) + repo stars (15%) + success rate (15%) + adoption rate (10%) + freshness (10%) + exploration bonus (10%). Domain filtering ensures platform-specific resources (iOS, Go, Rust) only surface for matching projects.
 
 ### Episode Encoding
 
@@ -438,16 +401,15 @@ claude-mem-lite/
   hook-handoff.mjs     # Cross-session handoff: state extraction, intent detection, injection
   hook-context.mjs     # CLAUDE.md context injection and token budgeting
   hook-episode.mjs     # Episode buffer management: atomic writes, pending entry merging
+  hook-memory.mjs      # Semantic memory injection on user prompt
   hook-semaphore.mjs   # LLM concurrency control: file-based semaphore for background workers
   schema.mjs           # Database schema: single source of truth for tables, migrations, FTS5
   tool-schemas.mjs     # Shared Zod schemas for MCP tool validation
-  utils.mjs            # Shared utilities: FTS5 query building, MinHash dedup, secret scrubbing
-  # Intelligent dispatch
-  dispatch.mjs         # 3-tier dispatch orchestration: fast filter, context signals, FTS5, Haiku
-  dispatch-inject.mjs  # Injection template rendering for skill/agent recommendations
-  dispatch-feedback.mjs # Closed-loop feedback: adoption detection, outcome tracking
+  utils.mjs            # Shared utilities: FTS5 query building, BM25 weight constants, MinHash dedup, secret scrubbing
+  # Resource registry
   registry.mjs         # Resource registry DB: schema, CRUD, FTS5, invocation tracking
   registry-retriever.mjs # FTS5 retrieval with synonym expansion and composite scoring
+  registry-indexer.mjs # Resource indexing pipeline
   registry-scanner.mjs # Filesystem scanner: reads content + hashes, delegates discovery
   resource-discovery.mjs # Shared discovery layer: flat dirs, plugin nesting, loose .md files
   haiku-client.mjs     # Unified Haiku LLM wrapper: direct API or CLI fallback
@@ -458,6 +420,7 @@ claude-mem-lite/
   scripts/
     setup.sh           # Setup hook: npm install + migration (hidden dir + old dir)
     post-tool-use.sh   # Bash pre-filter: skips noise in ~5ms, tracks Read paths
+    user-prompt-search.js # UserPromptSubmit hook: auto-search memory on user prompts
     convert-commands.mjs # Converts command .md → SKILL.md in managed plugins
     index-managed.mjs  # Offline indexer for managed resources
   # Test & benchmark (dev only)
@@ -495,7 +458,7 @@ npm run benchmark:gate    # CI gate: fails if metrics regress beyond 5% toleranc
 | Variable | Description | Default |
 |----------|-------------|---------|
 | `CLAUDE_MEM_DIR` | Custom data directory. All databases, runtime files, and managed resources are stored here. | `~/.claude-mem-lite/` |
-| `CLAUDE_MEM_MODEL` | LLM model for background calls (episode extraction, session summaries, dispatch). Accepts `haiku` or `sonnet`. | `haiku` |
+| `CLAUDE_MEM_MODEL` | LLM model for background calls (episode extraction, session summaries). Accepts `haiku` or `sonnet`. | `haiku` |
 | `CLAUDE_MEM_DEBUG` | Enable debug logging (`1` to enable). | _(disabled)_ |
 
 ## License

package/cli.mjs

@@ -1,4 +1,12 @@
 #!/usr/bin/env node
-import { main } from './install.mjs';
+const CLI_COMMANDS = new Set(['search', 'recent', 'recall', 'get', 'timeline', 'save', 'stats', 'context', 'help']);
 
-await main(process.argv.slice(2));
+const cmd = process.argv[2];
+
+if (CLI_COMMANDS.has(cmd)) {
+  const { run } = await import('./mem-cli.mjs');
+  await run(process.argv.slice(2));
+} else {
+  const { main } = await import('./install.mjs');
+  await main(process.argv.slice(2));
+}

package/commands/mem.md

@@ -2,39 +2,36 @@
 description: Search and manage project memory (observations, sessions, prompts)
 ---
 
-# Memory Skill
+# Memory
 
 Search and browse your project memory efficiently.
 
