Fix Assets sync: correct AQL pagination + Jira Cloud CMDB field format

Two bugs found during end-to-end smoke test on the stores cache
workaround (Forgejo #8):

1. Wrong pagination params on POST /object/aql. The endpoint uses
   startAt/maxResults as URL query params, not page/resultPerPage in the
   body. Passing the wrong param names caused the API to silently fall
   back to the default maxResults=25, so the first sync only cached 25
   stores. After fix: 2,661 stores paginated across 6 pages in 5.3s.

2. Wrong request-field shape for CMDB-object custom fields on Jira
   Cloud. resolveStoreAssetReference was returning
     [{ objectId: "75974" }]
   which is the legacy Data Center / Server shape. Cloud requires
     [{ id: "<workspaceId>:<objectId>" }]
   The old shape is silently accepted (HTTP 204) by REST and by
   POST /rest/servicedeskapi/request, but the field is never actually
   persisted \u2014 verified via direct REST GET showing customfield_10261:[].
   After fix: SS-20948 shows the store correctly populated.

- assetsSyncClient.js: use URLSearchParams to pass
  startAt/maxResults/includeAttributes; drop the old page/resultPerPage
  opts; jsdoc updated with pointer to Atlassian Assets API v1 docs.
- storesCache.js: pagination loop now advances by startAt +=
  values.length and trusts isLast (Atlassian caps `total` at 1000 as a
  hint on this endpoint, so we can't rely on it).
- assets.js: new buildStoreFieldRef(objectId) helper produces the
  correct Cloud shape from config.jira.assetsWorkspaceId; all three
  resolution paths (cache / live PAT / service-account fallback) route
  through it. Falls back to legacy shape with a loud error log if
  workspaceId is unset (misconfig).

End-to-end verified: SS-20948 was created for store 00782 (objectId
75974) via cache-only lookup and shows the field populated on both the
JSM request echo and a direct REST GET.

Co-authored-by: Cursor <cursoragent@cursor.com>

.env.example

@@ -45,6 +45,35 @@ JIRA_ASSETS_STORE_NUMBER_ATTRIBUTE_ID=
 # The custom field on the JSM request that holds the Store Assets reference.
 JIRA_STORE_CUSTOM_FIELD_ID=customfield_10261
 
+# --- Assets Stores cache (personal-PAT sync workaround) ---
+# The service account is silently filtered out of Object Type 109 (see
+# Forgejo issue #1). Until that's fixed, the app populates a local store-number
+# -> objectId cache using a *personal* Atlassian PAT that has the right role.
+# This PAT is used ONLY for reading the Stores schema; nothing that mutates
+# Jira state uses it.
+#
+# The .env file is the primary supported storage for the PAT (the bot runs on
+# a Linux host; macOS Keychain isn't available there). Because .env stays out
+# of source control (.gitignore) and app.log no longer echoes auth headers,
+# the cleartext token here is scoped to whoever has filesystem access on the
+# deploy host — lock the file down with `chmod 600 .env` and rotate the token
+# if that trust changes.
+#
+# For local dev on macOS you can instead source the token from Keychain via
+# bin/load-assets-sync-secret.sh (see README) and leave ASSETS_SYNC_TOKEN out
+# of .env entirely.
+ASSETS_SYNC_EMAIL=you@ae.com
+ASSETS_SYNC_TOKEN=REPLACE_ME
+
+# Where the local cache lives on disk (JSON). Gitignored. Regenerable via
+# POST /api/wxccai/admin/storesCache/refresh.
+STORES_CACHE_PATH=./data/stores.json
+# How often to run a full resync (hours). 0 disables the scheduler.
+STORES_CACHE_REFRESH_HOURS=24
+# Cache is considered "stale" after this many hours; boot-time refresh fires
+# if the on-disk snapshot is older than this.
+STORES_CACHE_STALE_AFTER_HOURS=48
+
 # --- xAI (Grok) ---
 XAI_API_KEY=REPLACE_ME
 XAI_BASE_URL=https://api.x.ai/v1

.gitignore

@@ -35,6 +35,11 @@ coverage/
 tmp/
 .tmp/
 
+# Stores cache and any other locally-generated app state.
+# Never contains secrets, but is machine-specific and can be regenerated
+# from Assets via POST /api/wxccai/admin/storesCache/refresh.
+data/
+
 # =============================================
 # Editor / IDE / OS
 # =============================================

README.md

@@ -55,10 +55,70 @@ Base path: `/api/wxccai`.
 
