@pyxmate/memory 0.39.0 → 0.41.0

package/dist/{chunk-TDPGRIJW.mjs → chunk-J3HY44AP.mjs}

@@ -1,6 +1,6 @@
 import {
   MemoryClient
-} from "./chunk-CK7VUON3.mjs";
+} from "./chunk-YT46GYB3.mjs";
 
 // ../dashboard/src/aggregations/consolidation-analytics.ts
 function analyzeConsolidationLog(entries) {

package/dist/{chunk-CK7VUON3.mjs → chunk-YT46GYB3.mjs}

@@ -107,6 +107,8 @@ var MemoryClient = class {
     }
     if (params.asOf) searchParams.set("asOf", params.asOf);
     if (params.anchorTime) searchParams.set("anchorTime", params.anchorTime);
+    const enumerationConcept = params.enumerationConcept?.trim();
+    if (enumerationConcept) searchParams.set("enumerationConcept", enumerationConcept);
     if (params.abstentionThreshold != null)
       searchParams.set("abstentionThreshold", String(params.abstentionThreshold));
     return this.fetchApi(`/api/memory/search?${searchParams}`);
@@ -733,10 +735,8 @@ var MemoryClient = class {
    * positive overlap. pyx never auto-prepends — the agent owns inclusion.
    */
   async fetchApplicableCorrections(input) {
-    const params = new URLSearchParams({
-      namespaceId: input.namespaceId,
-      taskShape: input.taskShape
-    });
+    const params = new URLSearchParams({ taskShape: input.taskShape });
+    if (input.namespaceId) params.set("namespaceId", input.namespaceId);
     if (input.project) params.set("project", input.project);
     if (input.limit !== void 0) params.set("limit", String(input.limit));
     return this.fetchApi(`/api/memory/corrections?${params}`);

package/dist/cli/pyx-mem.mjs

@@ -555,7 +555,7 @@ import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 
 // src/mcp/instructions.ts
-var PYX_MEMORY_INSTRUCTIONS = `Use pyx-memory to search durable project/user memory before assuming prior decisions, and to store concise facts after corrections, bug fixes, design decisions, integration discoveries, gotchas, explicit preferences, or "remember this" requests. Store decisions, not deliberation. Include topic and project. Pass eventTime (ISO-8601, when the fact happened or took effect) for any fact that can change or go stale \u2014 job/status changes, decisions that supersede earlier ones, dated events; recency ordering, dated ("as of") queries, and stale-vs-current conflict resolution all key off it. Search effort is retrieval depth: quick=strongest, deep=everything including archived/superseded. Use reinforce after memories were actually used, so they surface in quick/medium effort tiers. Use lineage for the history of how a fact changed; pass subject+relation for graph lineage or entryId for a superseded chain. When a question names a time \u2014 explicit or relative ("last year", "\uB450 \uB2EC \uC804") \u2014 resolve it to an absolute ISO-8601 timestamp yourself and pass it as search anchorTime: results then rank by proximity to this time instead of now, without excluding anything. When content names people, organizations, tools, places, events, or key concepts, you (the caller) must extract and pass both entities and relationships \u2014 the server does not auto-extract them. Edges matter as much as nodes: without relationships the graph cannot connect related memories. Use file ingest for documents/images worth persisting; images require a description.`;
+var PYX_MEMORY_INSTRUCTIONS = `Use pyx-memory to search durable project/user memory before assuming prior decisions, and to store concise facts after corrections, bug fixes, design decisions, integration discoveries, gotchas, explicit preferences, or "remember this" requests. Store decisions, not deliberation. Include topic and project. Pass eventTime (ISO-8601, when the fact happened or took effect) for any fact that can change or go stale \u2014 job/status changes, decisions that supersede earlier ones, dated events; recency ordering, dated ("as of") queries, and stale-vs-current conflict resolution all key off it. Search effort is retrieval depth: quick=strongest, deep=everything including archived/superseded. Use reinforce after memories were actually used, so they surface in quick/medium effort tiers. Use lineage for the history of how a fact changed; pass subject+relation for graph lineage or entryId for a superseded chain. Use record_correction when the user corrects a mistake you made (what was wrong, what to do instead, when it applies); call fetch_applicable_corrections before a task to retrieve the corrections that match it, then decide which to follow \u2014 pyx never auto-applies them. When a question names a time \u2014 explicit or relative ("last year", "\uB450 \uB2EC \uC804") \u2014 resolve it to an absolute ISO-8601 timestamp yourself and pass it as search anchorTime: results then rank by proximity to this time instead of now, without excluding anything. For count/list questions about a category, pass \`enumerationConcept\` = the language-agnostic/global category phrase (e.g. "fitness classes" / "\uC6B4\uB3D9 \uC218\uC5C5"); the caller-supplied hint is resolved by the embedding model, which covers a broad range of languages. pyx also auto-detects English/Korean count phrasing as a best-effort fallback when the hint is omitted. When content names people, organizations, tools, places, events, or key concepts, you (the caller) must extract and pass both entities and relationships \u2014 the server does not auto-extract them. Edges matter as much as nodes: without relationships the graph cannot connect related memories. For countable categories the user may later enumerate ("how many fitness classes / streaming services / pets do I have?"), add a canonical category CONCEPT node and an IS_A edge from each member to it. A language/synonym variant gives zero-cost exact-match resolution only when it is also a memberful CONCEPT node with member IS_A edges; there is no query-time alias\u2192canonical resolver. pyx resolves count/list category hints at query time via strict embedding over memberful CONCEPT nodes. Use file ingest for documents/images worth persisting; images require a description.`;
 