-## Commands
+## Quick Commands
 
-- `/mem search <query>` — FTS5 full-text search across all memories
-- `/mem recent [n]` — Show recent N observations (default 10)
+- `/mem search <query>` — Search all memories (FTS5 full-text search)
+- `/mem recent [n]` — Show recent N observations (default 5)
+- `/mem recall <file>` — History for a file before editing
+- `/mem timeline <id>` — Browse timeline around an observation
 - `/mem save <text>` — Save a manual memory/note
 - `/mem stats` — Show memory statistics
-- `/mem timeline <query>` — Browse timeline around a matching observation
 - `/mem cleanup` — Scan and interactively purge stale data
 - `/mem cleanup [N]d` — Purge stale data older than N days (e.g. `cleanup 60d`)
 - `/mem cleanup keep [N]d` — Purge stale data but retain last N days (e.g. `cleanup keep 14d`)
 
-## Efficient Search Workflow (3 steps, saves 10x tokens)
-
-1. **Search** → `mem_search(query="...")` → get compact ID index
-2. **Browse** → `mem_timeline(anchor=ID)` → see surrounding context
-3. **Detail** → `mem_get(ids=[...])` → get full content for specific IDs
-
 ## Instructions
 
 When the user invokes `/mem`, parse their intent:
 
-- `/mem search <query>` → call `mem_search` with the query
-- `/mem recent` or `/mem recent 20` → call `mem_search` with no query, limit=N
-- `/mem save <text>` → call `mem_save` with the text as content
-- `/mem stats` → call `mem_stats`
-- `/mem timeline <query>` → call `mem_timeline` with the query
+- `/mem search <query>` → run `claude-mem-lite search <query>` via Bash
+- `/mem recent` or `/mem recent 20` → run `claude-mem-lite recent [N]` via Bash
+- `/mem recall <file>` → run `claude-mem-lite recall <file>` via Bash
+- `/mem timeline <id>` → run `claude-mem-lite timeline --anchor <id>` via Bash
+- `/mem save <text>` → call `mem_save` MCP tool with the text as content
+- `/mem stats` → run `claude-mem-lite stats` via Bash
+- `/mem get <ids>` → run `claude-mem-lite get <ids>` via Bash
 - `/mem cleanup` → run `mem_maintain(action="scan")`, report pending purge count and stale items to user, ask for confirmation, then run `mem_maintain(action="execute", operations=["purge_stale"])` if confirmed
 - `/mem cleanup Nd` (e.g. `60d`) → same as above but use `retain_days=N` to only purge items older than N days
 - `/mem cleanup keep Nd` (e.g. `keep 14d`) → same as above with `retain_days=N`
-- `/mem <query>` (no subcommand) → treat as search, call `mem_search`
+- `/mem <query>` (no subcommand) → treat as search, run `claude-mem-lite search <query>` via Bash
 
-Always use the compact index from mem_search first, then mem_get for details only when needed. This minimizes token usage.
+Use Bash commands first. For detailed data, use `claude-mem-lite get <id>` via Bash.

package/commands/recall.md

@@ -0,0 +1,9 @@
+---
+description: Recall past observations for a file before editing
+argument-hint: <file_path>
+---
+
+## File Memory
+!`claude-mem-lite recall $ARGUMENTS 2>/dev/null || echo "No history found"`
+
+Consider these past observations before making changes.

package/commands/recent.md

@@ -0,0 +1,7 @@
+---
+description: Show recent memory observations
+argument-hint: [count]
+---
+
+## Recent Observations
+!`claude-mem-lite recent $ARGUMENTS 2>/dev/null || echo "No observations"`

package/commands/search.md

@@ -0,0 +1,9 @@
+---
+description: Search memory for past bugfixes, decisions, discoveries
+argument-hint: <query>
+---
+
+## Memory Search
+!`claude-mem-lite search $ARGUMENTS 2>/dev/null || echo "No results found"`
+
+Use `claude-mem-lite get <id>` via Bash for full details on any result.

package/commands/timeline.md

@@ -0,0 +1,7 @@
+---
+description: Browse memory timeline around an observation
+argument-hint: <observation_id>
+---
+
+## Timeline
+!`claude-mem-lite timeline --anchor $ARGUMENTS 2>/dev/null || echo "Not found"`

package/hook-context.mjs

@@ -3,7 +3,7 @@
 
 import { join } from 'path';
 import { readFileSync, writeFileSync, renameSync, unlinkSync } from 'fs';
-import { estimateTokens, truncate, debugLog, debugCatch } from './utils.mjs';
+import { estimateTokens, truncate, debugLog, debugCatch, DECAY_HALF_LIFE_BY_TYPE, DEFAULT_DECAY_HALF_LIFE_MS } from './utils.mjs';
 