 - `POST /issueTranscript/:jiraKey` — attaches audio + JSON transcript + human-readable transcript, then posts a restricted-visibility summary comment.
 
+### Admin
+
+- `GET /admin/storesCache/status` — snapshot of the local Assets store cache: `{ storeCount, lastSyncAt, ageSeconds, syncing, assetsSyncConfigured, ... }`. Safe for health checks.
+- `POST /admin/storesCache/refresh` — force an immediate resync via the personal PAT (see below). Takes a few seconds and returns the new status.
+
 ### Debug (non-production only)
 
 - `GET /debug/assetsProbe?storeNumber=305` — runs several AQL variants against Jira Assets and returns visible schemas + object-type detail + a computed diagnosis. Returns 404 when `NODE_ENV=production`.
 
+## Stores cache (Assets workaround)
+
+`createSSRequest` needs to translate a store number into an Assets object id. The shared service account is silently filtered out of the Store object type, so the app maintains a local `storeNumber → objectId` cache that's populated from a **personal Atlassian PAT** (a real human account with the right Assets role). The service account is still used for everything else (creating tickets, comments, attachments).
+
+Both storage patterns end up in the same place — `process.env.ASSETS_SYNC_TOKEN` — so the runtime code path is identical. Pick whichever fits the host.
+
+### Setup A — production / Linux host (`.env`)
+
+Put the values directly in `.env` (which is gitignored) and lock the file down:
+
+```bash
+cat >> .env <<'EOF'
+ASSETS_SYNC_EMAIL=you@ae.com
+ASSETS_SYNC_TOKEN=<paste-your-atlassian-api-token>
+EOF
+
+chmod 600 .env    # only the bot user can read it
+```
+
+Then start normally:
+
+```bash
+npm start
+```
+
+Rotate the token in Atlassian → Account → Security → API tokens whenever the trust boundary on the host changes (new operator, offboarding, suspected leak). The app reloads it on the next process start.
+
+### Setup B — local dev on macOS (Keychain)
+
+If you're running the app on a Mac and would rather not keep the PAT in `.env`, use the wrapper script — it pulls the token from Keychain into `ASSETS_SYNC_TOKEN` before exec'ing the process:
+
+```bash
+security add-generic-password \
+    -s jira-assets-sync \
+    -a you@ae.com \
+    -w '<paste-your-atlassian-api-token>' \
+    -U
+
+echo 'ASSETS_SYNC_EMAIL=you@ae.com' >> .env   # email in .env, token stays in Keychain
+
+./bin/load-assets-sync-secret.sh npm start
+```
+
+If `ASSETS_SYNC_TOKEN` is already in the process env (Setup A above), the wrapper is a no-op and skips the Keychain lookup.
+
+### Runtime behavior
+
+On boot the app loads the on-disk cache at `data/stores.json`, kicks off a background refresh if the snapshot is missing or older than `STORES_CACHE_STALE_AFTER_HOURS`, and schedules a periodic full resync every `STORES_CACHE_REFRESH_HOURS`. `resolveStoreAssetReference` then serves lookups from memory (sub-ms) with a live PAT lookup as fallback for brand-new stores.
+
+Force a refresh at any time:
+
+```bash
+curl -X POST http://localhost:1866/api/wxccai/admin/storesCache/refresh
+```
+
 ## Jira Assets gotcha
 
 `createSSRequest` resolves a store number to an Assets object reference (`customfield_10261`). Two independent permission layers must both grant access, or every AQL query silently returns `total: 0`:

bin/load-assets-sync-secret.sh

@@ -0,0 +1,78 @@
+#!/usr/bin/env sh
+# -----------------------------------------------------------------------------
+# load-assets-sync-secret.sh
+#
+# Loads the personal Atlassian PAT used for the Assets → Stores cache sync
+# from macOS Keychain into ASSETS_SYNC_TOKEN, then execs whatever command you
+# passed as arguments.
+#
+# Rationale:
+#   The service account is silently filtered out of Object Type 109 (Store
+#   Address / Hierarchy) — see Forgejo issue #1. The stores cache is populated
+#   using a personal PAT instead. On a shared dev Mac, Keychain is a nicer home
+#   for that PAT than .env (encrypted at rest, per-user access). On the prod
+#   Linux host where the bot actually runs, put the PAT in .env directly with
+#   `chmod 600 .env` — this script exits as a no-op if ASSETS_SYNC_TOKEN is
+#   already exported into the process env.
+#
+# Setup (one-time, per machine):
+#
+#   # 1. Store the token in Keychain
+#   security add-generic-password \
+#       -s jira-assets-sync \
+#       -a mcqueenj@ae.com \
+#       -w '<paste-your-atlassian-api-token-here>' \
+#       -U
+#
+#   # 2. Set the email (either exported here or in ~/.zshrc)
+#   export ASSETS_SYNC_EMAIL="mcqueenj@ae.com"
+#
+# Usage:
+#
+#   ./bin/load-assets-sync-secret.sh npm start
+#   ./bin/load-assets-sync-secret.sh node src/app.js
+#
+# Environment variables (override defaults if needed):
+#
+#   ASSETS_SYNC_KEYCHAIN_SERVICE  Keychain service name (default: jira-assets-sync)
+#   ASSETS_SYNC_KEYCHAIN_ACCOUNT  Keychain account name (default: value of $ASSETS_SYNC_EMAIL)
+#   ASSETS_SYNC_TOKEN             If already set, skip the Keychain read entirely.
+# -----------------------------------------------------------------------------
+
+set -eu
+
+if [ -z "${ASSETS_SYNC_TOKEN:-}" ]; then
+    SERVICE="${ASSETS_SYNC_KEYCHAIN_SERVICE:-jira-assets-sync}"
+    ACCOUNT="${ASSETS_SYNC_KEYCHAIN_ACCOUNT:-${ASSETS_SYNC_EMAIL:-}}"
+
+    if [ -z "$ACCOUNT" ]; then
+        printf 'load-assets-sync-secret.sh: neither ASSETS_SYNC_TOKEN nor an account name is set.\n' >&2
+        printf '  Export ASSETS_SYNC_EMAIL=you@ae.com or ASSETS_SYNC_KEYCHAIN_ACCOUNT=<account>.\n' >&2
+        exit 1
+    fi
+
+    if ! command -v security >/dev/null 2>&1; then
+        printf 'load-assets-sync-secret.sh: `security` not found (this script is macOS-only).\n' >&2
+        printf '  On Linux/prod, export ASSETS_SYNC_TOKEN directly from your secret manager.\n' >&2
+        exit 1
+    fi
+
+    if ! ASSETS_SYNC_TOKEN=$(security find-generic-password -s "$SERVICE" -a "$ACCOUNT" -w 2>/dev/null); then
+        printf 'load-assets-sync-secret.sh: Keychain lookup failed for service="%s" account="%s".\n' "$SERVICE" "$ACCOUNT" >&2
+        printf '  Store the token with:\n' >&2
+        printf '    security add-generic-password -s %s -a %s -w '\''<token>'\'' -U\n' "$SERVICE" "$ACCOUNT" >&2
+        exit 1
+    fi
+    export ASSETS_SYNC_TOKEN
+fi
+
+# Only echoes existence, never the token itself.
+printf 'load-assets-sync-secret.sh: ASSETS_SYNC_TOKEN loaded (%d chars) for %s\n' \
+    "${#ASSETS_SYNC_TOKEN}" "${ASSETS_SYNC_EMAIL:-<unset ASSETS_SYNC_EMAIL>}" >&2
+
+if [ $# -eq 0 ]; then
+    printf 'load-assets-sync-secret.sh: no command given; exiting after loading token.\n' >&2
+    exit 0
+fi
+
+exec "$@"

src/app.js

@@ -2,6 +2,7 @@ import express from 'express';
 import config from './config/index.js';
 import wxccRoutes from './routes/wxccRoutes.js';
 import { getDetailedHealth } from './services/healthService.js';
+import * as storesCache from './services/jira/storesCache.js';
 import logger from './utilities/logger.js';
 
 const app = express();
@@ -106,6 +107,8 @@ app.get('/', (req, res) => {
             closeTicket: "POST /api/wxccai/ticket/:key/close  body: { transitionName?, resolution?, comment?, internal? }",
             createSSRequest: "POST /api/wxccai/createSSRequest  (store support ticket creator; storeNumber auto-resolved via Assets)",
             ssRequestTypes: "GET /api/wxccai/ssRequestTypes  (lists exact subType values + config notes)",
+            storesCacheStatus: "GET /api/wxccai/admin/storesCache/status  (Assets store cache state)",
+            storesCacheRefresh: "POST /api/wxccai/admin/storesCache/refresh  (force a resync via personal PAT)",
             assetsProbe: "GET /api/wxccai/debug/assetsProbe?storeNumber=305  (non-production only)"
         }
     });
@@ -131,4 +134,10 @@ app.use((req, res) => {
 
