agent-working-memory 0.7.10 → 0.7.11

package/README.md

@@ -393,6 +393,7 @@ npm run test:locomo   # LoCoMo industry benchmark (28.2%)
 | `AWM_DISABLE_POOL_FILTER` | *(unset)* | Set to `1` to disable the candidate pool reduction (0.7.7+). Reverts recall to scoring all active candidates — slower but useful for A/B testing if a recall regression appears |
 | `AWM_DISABLE_SLIM_CACHE` | *(unset)* | Set to `1` to disable the in-memory slim cache (0.7.10+). Reverts to per-recall SQL fetch — slower but useful if cache invariants are suspected of drift |
 | `AWM_DISABLE_RERANK_SKIP` | *(unset)* | Set to `1` to disable the reranker skip on clear-winner queries (0.7.10+). Forces every recall through the cross-encoder |
+| `AWM_DISABLE_EXPANSION_CACHE` | *(unset)* | Set to `1` to disable the query expansion skip heuristic + LRU cache (0.7.11+). Forces every recall through the flan-t5-small expander |
 | `AWM_WORKSPACE` | *(unset)* | Default workspace for cross-agent recall in hive setups |
 
 ## Tech Stack
@@ -412,6 +413,10 @@ npm run test:locomo   # LoCoMo industry benchmark (28.2%)
 
 All three ML models run locally via ONNX. No external API calls for retrieval. The entire system is a single SQLite file + a Node.js process.
 
+## What's New in v0.7.11
+
+- **Query expansion skip + LRU cache** — flan-t5-small was 164ms per recall (18% of post-0.7.10 floor). Two fixes in `core/query-expander.ts`: (1) skip heuristic for long/specific queries (>50 chars OR ≥5 distinct meaningful tokens), and (2) 500-entry LRU cache for repeated queries. ~30% of typical agent recalls hit the skip; repeated recalls hit the cache. Avg savings: ~100-150ms per recall. Recall quality A/B: 8/8 top-1, 4.63/5 top-5. Disable via `AWM_DISABLE_EXPANSION_CACHE=1`.
+
 ## What's New in v0.7.10
 
 - **Recall latency 1.4s → 0.9s median (~35% on top of 0.7.9)** — two more fixes after phase-breakdown showed the slim fetch was still 310ms (Buffer→Float32Array on every recall) and the reranker was 354ms (40% of remaining cost):
@@ -482,7 +487,7 @@ See [CHANGELOG.md](CHANGELOG.md) for full details.
 
 ## Project Status
 
-AWM is in active development (v0.7.10). The core memory pipeline, consolidation system, multi-agent coordination, and MCP integration are stable and used daily in production coding workflows.
+AWM is in active development (v0.7.11). The core memory pipeline, consolidation system, multi-agent coordination, and MCP integration are stable and used daily in production coding workflows.
 
 - Core retrieval and consolidation: **stable**
 - MCP tools and Claude Code integration: **stable**

package/dist/api/routes.js

@@ -527,7 +527,7 @@ export function registerRoutes(app, deps) {
         const base = {
             status: 'ok',
             timestamp: new Date().toISOString(),
-            version: '0.7.10',
+            version: '0.7.11',
             coordination: coordEnabled,
         };
         if (coordEnabled) {

package/dist/cli.js

@@ -307,7 +307,7 @@ async function exportMemories() {
     // Collect unique agents
     const agents = [...new Set(memories.map((m) => m.agent_id))];
     const exportData = {
-        version: '0.7.10',
+        version: '0.7.11',
         exported_at: new Date().toISOString(),
         source_db: dbPath,
         agent_filter: agentFilter,

package/dist/core/query-expander.d.ts

@@ -19,6 +19,13 @@ export declare function getExpander(): Promise<Text2TextGenerationPipeline>;
  * Expand a query with related terms and synonyms.
  * Returns the original query + generated expansion terms.
  * Falls back to the original query on any error.
+ *
+ * Optimization (0.7.11+):
+ * - Skip heuristic for long/specific queries (~30% of typical agent recalls)
+ * - LRU cache for repeated queries (cache hit ≈ 0ms vs 164ms cold)
+ * - Disable both via AWM_DISABLE_EXPANSION_CACHE=1
  */
 export declare function expandQuery(originalQuery: string): Promise<string>;
+/** Clear the expansion cache (used by tests + cache invalidation if needed). */
+export declare function clearExpansionCache(): void;
 //# sourceMappingURL=query-expander.d.ts.map