 /**
  * Infer the project directory from environment variables or cwd.
@@ -83,9 +83,10 @@ export function selectWithTokenBudget(db, project, budget = 2000) {
   let totalTokens = 0;
 
   // Score each candidate: value = recency * importance, cost = tokens
+  // Recency uses exponential half-life (consistent with server.mjs BM25 scoring)
   const scoredObs = obsPool.map(o => {
-    const ageDays = (now_ms - o.created_at_epoch) / 86400000;
-    const recency = 1 / (1 + ageDays);
+    const halfLifeMs = DECAY_HALF_LIFE_BY_TYPE[o.type] || DEFAULT_DECAY_HALF_LIFE_MS;
+    const recency = 1.0 + Math.exp(-0.693 * (now_ms - o.created_at_epoch) / halfLifeMs);
     const impBoost = 0.5 + 0.5 * (o.importance || 1);
     const lessonBoost = o.lesson_learned ? 1.3 : 1.0;
     const value = recency * impBoost * lessonBoost;
@@ -94,8 +95,7 @@ export function selectWithTokenBudget(db, project, budget = 2000) {
   });
 
   const scoredSess = sessPool.map(s => {
-    const ageDays = (now_ms - s.created_at_epoch) / 86400000;
-    const recency = 1 / (1 + ageDays);
+    const recency = 1.0 + Math.exp(-0.693 * (now_ms - s.created_at_epoch) / DEFAULT_DECAY_HALF_LIFE_MS);
     const value = recency * 1.5; // Session summaries slightly boosted
     const cost = estimateTokens((s.request || '') + (s.completed || '') + (s.next_steps || ''));
     return { ...s, value, cost, valueDensity: cost > 0 ? value / Math.sqrt(cost) : 0 };

package/hook-llm.mjs

@@ -6,7 +6,7 @@ import { existsSync, readFileSync, unlinkSync, readdirSync } from 'fs';
 import {
   jaccardSimilarity, truncate, clampImportance, computeRuleImportance,
   inferProject, parseJsonFromLLM,
-  computeMinHash, estimateJaccardFromMinHash, cjkBigrams, EDIT_TOOLS, debugCatch, debugLog,
+  computeMinHash, estimateJaccardFromMinHash, cjkBigrams, EDIT_TOOLS, debugCatch, debugLog, OBS_BM25,
 } from './utils.mjs';
 import { acquireLLMSlot, releaseLLMSlot } from './hook-semaphore.mjs';
 import {
@@ -139,7 +139,7 @@ function linkRelatedObservations(db, savedId, obs, episode) {
           SELECT o.id FROM observations_fts
           JOIN observations o ON observations_fts.rowid = o.id
           WHERE observations_fts MATCH ? AND o.id != ? AND o.project = ?
-          ORDER BY bm25(observations_fts, 10, 5, 5, 3, 3, 2)
+          ORDER BY ${OBS_BM25}
           LIMIT 5
         `).all(ftsQuery, newObs.id, episode.project);
         for (const m of ftsMatches) candidates.add(m.id);

package/hook-memory.mjs

@@ -1,7 +1,7 @@
 // claude-mem-lite — Semantic Memory Injection
 // Search past observations for relevant memories to inject as context at user-prompt time.
 
-import { sanitizeFtsQuery, debugCatch } from './utils.mjs';
+import { sanitizeFtsQuery, debugCatch, OBS_BM25 } from './utils.mjs';
 