 app.listen(config.port, () => {
     logger.info(`Server running on port ${config.port} in ${config.nodeEnv} mode`);
+
+    // Kick off the stores cache: load from disk, schedule periodic refresh,
+    // and (if stale) start a background full refresh. Never blocks startup.
+    storesCache.init().catch(err => {
+        logger.error('storesCache.init failed at boot', { error: err.message });
+    });
 });

src/config/index.js

@@ -1,9 +1,23 @@
 import dotenv from 'dotenv';
+import path from 'node:path';
 dotenv.config();
 
 const cloudId = process.env.JIRA_CLOUD_ID?.trim();
 let baseUrl = process.env.JIRA_BASE_URL?.trim();
 
+const intFromEnv = (name, fallback) => {
+    const raw = process.env[name];
+    if (raw === undefined || raw === null || raw === '') return fallback;
+    const n = Number.parseInt(raw, 10);
+    return Number.isFinite(n) ? n : fallback;
+};
+
+const boolFromEnv = (name, fallback) => {
+    const raw = process.env[name];
+    if (raw === undefined || raw === null || raw === '') return fallback;
+    return /^(1|true|yes|on)$/i.test(String(raw).trim());
+};
+
 if (cloudId) {
     // When JIRA_CLOUD_ID is provided, use the api.atlassian.com/ex/jira gateway form.
     // This is the current style for addressing a specific Jira Cloud site (e.g. for
@@ -56,6 +70,34 @@ export const config = {
 
         // The custom field ID on the request that holds the Store Assets reference
         storeCustomFieldId: process.env.JIRA_STORE_CUSTOM_FIELD_ID || 'customfield_10261',
+
+        // --- Personal-account credentials for the Stores cache sync ---
+        // Compartmentalized from the service account above. Used ONLY by
+        // src/services/jira/assetsSyncClient.js to populate storesCache. This
+        // exists because the service account is silently filtered out for the
+        // Store object type (see Forgejo issue #1 for the underlying
+        // permission problem). Never used for anything that mutates state on
+        // Atlassian's side.
+        //
+        // The token is expected to be exported into the process env from macOS
+        // Keychain via bin/load-assets-sync-secret.sh — it should NOT sit in
+        // .env in cleartext.
+        assetsSyncEmail: process.env.ASSETS_SYNC_EMAIL?.trim() || null,
+        assetsSyncToken: process.env.ASSETS_SYNC_TOKEN || null,
+    },
+
+    // Local cache of the Assets Store schema. Backed by a JSON file on disk;
+    // refreshed periodically via the personal PAT above.
+    storesCache: {
+        enabled: boolFromEnv('STORES_CACHE_ENABLED', true),
+        // Where the on-disk snapshot lives. Default is data/stores.json under
+        // the process cwd; gitignored.
+        path: process.env.STORES_CACHE_PATH?.trim()
+            || path.resolve(process.cwd(), 'data', 'stores.json'),
+        refreshIntervalHours: intFromEnv('STORES_CACHE_REFRESH_HOURS', 24),
+        staleAfterHours: intFromEnv('STORES_CACHE_STALE_AFTER_HOURS', 48),
+        pageSize: intFromEnv('STORES_CACHE_PAGE_SIZE', 500),
+        maxPages: intFromEnv('STORES_CACHE_MAX_PAGES', 200),
     },
 
     xai: {

src/routes/wxccRoutes.js

@@ -1,22 +1,9 @@
 import express from 'express';
-import axios from 'axios';
-import axiosRetry from 'axios-retry';
 import { logger, webexLogger } from '../utilities/logger.js';
 import * as jiraService from '../services/jiraService.js';
 import grokService from '../services/grokService.js';
 import config from '../config/index.js';
 
-// Configure retry for transient failures
-axiosRetry(axios, {
-    retries: 3,
-    retryDelay: (retryCount) => Math.pow(2, retryCount) * 1000,
-    retryCondition: (error) => {
-        return axiosRetry.isNetworkOrIdempotentRequestError(error) ||
-               error.response?.status === 429 ||
-               error.response?.status >= 500;
-     }
-});
-
 const router = express.Router();
 