package/dist/core/query-expander.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"query-expander.d.ts","sourceRoot":"","sources":["../../src/core/query-expander.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;GAUG;AAEH,OAAO,EAAY,KAAK,2BAA2B,EAAE,MAAM,2BAA2B,CAAC;AAMvF;;;GAGG;AACH,wBAAsB,WAAW,IAAI,OAAO,CAAC,2BAA2B,CAAC,CAaxE;AAED;;;;GAIG;AACH,wBAAsB,WAAW,CAAC,aAAa,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAoBxE"}
+{"version":3,"file":"query-expander.d.ts","sourceRoot":"","sources":["../../src/core/query-expander.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;GAUG;AAEH,OAAO,EAAY,KAAK,2BAA2B,EAAE,MAAM,2BAA2B,CAAC;AAMvF;;;GAGG;AACH,wBAAsB,WAAW,IAAI,OAAO,CAAC,2BAA2B,CAAC,CAaxE;AA6BD;;;;;;;;;GASG;AACH,wBAAsB,WAAW,CAAC,aAAa,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CA8CxE;AAED,gFAAgF;AAChF,wBAAgB,mBAAmB,IAAI,IAAI,CAE1C"}

package/dist/core/query-expander.js

@@ -33,12 +33,58 @@ export async function getExpander() {
     });
     return initPromise;
 }
+/**
+ * LRU cache of normalized-query → expanded-query mappings.
+ * Map preserves insertion order — re-set on hit moves entry to the end (most-recent),
+ * delete-first when over capacity drops the least-recent. ~500 entries × ~200 chars
+ * each ≈ 100KB, negligible memory cost.
+ *
+ * Why: phase-breakdown spike (2026-05-08) showed expandQuery at ~164ms per call.
+ * Agent recall patterns repeat — cache hits are common.
+ */
+const expansionCache = new Map();
+const EXPANSION_CACHE_LIMIT = 500;
+/**
+ * Heuristic: skip expansion when the query is already specific.
+ * Long, multi-token queries don't benefit from synonyms — they're already
+ * narrow enough that flan-t5's general-vocabulary expansion adds noise more
+ * than recall. Exact thresholds are conservative; false-skips would only
+ * affect candidates that BM25 catches anyway.
+ */
+function shouldSkipExpansion(normalized) {
+    if (normalized.length === 0)
+        return true;
+    if (normalized.length > 50)
+        return true;
+    // ≥5 distinct meaningful tokens = already specific
+    const tokens = new Set(normalized.split(/\s+/).filter(t => t.length > 2));
+    return tokens.size >= 5;
+}
 /**
  * Expand a query with related terms and synonyms.
  * Returns the original query + generated expansion terms.
  * Falls back to the original query on any error.
+ *
+ * Optimization (0.7.11+):
+ * - Skip heuristic for long/specific queries (~30% of typical agent recalls)
+ * - LRU cache for repeated queries (cache hit ≈ 0ms vs 164ms cold)
+ * - Disable both via AWM_DISABLE_EXPANSION_CACHE=1
  */
 export async function expandQuery(originalQuery) {
+    const normalized = originalQuery.toLowerCase().trim();
+    const optimizationsEnabled = process.env.AWM_DISABLE_EXPANSION_CACHE !== '1';
+    if (optimizationsEnabled) {
+        if (shouldSkipExpansion(normalized)) {
+            return originalQuery;
+        }
+        const cached = expansionCache.get(normalized);
+        if (cached !== undefined) {
+            // Move to end (most-recent) for LRU semantics
+            expansionCache.delete(normalized);
+            expansionCache.set(normalized, cached);
+            return cached;
+        }
+    }
     try {
         const expander = await getExpander();
         const prompt = `Expand this search query with synonyms and related terms. Only output the additional terms, not the original query. Query: ${originalQuery}. Additional terms:`;
@@ -48,13 +94,26 @@ export async function expandQuery(originalQuery) {
         });
         const expanded = Array.isArray(result) ? result[0]?.generated_text ?? '' : '';
         const cleanExpanded = expanded.trim();
-        if (cleanExpanded && cleanExpanded.length > 2) {
-            return `${originalQuery} ${cleanExpanded}`;
+        const finalQuery = cleanExpanded && cleanExpanded.length > 2
+            ? `${originalQuery} ${cleanExpanded}`
+            : originalQuery;
+        // Cache the result (LRU eviction when over capacity)
+        if (optimizationsEnabled) {
+            if (expansionCache.size >= EXPANSION_CACHE_LIMIT) {
+                const oldestKey = expansionCache.keys().next().value;
+                if (oldestKey !== undefined)
+                    expansionCache.delete(oldestKey);
+            }
+            expansionCache.set(normalized, finalQuery);
         }
-        return originalQuery;
+        return finalQuery;
     }
     catch {
         return originalQuery;
     }
 }
+/** Clear the expansion cache (used by tests + cache invalidation if needed). */
+export function clearExpansionCache() {
+    expansionCache.clear();
+}
 //# sourceMappingURL=query-expander.js.map