 const MAX_MEMORY_INJECTIONS = 3;
 const MEMORY_LOOKBACK_MS = 60 * 86400000; // 60 days
@@ -32,7 +32,7 @@ export function searchRelevantMemories(db, userPrompt, project, excludeIds = [])
     // Phase 1: Same-project search (highest priority)
     const selectStmt = db.prepare(`
       SELECT o.id, o.type, o.title, o.importance, o.lesson_learned, o.project,
-             bm25(observations_fts, 10, 5, 5, 3, 3, 2) as relevance
+             ${OBS_BM25} as relevance
       FROM observations_fts
       JOIN observations o ON o.id = observations_fts.rowid
       WHERE observations_fts MATCH ?
@@ -40,7 +40,7 @@ export function searchRelevantMemories(db, userPrompt, project, excludeIds = [])
         AND o.importance >= 1
         AND o.created_at_epoch > ?
         AND COALESCE(o.compressed_into, 0) = 0
-      ORDER BY bm25(observations_fts, 10, 5, 5, 3, 3, 2)
+      ORDER BY ${OBS_BM25}
       LIMIT 10
     `);
     const rows = selectStmt.all(ftsQuery, project, cutoff);
@@ -51,7 +51,7 @@ export function searchRelevantMemories(db, userPrompt, project, excludeIds = [])
     try {
       crossRows = db.prepare(`
         SELECT o.id, o.type, o.title, o.importance, o.lesson_learned, o.project,
-               bm25(observations_fts, 10, 5, 5, 3, 3, 2) as relevance
+               ${OBS_BM25} as relevance
         FROM observations_fts
         JOIN observations o ON o.id = observations_fts.rowid
         WHERE observations_fts MATCH ?
@@ -60,7 +60,7 @@ export function searchRelevantMemories(db, userPrompt, project, excludeIds = [])
           AND o.importance >= 2
           AND o.created_at_epoch > ?
           AND COALESCE(o.compressed_into, 0) = 0
-        ORDER BY bm25(observations_fts, 10, 5, 5, 3, 3, 2)
+        ORDER BY ${OBS_BM25}
         LIMIT 5
       `).all(ftsQuery, project, cutoff);
     } catch (e) { debugCatch(e, 'crossProjectSearch'); }

package/hook-semaphore.mjs

@@ -62,7 +62,8 @@ export async function acquireLLMSlot() {
             active++;
           }
         } catch {
-          active++;
+          // Corrupt/unreadable semaphore file — treat as stale and remove
+          try { unlinkSync(fp); } catch {}
         }
       }
     } catch {}

package/hook-shared.mjs

@@ -4,10 +4,9 @@
 import { execFileSync, spawn } from 'child_process';
 import { randomUUID } from 'crypto';
 import { join } from 'path';
-import { existsSync, readFileSync, writeFileSync, appendFileSync, mkdirSync, renameSync, unlinkSync } from 'fs';
+import { existsSync, readFileSync, writeFileSync, mkdirSync, renameSync } from 'fs';
 import { inferProject, debugCatch } from './utils.mjs';
-import { ensureDb, DB_DIR, REGISTRY_DB_PATH } from './schema.mjs';
-import { ensureRegistryDb } from './registry.mjs';
+import { ensureDb, DB_DIR } from './schema.mjs';
 import { getClaudePath as getClaudePathShared, resolveModel as resolveModelShared } from './haiku-client.mjs';
 
 // ─── Constants ────────────────────────────────────────────────────────────────
@@ -24,7 +23,6 @@ export const STALE_LOCK_MS = 30000;                       // 30s
 export const DEDUP_WINDOW_MS = 5 * 60 * 1000;            // 5 min (title dedup)
 export const RELATED_OBS_WINDOW_MS = 7 * 86400000;       // 7 days
 export const FALLBACK_OBS_WINDOW_MS = RELATED_OBS_WINDOW_MS; // same window
-export const RESOURCE_RESCAN_INTERVAL_MS = 60 * 60 * 1000;    // 1 hour
 
 // Handoff system constants
 export const HANDOFF_EXPIRY_CLEAR = 6 * 3600000;                // 6 hours (covers lunch/meeting breaks)
@@ -69,20 +67,6 @@ export function openDb() {
   }
 }
 
-// ─── Registry Database (dispatch system) ─────────────────────────────────────
-let _registryDb = null;
-
-export function getRegistryDb() {
-  if (_registryDb) return _registryDb;
-  try { _registryDb = ensureRegistryDb(REGISTRY_DB_PATH); } catch (e) { debugCatch(e, 'getRegistryDb'); }
-  return _registryDb;
-}
-
-export function closeRegistryDb() {
-  if (_registryDb) try { _registryDb.close(); } catch {}
-  _registryDb = null;
-}
-
 // ─── LLM via claude CLI ─────────────────────────────────────────────────────
 
 export function callLLM(prompt, timeoutMs = 15000) {
@@ -127,75 +111,6 @@ export function spawnBackground(bgEvent, ...extraArgs) {
 