 // ========================
@@ -325,6 +312,40 @@ router.post('/wxccai/ticket/:key/close', async (req, res) => {
     }
 });
 
+// ========================
+// Stores cache admin
+// GET  /api/wxccai/admin/storesCache/status   — snapshot of cache state
+// POST /api/wxccai/admin/storesCache/refresh  — force an immediate resync
+//
+// These are admin endpoints (not user-facing). The status endpoint is safe to
+// hit from a health check; the refresh endpoint triggers an AQL walk of the
+// entire Store object type via the personal PAT and takes a few seconds on a
+// warm connection. Refresh is a no-op if credentials aren't configured or a
+// sync is already in progress.
+// ========================
+router.get('/wxccai/admin/storesCache/status', (req, res) => {
+    try {
+        res.json(jiraService.storesCache.status());
+    } catch (err) {
+        logger.error('storesCache status failed', { error: err.message });
+        res.status(500).json({ error: err.message });
+    }
+});
+
+router.post('/wxccai/admin/storesCache/refresh', async (req, res) => {
+    try {
+        const result = await jiraService.storesCache.refresh({ force: true });
+        res.json({ success: true, ...result, status: jiraService.storesCache.status() });
+    } catch (err) {
+        logger.error('storesCache refresh failed', { error: err.message });
+        res.status(500).json({
+            success: false,
+            error: err.message,
+            status: jiraService.storesCache.status(),
+        });
+    }
+});
+
 // ========================
 // Assets diagnostic (non-production only)
 // GET /api/wxccai/debug/assetsProbe?storeNumber=305