package/dist/core/query-expander.js.map

@@ -1 +1 @@
-{"version":3,"file":"query-expander.js","sourceRoot":"","sources":["../../src/core/query-expander.ts"],"names":[],"mappings":"AAAA,gDAAgD;AAChD,sCAAsC;AACtC;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,QAAQ,EAAoC,MAAM,2BAA2B,CAAC;AAEvF,MAAM,QAAQ,GAAG,sBAAsB,CAAC;AACxC,IAAI,QAAQ,GAAuC,IAAI,CAAC;AACxD,IAAI,WAAW,GAAgD,IAAI,CAAC;AAEpE;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,WAAW;IAC/B,IAAI,QAAQ;QAAE,OAAO,QAAQ,CAAC;IAC9B,IAAI,WAAW;QAAE,OAAO,WAAW,CAAC;IAEpC,WAAW,GAAG,QAAQ,CAAC,sBAAsB,EAAE,QAAQ,EAAE;QACvD,KAAK,EAAE,MAAM;KACd,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE;QACb,QAAQ,GAAG,IAAmC,CAAC;QAC/C,OAAO,CAAC,GAAG,CAAC,0BAA0B,QAAQ,EAAE,CAAC,CAAC;QAClD,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC,CAAC;IAEH,OAAO,WAAW,CAAC;AACrB,CAAC;AAED;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,WAAW,CAAC,aAAqB;IACrD,IAAI,CAAC;QACH,MAAM,QAAQ,GAAG,MAAM,WAAW,EAAE,CAAC;QACrC,MAAM,MAAM,GAAG,8HAA8H,aAAa,qBAAqB,CAAC;QAEhL,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,MAAM,EAAE;YACpC,cAAc,EAAE,EAAE;YAClB,oBAAoB,EAAE,CAAC;SACxB,CAAC,CAAC;QAEH,MAAM,QAAQ,GAAG,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,CAAE,MAAM,CAAC,CAAC,CAAS,EAAE,cAAc,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACvF,MAAM,aAAa,GAAG,QAAQ,CAAC,IAAI,EAAE,CAAC;QAEtC,IAAI,aAAa,IAAI,aAAa,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC9C,OAAO,GAAG,aAAa,IAAI,aAAa,EAAE,CAAC;QAC7C,CAAC;QACD,OAAO,aAAa,CAAC;IACvB,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,aAAa,CAAC;IACvB,CAAC;AACH,CAAC"}
+{"version":3,"file":"query-expander.js","sourceRoot":"","sources":["../../src/core/query-expander.ts"],"names":[],"mappings":"AAAA,gDAAgD;AAChD,sCAAsC;AACtC;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,QAAQ,EAAoC,MAAM,2BAA2B,CAAC;AAEvF,MAAM,QAAQ,GAAG,sBAAsB,CAAC;AACxC,IAAI,QAAQ,GAAuC,IAAI,CAAC;AACxD,IAAI,WAAW,GAAgD,IAAI,CAAC;AAEpE;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,WAAW;IAC/B,IAAI,QAAQ;QAAE,OAAO,QAAQ,CAAC;IAC9B,IAAI,WAAW;QAAE,OAAO,WAAW,CAAC;IAEpC,WAAW,GAAG,QAAQ,CAAC,sBAAsB,EAAE,QAAQ,EAAE;QACvD,KAAK,EAAE,MAAM;KACd,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE;QACb,QAAQ,GAAG,IAAmC,CAAC;QAC/C,OAAO,CAAC,GAAG,CAAC,0BAA0B,QAAQ,EAAE,CAAC,CAAC;QAClD,