 export function sleep(ms) { return new Promise(r => setTimeout(r, ms)); }
 
-// ─── Injection Budget (per hook invocation, in-memory) ───────────────────────
-// Limits context injections within a single hook process to prevent context bloat.
-// Note: each hook event runs in a separate process, so this is per-invocation,
-// not per-session. Session-level dedup is handled by cooldown/sessionId checks.
-
-export const MAX_INJECTIONS_PER_SESSION = 3;
-let _injectionCount = 0;
-
-export function getInjectionCount() { return _injectionCount; }
-export function incrementInjection() { _injectionCount++; }
-export function resetInjectionBudget() { _injectionCount = 0; }
-export function hasInjectionBudget() { return _injectionCount < MAX_INJECTIONS_PER_SESSION; }
-
-// ─── Previous Session Context (for user-prompt dispatch enrichment) ──────────
-// Session-start caches next_steps; first user-prompt reads+clears for richer dispatch.
-
-// ─── Tool Event Tracking (for dispatch feedback) ────────────────────────────
-// PostToolUse appends feedback-relevant tool events (Skill, Task, Edit, Write, Bash errors).
-// Stop handler reads them and passes to collectFeedback for adoption/outcome detection.
-
-export function toolEventsFile() {
-  return join(RUNTIME_DIR, `tool-events-${inferProject()}.jsonl`);
-}
-
-/**
- * Append a tool event for feedback tracking.
- * Only call for feedback-relevant events (Skill, Task, Edit, Write, Bash).
- * @param {object} event { tool_name, tool_input, tool_response }
- */
-export function appendToolEvent(event) {
-  try {
-    appendFileSync(toolEventsFile(), JSON.stringify(event) + '\n');
-  } catch {}
-}
-
-/**
- * Read all tracked tool events and remove the file.
- * Uses rename→read→delete for atomicity.
- * @returns {object[]} Array of tool event objects
- */
-export function readAndClearToolEvents() {
-  const file = toolEventsFile();
-  const claimFile = file + `.claim-${process.pid}-${Date.now()}`;
-  try {
-    renameSync(file, claimFile);
-    const raw = readFileSync(claimFile, 'utf8');
-    try { unlinkSync(claimFile); } catch {}
-    return raw.trim().split('\n').filter(Boolean).map(line => {
-      try { return JSON.parse(line); } catch { return null; }
-    }).filter(Boolean);
-  } catch {
-    return [];
-  }
-}
-
-/**
- * Read tracked tool events WITHOUT clearing the file.
- * Used by dispatch to detect active suites mid-session.
- * @returns {object[]} Array of tool event objects
- */
-export function peekToolEvents() {
-  try {
-    const raw = readFileSync(toolEventsFile(), 'utf8');
-    return raw.trim().split('\n').filter(Boolean).map(line => {
-      try { return JSON.parse(line); } catch { return null; }
-    }).filter(Boolean);
-  } catch { return []; }
-}
-
 /**
  * Extract partial response from CLI error output (timeout/error recovery).
  * @param {Error} error The caught error from execFileSync

package/hook-update.mjs

@@ -194,13 +194,12 @@ export function getCurrentVersion() {
 // ── Source files to copy (must match install.mjs SOURCE_FILES) ──
 const SOURCE_FILES = [
   'server.mjs', 'server-internals.mjs', 'tool-schemas.mjs',
-  'hook.mjs', 'hook-shared.mjs', 'hook-llm.mjs', 'hook-memory.mjs',
+  'hook.mjs', 'hook-shared.mjs', 'hook-llm.mjs', 'hook-memory.mjs', 'skip-tools.mjs',
   'hook-semaphore.mjs', 'hook-episode.mjs', 'hook-context.mjs', 'hook-handoff.mjs', 'hook-update.mjs',
   'haiku-client.mjs', 'utils.mjs', 'schema.mjs', 'package.json', 'package-lock.json', 'skill.md',
   'registry.mjs', 'registry-scanner.mjs', 'registry-indexer.mjs',
   'registry-retriever.mjs', 'resource-discovery.mjs',
-  'dispatch.mjs', 'dispatch-inject.mjs', 'dispatch-feedback.mjs', 'dispatch-patterns.mjs', 'dispatch-workflow.mjs',
-  'install.mjs', 'install-metadata.mjs',
+  'install.mjs', 'install-metadata.mjs', 'mem-cli.mjs',
 ];
 const SWITCHABLE_PATHS = [...SOURCE_FILES, 'scripts', 'registry', 'node_modules'];