src/services/jira/assets.js

@@ -0,0 +1,732 @@
+// Jira Assets (formerly Insight / CMDB) integration.
+//
+// Store objects are "service objects" and therefore live behind the
+// workspace-scoped API:
+//   POST https://api.atlassian.com/jsm/assets/workspace/{workspaceId}/v1/object/aql
+//
+// This module contains:
+//   - Low-level helpers (workspace discovery, AQL POST, Assets GET) that
+//     never throw on non-2xx so callers can inspect what happened.
+//   - resolveStoreAssetReference: production path — store number → object ref.
+//   - probeAssetsForStore: dev-only diagnostic that runs many AQL variants
+//     plus schema/object-type introspection and returns a plain-English
+//     diagnosis of what's misconfigured.
+import axios from 'axios';
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+import { jiraClient } from './client.js';
+import * as storesCache from './storesCache.js';
+import { assetsAql as syncAssetsAql, isAssetsSyncConfigured } from './assetsSyncClient.js';
+
+/**
+ * Resolve the Assets workspace id. Prefers the explicit env var; otherwise
+ * discovers via /rest/servicedeskapi/assets/workspace.
+ */
+async function getAssetsWorkspaceId() {
+    if (config.jira.assetsWorkspaceId) {
+        return config.jira.assetsWorkspaceId;
+    }
+
+    const list = await listAssetsWorkspacesRaw();
+    const first = list.workspaces[0];
+    if (first?.workspaceId) {
+        logger.info(`Discovered Assets workspaceId via /rest/servicedeskapi/assets/workspace: ${first.workspaceId}`);
+        if (list.workspaces.length > 1) {
+            logger.warn(`Multiple Assets workspaces are visible to this account (${list.workspaces.length}); using the first. Set JIRA_ASSETS_WORKSPACE_ID explicitly to disambiguate.`, {
+                workspaces: list.workspaces.map(w => w.workspaceId)
+            });
+        }
+        return first.workspaceId;
+    }
+
+    throw new Error('JIRA_ASSETS_WORKSPACE_ID is required for Assets object lookup (or ensure /rest/servicedeskapi/assets/workspace is accessible).');
+}
+
+/**
+ * Return the full list of Assets workspaces the current account can see, plus
+ * the raw payload for diagnostics. Never throws.
+ */
+async function listAssetsWorkspacesRaw() {
+    try {
+        const resp = await jiraClient.get('/rest/servicedeskapi/assets/workspace');
+        const data = resp.data;
+
+        let entries = [];
+        if (Array.isArray(data)) entries = data;
+        else if (Array.isArray(data?.values)) entries = data.values;
+        else if (Array.isArray(data?.workspaces)) entries = data.workspaces;
+        else if (data && typeof data === 'object') entries = [data];
+
+        const workspaces = entries
+            .map(e => ({ workspaceId: e.workspaceId || e.id || e.key || e.workspaceID || null }))
+            .filter(w => w.workspaceId);
+
+        return { httpStatus: resp.status, workspaces, raw: data };
+    } catch (e) {
+        return {
+            httpStatus: e.response?.status || 'network',
+            workspaces: [],
+            raw: e.response?.data || null,
+            error: e.message
+        };
+    }
+}
+
+/**
+ * Filter out response headers that would leak scope/tenant/session identifiers
+ * or noise (cookies, tracing tokens, CORS bookkeeping). We keep just the
+ * Atlassian informational headers useful for debugging (rate-limit, tracing,
+ * deprecation, request id).
+ */
+function pickInterestingHeaders(headers = {}) {
+    const wanted = new Set([
+        'content-type', 'content-length',
+        'x-request-id', 'x-arequestid', 'x-arequest-id',
+        'x-ratelimit-limit', 'x-ratelimit-remaining', 'x-ratelimit-reset',
+        'x-atlassian-request-id', 'x-atlassian-trace-id',
+        'atl-traceid', 'atl-request-id',
+        'x-atlassian-server-status', 'x-atlassian-cursor',
+        'x-content-type-options', 'x-frame-options',
+        'deprecation', 'sunset', 'warning', 'retry-after'
+    ]);
+    const out = {};
+    for (const [k, v] of Object.entries(headers)) {
+        if (wanted.has(k.toLowerCase())) out[k] = v;
+    }
+    return out;
+}
+
+/**
+ * Low-level AQL POST helper. Never throws on non-2xx.
+ * Returns { status, statusText, data, headers, requestUrl, requestBody, workspaceId, error }.
+ * Uses bare axios (not jiraClient) because the URL is api.atlassian.com, not
+ * the Jira baseURL. Reuses jiraClient's Authorization header.
+ */
+async function runAssetsAql(qlQuery, { resultPerPage = 5, includeAttributes = true, extraBody = {} } = {}) {
+    const workspaceId = await getAssetsWorkspaceId();
+    const aqlUrl = `https://api.atlassian.com/jsm/assets/workspace/${workspaceId}/v1/object/aql`;
+
+    const authHeader = jiraClient.defaults.headers.Authorization
+        || jiraClient.defaults.headers.common?.Authorization;
+
+    const body = { qlQuery, resultPerPage, includeAttributes, ...extraBody };
+
+    logger.debug('Assets AQL request', { aqlUrl, qlQuery, resultPerPage });
+
+    try {
+        const resp = await axios.post(aqlUrl, body, {
+            headers: {
+                'Authorization': authHeader,
+                'Content-Type': 'application/json',
+                'Accept': 'application/json'
+            },
+            timeout: 15000,