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC,CAAC;IAEH,OAAO,WAAW,CAAC;AACrB,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,cAAc,GAAG,IAAI,GAAG,EAAkB,CAAC;AACjD,MAAM,qBAAqB,GAAG,GAAG,CAAC;AAElC;;;;;;GAMG;AACH,SAAS,mBAAmB,CAAC,UAAkB;IAC7C,IAAI,UAAU,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IACzC,IAAI,UAAU,CAAC,MAAM,GAAG,EAAE;QAAE,OAAO,IAAI,CAAC;IACxC,mDAAmD;IACnD,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,UAAU,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC;IAC1E,OAAO,MAAM,CAAC,IAAI,IAAI,CAAC,CAAC;AAC1B,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,CAAC,KAAK,UAAU,WAAW,CAAC,aAAqB;IACrD,MAAM,UAAU,GAAG,aAAa,CAAC,WAAW,EAAE,CAAC,IAAI,EAAE,CAAC;IACtD,MAAM,oBAAoB,GAAG,OAAO,CAAC,GAAG,CAAC,2BAA2B,KAAK,GAAG,CAAC;IAE7E,IAAI,oBAAoB,EAAE,CAAC;QACzB,IAAI,mBAAmB,CAAC,UAAU,CAAC,EAAE,CAAC;YACpC,OAAO,aAAa,CAAC;QACvB,CAAC;QACD,MAAM,MAAM,GAAG,cAAc,CAAC,GAAG,CAAC,UAAU,CAAC,CAAC;QAC9C,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;YACzB,8CAA8C;YAC9C,cAAc,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC;YAClC,cAAc,CAAC,GAAG,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC;YACvC,OAAO,MAAM,CAAC;QAChB,CAAC;IACH,CAAC;IAED,IAAI,CAAC;QACH,MAAM,QAAQ,GAAG,MAAM,WAAW,EAAE,CAAC;QACrC,MAAM,MAAM,GAAG,8HAA8H,aAAa,qBAAqB,CAAC;QAEhL,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,MAAM,EAAE;YACpC,cAAc,EAAE,EAAE;YAClB,oBAAoB,EAAE,CAAC;SACxB,CAAC,CAAC;QAEH,MAAM,QAAQ,GAAG,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,CAAE,MAAM,CAAC,CAAC,CAAS,EAAE,cAAc,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACvF,MAAM,aAAa,GAAG,QAAQ,CAAC,IAAI,EAAE,CAAC;QAEtC,MAAM,UAAU,GAAG,aAAa,IAAI,aAAa,CAAC,MAAM,GAAG,CAAC;YAC1D,CAAC,CAAC,GAAG,aAAa,IAAI,aAAa,EAAE;YACrC,CAAC,CAAC,aAAa,CAAC;QAElB,qDAAqD;QACrD,IAAI,oBAAoB,EAAE,CAAC;YACzB,IAAI,cAAc,CAAC,IAAI,IAAI,qBAAqB,EAAE,CAAC;gBACjD,MAAM,SAAS,GAAG,cAAc,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC;gBACrD,IAAI,SAAS,KAAK,SAAS;oBAAE,cAAc,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;YAChE,CAAC;YACD,cAAc,CAAC,GAAG,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;QAC7C,CAAC;QAED,OAAO,UAAU,CAAC;IACpB,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,aAAa,CAAC;IACvB,CAAC;AACH,CAAC;AAED,gFAAgF;AAChF,MAAM,UAAU,mBAAmB;IACjC,cAAc,CAAC,KAAK,EAAE,CAAC;AACzB,CAAC"}