 // src/mcp/sampling.ts
 function createSamplingClient(server) {
@@ -713,7 +713,7 @@ var scopeShape = {
 
 // src/mcp/tools/corrections.ts
 var recordInputShape = {
-  namespaceId: z2.string().min(1).describe("Namespace id to record the correction into."),
+  namespaceId: z2.string().min(1).optional().describe("Namespace id to record into. Omit in single-tenant deployments (namespace-free)."),
   whatWasWrong: z2.string().min(1).describe("What the agent did wrong."),
   whatToDoInstead: z2.string().min(1).describe("The corrective instruction to follow instead."),
   appliesWhen: z2.string().min(1).describe("When this correction applies \u2014 the task context future tasks are matched against."),
@@ -738,7 +738,7 @@ var recordCorrectionTool = {
       method: "POST",
       path: "/api/memory/corrections",
       body: {
-        namespaceId: args.namespaceId,
+        ...args.namespaceId !== void 0 ? { namespaceId: args.namespaceId } : {},
         whatWasWrong: args.whatWasWrong,
         whatToDoInstead: args.whatToDoInstead,
         appliesWhen: args.appliesWhen,
@@ -751,7 +751,7 @@ var recordCorrectionTool = {
   }
 };
 var fetchInputShape = {
-  namespaceId: z2.string().min(1).describe("Namespace id whose corrections to fetch."),
+  namespaceId: z2.string().min(1).optional().describe("Namespace id whose corrections to fetch. Omit in single-tenant deployments."),
   taskShape: z2.string().min(1).describe("Shape of the task about to run \u2014 scored against each correction."),
   project: z2.string().optional().describe("Optional project filter."),
   limit: z2.number().int().positive().optional().describe("Max corrections to return (cap 5)."),
@@ -1091,7 +1091,7 @@ import { z as z10 } from "zod";
 var inputShape7 = {
   query: z10.string().min(1).describe("Required natural-language search text."),
   limit: z10.number().int().min(1).max(100).optional().describe(
-    'Target result count; server clamps to 1\u2013100. Default 10. The hybrid strategy may widen enumeration/count queries ("how many \u2026", "list all/every \u2026") up to 30 for set-completeness recall.'
+    "Target result count; server clamps to 1\u2013100. Default 10. The hybrid strategy may widen count/list searches up to 30 for set-completeness recall when the caller passes `enumerationConcept`; English/Korean marker auto-detect is a best-effort fallback."
   ),
   strategy: z10.enum(["naive", "graph", "hybrid"]).optional().describe(
     "RAG strategy. Defaults to `hybrid` (cross-encoder reranking, multi-entity decomposition, confidence scoring) and is sent explicitly when omitted; pass `naive` for a lighter vector-only search or `graph` for graph-augmented retrieval."
@@ -1108,6 +1108,9 @@ var inputShape7 = {
   anchorTime: z10.string().optional().describe(
     'Soft recency ANCHOR (ISO-8601) \u2014 ranks results by proximity to this time instead of now; never excludes anything. When the question names a relative time ("two months ago", "last year", "3\uB144 \uC804"), resolve it against the current date yourself and pass the absolute timestamp here; prefer this over eventTimeStart/End unless a strict window is required, because hard filters drop last-known-before facts.'
   ),
+  enumerationConcept: z10.string().trim().min(1).optional().describe(
+    'For count/list questions about a category, pass the language-agnostic/global category phrase here (for example, "fitness classes" or "\uC6B4\uB3D9 \uC218\uC5C5"). The embedding model resolves the caller-supplied hint across a broad range of languages.'
+  ),
   ...scopeShape
 };
 var searchMemoriesTool = {
@@ -1141,7 +1144,8 @@ var searchMemoriesTool = {
         eventTimeStart: args.eventTimeStart,
         eventTimeEnd: args.eventTimeEnd,
         asOf: args.asOf,
-        anchorTime: args.anchorTime
+        anchorTime: args.anchorTime,
+        enumerationConcept: args.enumerationConcept
       },
       scope: args
     });
@@ -1283,7 +1287,7 @@ var inputShape9 = {
       type: z12.enum(relationshipTypes).describe("Relationship type.")
     })
   ).optional().describe(
-    "Edges between entities; source and target must be entity names from this request. Relationships matter as much as entities because graph traversal needs edges to connect related memories."
+    'Edges between entities; source and target must be entity names from this request. Relationships matter as much as entities because graph traversal needs edges to connect related memories. For countable categories the user may later enumerate ("how many fitness classes / streaming services / pets do I have?"), add a canonical category CONCEPT node and an IS_A edge from each member to it. A language/synonym variant gives zero-cost exact-match resolution only when it is also a memberful CONCEPT node with member IS_A edges; there is no query-time alias\u2192canonical resolver. pyx resolves un-enumerated count-noun variants at query time via strict embedding over memberful CONCEPT nodes.'
   ),
   extractEntities: z12.boolean().optional().describe(
     "Override extraction: false skips caller-side extraction; true requires caller entities or MCP sampling and errors loudly if neither is available."
@@ -1420,7 +1424,7 @@ var ALL_TOOL_NAMES = ALL_TOOLS.map((t) => t.name);
 // src/mcp/server.ts
 async function runMcpServer(opts) {
   const fetchImpl = opts.fetchImpl ?? fetch;
-  const version = opts.version ?? (true ? "0.39.0" : "0.0.0-dev");
+  const version = opts.version ?? (true ? "0.41.0" : "0.0.0-dev");
   const server = new McpServer(
     { name: "pyx-memory", version },
     { instructions: PYX_MEMORY_INSTRUCTIONS, capabilities: { tools: {} } }