+            validateStatus: () => true
+        });
+        return {
+            status: resp.status,
+            statusText: resp.statusText,
+            data: resp.data ?? null,
+            headers: pickInterestingHeaders(resp.headers || {}),
+            requestUrl: aqlUrl,
+            requestBody: body,
+            workspaceId
+        };
+    } catch (err) {
+        return {
+            status: err.response?.status || 'network',
+            statusText: err.response?.statusText || err.code || 'error',
+            data: err.response?.data || null,
+            headers: pickInterestingHeaders(err.response?.headers || {}),
+            requestUrl: aqlUrl,
+            requestBody: body,
+            workspaceId,
+            error: err.message
+        };
+    }
+}
+
+/**
+ * Low-level GET helper against api.atlassian.com Assets endpoints.
+ * Same shape as runAssetsAql. Use for schema/objecttype introspection.
+ */
+async function runAssetsGet(path) {
+    const workspaceId = await getAssetsWorkspaceId();
+    const url = `https://api.atlassian.com/jsm/assets/workspace/${workspaceId}/v1${path}`;
+
+    const authHeader = jiraClient.defaults.headers.Authorization
+        || jiraClient.defaults.headers.common?.Authorization;
+
+    try {
+        const resp = await axios.get(url, {
+            headers: { 'Authorization': authHeader, 'Accept': 'application/json' },
+            timeout: 15000,
+            validateStatus: () => true
+        });
+        return {
+            status: resp.status,
+            statusText: resp.statusText,
+            data: resp.data ?? null,
+            headers: pickInterestingHeaders(resp.headers || {}),
+            requestUrl: url,
+            workspaceId
+        };
+    } catch (err) {
+        return {
+            status: err.response?.status || 'network',
+            statusText: err.response?.statusText || err.code || 'error',
+            data: err.response?.data || null,
+            headers: pickInterestingHeaders(err.response?.headers || {}),
+            requestUrl: url,
+            workspaceId,
+            error: err.message
+        };
+    }
+}
+
+/**
+ * List all Assets schemas visible to the current token. Critical diagnostic:
+ * if this returns zero schemas, the token has no Assets access at all
+ * (regardless of what workspace id is used).
+ */
+export async function listAssetsSchemas() {
+    return runAssetsGet('/objectschema/list');
+}
+
+/**
+ * Fetch a single Assets object type (id, name, attributes). If HTTP 200,
+ * the token can see the type — and the attribute names in the response are
+ * authoritative for AQL queries.
+ */
+export async function getAssetsObjectType(objectTypeId) {
+    const [detail, attributes] = await Promise.all([
+        runAssetsGet(`/objecttype/${objectTypeId}`),
+        runAssetsGet(`/objecttype/${objectTypeId}/attributes`)
+    ]);
+    return { detail, attributes };
+}
+
+/**
+ * Flatten one Assets AQL "value" (object entry) into a compact shape suitable
+ * for humans debugging attribute names/values. Different Assets tenants return
+ * subtly different envelopes (attributes[].objectTypeAttribute vs typeAttribute,
+ * objectAttributeValues[].value vs displayValue), so we're defensive.
+ */
+function summarizeAssetsObject(obj) {
+    if (!obj || typeof obj !== 'object') return null;
+
+    const attributes = Array.isArray(obj.attributes) ? obj.attributes.map(attr => {
+        const meta = attr.objectTypeAttribute || attr.typeAttribute || {};
+        const rawValues = Array.isArray(attr.objectAttributeValues) ? attr.objectAttributeValues : [];
+        const values = rawValues.map(v => v.displayValue ?? v.value ?? v.searchValue ?? null).filter(v => v !== null);
+        return {
+            id: attr.objectTypeAttributeId || meta.id || null,
+            name: meta.name || null,
+            values
+        };
+    }) : [];
+
+    return {
+        id: obj.id || null,
+        objectKey: obj.objectKey || null,
+        name: obj.label || obj.name || null,
+        objectType: obj.objectType?.name || null,
+        objectTypeId: obj.objectType?.id || null,
+        attributes
+    };
+}
+
+/**
+ * Build the JSM/Jira Cloud request-field value for a CMDB-object custom field.
+ * On Jira Cloud the CMDB field expects `{ id: "<workspaceId>:<objectId>" }` —
+ * NOT the old `{ objectId }` shape from Data Center / Server, which is
+ * silently accepted (HTTP 204) but never actually persists to the ticket.
+ */
+function buildStoreFieldRef(objectId) {
+    const workspaceId = config.jira.assetsWorkspaceId;
+    if (!workspaceId) {
+        // Fall back to the legacy shape so at least *something* is sent. The
+        // caller will get an empty field on the created ticket, but we log the
+        // config problem loudly.
+        logger.error('JIRA_ASSETS_WORKSPACE_ID not configured; store custom field write will silently no-op on Jira Cloud');
+        return [{ objectId: String(objectId) }];
+    }
+    return [{ id: `${workspaceId}:${objectId}` }];
+}
+
+/**
+ * Resolve a store number (e.g. "00305" or 305) to the Assets object reference
+ * used for the Store custom field on a JSM request:
+ *   customfield_10261: [ { "id": "<workspaceId>:<objectId>" } ]