package/dist/index.js

@@ -175,7 +175,7 @@ async function main() {
     getExpander().catch(err => console.warn('Query expander model unavailable:', err.message));
     // Start server
     await app.listen({ port: PORT, host: '0.0.0.0' });
-    console.log(`AgentWorkingMemory v0.7.10 listening on port ${PORT}`);
+    console.log(`AgentWorkingMemory v0.7.11 listening on port ${PORT}`);
     // Graceful shutdown
     const shutdown = async () => {
         clearInterval(backupTimer);

package/dist/mcp.js

@@ -71,7 +71,7 @@ import { queryPeerDecisions, formatPeerDecisions } from './coordination/peer-dec
 const INCOGNITO = process.env.AWM_INCOGNITO === '1' || process.env.AWM_INCOGNITO === 'true';
 if (INCOGNITO) {
     console.error('AWM: incognito mode — all memory tools disabled, nothing will be recorded');
-    const server = new McpServer({ name: 'agent-working-memory', version: '0.7.10' });
+    const server = new McpServer({ name: 'agent-working-memory', version: '0.7.11' });
     const transport = new StdioServerTransport();
     server.connect(transport).catch(err => {
         console.error('MCP server failed:', err);
@@ -102,7 +102,7 @@ else {
     let coordDb = null;
     const server = new McpServer({
         name: 'agent-working-memory',
-        version: '0.7.10',
+        version: '0.7.11',
     });
     server.registerResource('awm-overview', 'awm://server/overview', {
         title: 'AWM Overview',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-working-memory",
-  "version": "0.7.10",
+  "version": "0.7.11",
   "description": "Cognitive memory layer for AI agents — activation-based retrieval, salience filtering, associative connections",
   "type": "module",
   "main": "dist/index.js",

package/src/api/routes.ts

@@ -705,7 +705,7 @@ export function registerRoutes(app: FastifyInstance, deps: MemoryDeps): void {
     const base: Record<string, unknown> = {
       status: 'ok',
       timestamp: new Date().toISOString(),
-      version: '0.7.10',
+      version: '0.7.11',
       coordination: coordEnabled,
     };
     if (coordEnabled) {

package/src/cli.ts

@@ -334,7 +334,7 @@ async function exportMemories() {
   const agents = [...new Set(memories.map((m: any) => m.agent_id))];
 
   const exportData = {
-    version: '0.7.10',
+    version: '0.7.11',
     exported_at: new Date().toISOString(),
     source_db: dbPath,
     agent_filter: agentFilter,

package/src/core/query-expander.ts

@@ -37,12 +37,60 @@ export async function getExpander(): Promise<Text2TextGenerationPipeline> {
   return initPromise;
 }
 