package/dist/dashboard.mjs

@@ -11,8 +11,8 @@ import {
   toGraphologyFormat,
   transformGraphData,
   unreachableHealth
-} from "./chunk-TDPGRIJW.mjs";
-import "./chunk-CK7VUON3.mjs";
+} from "./chunk-J3HY44AP.mjs";
+import "./chunk-YT46GYB3.mjs";
 import "./chunk-KSTI4M52.mjs";
 export {
   DashboardClient,

package/dist/index.d.ts

@@ -394,7 +394,8 @@ declare class MemoryClient implements ExtendedMemoryInterface {
      * the body carries only the correction fields. Returns `{ id, createdAt }`.
      */
     recordCorrection(input: {
-        namespaceId: string;
+        /** Omit in single-tenant deployments — corrections are stored namespace-free. */
+        namespaceId?: string;
         whatWasWrong: string;
         whatToDoInstead: string;
         appliesWhen: string;
@@ -410,7 +411,8 @@ declare class MemoryClient implements ExtendedMemoryInterface {
      * positive overlap. pyx never auto-prepends — the agent owns inclusion.
      */
     fetchApplicableCorrections(input: {
-        namespaceId: string;
+        /** Omit in single-tenant deployments — reads the namespace-free scope. */
+        namespaceId?: string;
         taskShape: string;
         project?: string;
         limit?: number;
@@ -585,6 +587,14 @@ interface MemorySearchParams {
      * over the query-text year parse (hybrid strategy).
      */
     anchorTime?: string;
+    /**
+     * The category the user is asking to count/list (e.g. "fitness classes",
+     * "운동 수업", any language). Presence signals enumeration intent and supplies
+     * the category for graph resolution; the caller (which understands the user's
+     * language) sets it. When omitted, pyx falls back to a deterministic EN/KO
+     * marker detector.
+     */
+    enumerationConcept?: string;
     /** Confidence threshold below which the system recommends abstention (0.0–1.0). Default: 0.3. */
     abstentionThreshold?: number;
     /**
@@ -878,7 +888,12 @@ interface MemoryIngestRequest {
  */
 interface CorrectionInput {
     tenantId: string;
-    namespaceId: string;
+    /**
+     * Namespace partition. Required in multi-tenant mode (the isolation
+     * boundary). Optional in single-tenant mode, where corrections are stored
+     * namespace-free (`namespace_id IS NULL`) exactly like ordinary writes.
+     */
+    namespaceId?: string;
     /** What the agent did wrong (required free text). */
     whatWasWrong: string;
     /** The corrective instruction (required free text) — stored as `entry.content`. */
@@ -895,7 +910,8 @@ interface CorrectionInput {
 /** Input to `Memory.fetchApplicableCorrections`. */
 interface FetchCorrectionsInput {
     tenantId: string;
-    namespaceId: string;
+    /** Namespace partition; omit to read the namespace-free scope (single-tenant mode). */
+    namespaceId?: string;
     /** Shape of the task about to run — scored against each correction's `appliesWhen`. */
     taskShape: string;
     /** Optional project filter — keeps only project-matching or project-absent rows. */

package/dist/index.mjs

@@ -1,7 +1,7 @@
 import {
   MemoryClient,
   MemoryServerError
-} from "./chunk-CK7VUON3.mjs";
+} from "./chunk-YT46GYB3.mjs";
 import {
   DEFAULTS,
   DEPRECATED_RAG_STRATEGIES,

package/dist/react.mjs

@@ -11,8 +11,8 @@ import {
   toGraphologyFormat,
   transformGraphData,
   unreachableHealth
-} from "./chunk-TDPGRIJW.mjs";
-import "./chunk-CK7VUON3.mjs";
+} from "./chunk-J3HY44AP.mjs";
+import "./chunk-YT46GYB3.mjs";
 import "./chunk-KSTI4M52.mjs";
 
 // ../dashboard/src/hooks/use-consolidation-log.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pyxmate/memory",
-  "version": "0.39.0",
+  "version": "0.41.0",
   "type": "module",
   "description": "SDK for pyx-memory — Memory as a Service for AI agents",
   "license": "MIT",