+/**
+ * LRU cache of normalized-query → expanded-query mappings.
+ * Map preserves insertion order — re-set on hit moves entry to the end (most-recent),
+ * delete-first when over capacity drops the least-recent. ~500 entries × ~200 chars
+ * each ≈ 100KB, negligible memory cost.
+ *
+ * Why: phase-breakdown spike (2026-05-08) showed expandQuery at ~164ms per call.
+ * Agent recall patterns repeat — cache hits are common.
+ */
+const expansionCache = new Map<string, string>();
+const EXPANSION_CACHE_LIMIT = 500;
+
+/**
+ * Heuristic: skip expansion when the query is already specific.
+ * Long, multi-token queries don't benefit from synonyms — they're already
+ * narrow enough that flan-t5's general-vocabulary expansion adds noise more
+ * than recall. Exact thresholds are conservative; false-skips would only
+ * affect candidates that BM25 catches anyway.
+ */
+function shouldSkipExpansion(normalized: string): boolean {
+  if (normalized.length === 0) return true;
+  if (normalized.length > 50) return true;
+  // ≥5 distinct meaningful tokens = already specific
+  const tokens = new Set(normalized.split(/\s+/).filter(t => t.length > 2));
+  return tokens.size >= 5;
+}
+
 /**
  * Expand a query with related terms and synonyms.
  * Returns the original query + generated expansion terms.
  * Falls back to the original query on any error.
+ *
+ * Optimization (0.7.11+):
+ * - Skip heuristic for long/specific queries (~30% of typical agent recalls)
+ * - LRU cache for repeated queries (cache hit ≈ 0ms vs 164ms cold)
+ * - Disable both via AWM_DISABLE_EXPANSION_CACHE=1
  */
 export async function expandQuery(originalQuery: string): Promise<string> {
+  const normalized = originalQuery.toLowerCase().trim();
+  const optimizationsEnabled = process.env.AWM_DISABLE_EXPANSION_CACHE !== '1';
+
+  if (optimizationsEnabled) {
+    if (shouldSkipExpansion(normalized)) {
+      return originalQuery;
+    }
+    const cached = expansionCache.get(normalized);
+    if (cached !== undefined) {
+      // Move to end (most-recent) for LRU semantics
+      expansionCache.delete(normalized);
+      expansionCache.set(normalized, cached);
+      return cached;
+    }
+  }
+
   try {
     const expander = await getExpander();
     const prompt = `Expand this search query with synonyms and related terms. Only output the additional terms, not the original query. Query: ${originalQuery}. Additional terms:`;
@@ -55,11 +103,26 @@ export async function expandQuery(originalQuery: string): Promise<string> {
     const expanded = Array.isArray(result) ? (result[0] as any)?.generated_text ?? '' : '';
     const cleanExpanded = expanded.trim();
 
-    if (cleanExpanded && cleanExpanded.length > 2) {
-      return `${originalQuery} ${cleanExpanded}`;
+    const finalQuery = cleanExpanded && cleanExpanded.length > 2
+      ? `${originalQuery} ${cleanExpanded}`
+      : originalQuery;
+
+    // Cache the result (LRU eviction when over capacity)
+    if (optimizationsEnabled) {
+      if (expansionCache.size >= EXPANSION_CACHE_LIMIT) {
+        const oldestKey = expansionCache.keys().next().value;
+        if (oldestKey !== undefined) expansionCache.delete(oldestKey);
+      }
+      expansionCache.set(normalized, finalQuery);
     }
-    return originalQuery;
+
+    return finalQuery;
   } catch {
     return originalQuery;
   }
 }
+
+/** Clear the expansion cache (used by tests + cache invalidation if needed). */
+export function clearExpansionCache(): void {
+  expansionCache.clear();
+}

package/src/index.ts

@@ -177,7 +177,7 @@ async function main() {
 
   // Start server
   await app.listen({ port: PORT, host: '0.0.0.0' });
-  console.log(`AgentWorkingMemory v0.7.10 listening on port ${PORT}`);
+  console.log(`AgentWorkingMemory v0.7.11 listening on port ${PORT}`);
 
   // Graceful shutdown
   const shutdown = async () => {

package/src/mcp.ts

@@ -78,7 +78,7 @@ const INCOGNITO = process.env.AWM_INCOGNITO === '1' || process.env.AWM_INCOGNITO
 
 if (INCOGNITO) {
   console.error('AWM: incognito mode — all memory tools disabled, nothing will be recorded');
-  const server = new McpServer({ name: 'agent-working-memory', version: '0.7.10' });
+  const server = new McpServer({ name: 'agent-working-memory', version: '0.7.11' });
   const transport = new StdioServerTransport();
   server.connect(transport).catch(err => {
     console.error('MCP server failed:', err);
@@ -115,7 +115,7 @@ let coordDb: import('better-sqlite3').Database | null = null;
 
 const server = new McpServer({
   name: 'agent-working-memory',
-  version: '0.7.10',
+  version: '0.7.11',
 });
 
 server.registerResource(