+ *
+ * Resolution order (each step logged so we can tell which path served the
+ * lookup):
+ *   1. Local stores cache (populated by the personal-PAT sync). Sub-ms; no
+ *      external call.
+ *   2. Live AQL via the personal PAT — handles brand-new stores that landed
+ *      between scheduled syncs.
+ *   3. Legacy AQL via the service account (kept as a fallback; currently
+ *      blocked by the Object-Type-109 permission wall tracked in Forgejo #1,
+ *      but will start working when that's resolved).
+ *
+ * Only throws once all three paths have missed.
+ */
+export async function resolveStoreAssetReference(rawStoreNumber) {
+    if (!rawStoreNumber) return null;
+
+    const normalized = String(rawStoreNumber).padStart(5, '0');
+    const objectTypeId = config.jira.assetsStoreObjectTypeId || '109';
+    const attribute = config.jira.assetsStoreNumberAttribute || 'Store Number';
+    const attrId = config.jira.assetsStoreNumberAttributeId;
+
+    // 1. Cache hit — the common case after the first sync.
+    const cached = storesCache.get(normalized);
+    if (cached) {
+        logger.info('Resolved store via cache', {
+            storeNumber: normalized,
+            objectId: cached.objectId,
+            cacheAge: storesCache.status().ageSeconds,
+        });
+        return buildStoreFieldRef(cached.objectId);
+    }
+
+    // 2. Live lookup via the personal-PAT sync client (handles brand-new
+    //    stores not yet in the cache).
+    if (isAssetsSyncConfigured()) {
+        const ql = `objectTypeId = ${objectTypeId} AND "${attribute}" = "${normalized}"`;
+        logger.info('Cache miss; trying live PAT lookup for store', { storeNumber: normalized, ql });
+        const r = await syncAssetsAql(ql, { maxResults: 1, includeAttributes: false });
+        if (r.status === 200) {
+            const values = Array.isArray(r.data?.values) ? r.data.values : [];
+            if (values.length > 0) {
+                const objectId = extractObjectIdFromResponse(r.data);
+                if (objectId) {
+                    logger.info('Resolved store via live PAT lookup', {
+                        storeNumber: normalized,
+                        objectId: String(objectId),
+                    });
+                    return buildStoreFieldRef(objectId);
+                }
+            }
+            logger.warn('Live PAT lookup returned 200 but no matching object', {
+                storeNumber: normalized,
+                total: r.data?.total,
+            });
+        } else {
+            logger.warn('Live PAT lookup failed', {
+                storeNumber: normalized,
+                status: r.status,
+                error: r.error,
+            });
+        }
+    }
+
+    // 3. Fallback: legacy service-account AQL. Kept behind the cache/PAT
+    //    layers so it costs nothing on the happy path.
+    const queries = [
+        `objectTypeId = ${objectTypeId} AND "${attribute}" = "${normalized}"`
+    ];
+    if (attrId) {
+        queries.push(`objectTypeId = ${objectTypeId} AND attribute[${attrId}] = "${normalized}"`);
+    }
+
+    let lastResult = null;
+
+    for (const qlQuery of queries) {
+        logger.info('Assets AQL lookup for store number (service-account fallback)', { qlQuery, storeNumber: normalized });
+
+        const result = await runAssetsAql(qlQuery, { resultPerPage: 1, includeAttributes: true });
+        lastResult = result;
+
+        if (result.status !== 200) {
+            logger.error('Assets AQL variant failed', {
+                qlQuery,
+                status: result.status,
+                error: result.error,
+                atlassianError: result.data
+            });
+            continue;
+        }
+
+        const data = result.data || {};
+        const values = Array.isArray(data.values) ? data.values : [];
+        const total = typeof data.total === 'number' ? data.total : values.length;
+
+        if (total === 0 || values.length === 0) {
+            logger.warn('Assets AQL returned zero results for variant', { qlQuery, total, storeNumber: normalized });
+            continue;
+        }
+
+        const objectId = extractObjectIdFromResponse(data);
+        if (!objectId) {
+            logger.warn('Assets AQL returned results but no extractable id', {
+                qlQuery,
+                storeNumber: normalized,
+                valuesSample: values[0]
+            });
+            continue;
+        }
+
+        logger.info('Resolved store via service-account AQL fallback', { storeNumber: normalized, objectId: String(objectId) });
+        return buildStoreFieldRef(objectId);
+    }
+
+    // All three paths missed — surface a message that tells the caller which
+    // routes were tried and what the cache looked like.
+    const cacheStatus = storesCache.status();
+    const total = lastResult?.data?.total ?? 'unknown';
+    const cacheHint = cacheStatus.assetsSyncConfigured
+        ? `cache has ${cacheStatus.storeCount} entries, last synced ${cacheStatus.lastSyncAt || 'never'}`
+        : 'cache is not populated (ASSETS_SYNC_EMAIL/ASSETS_SYNC_TOKEN not configured)';
+    throw new Error(
+        `Failed to resolve Store Number ${normalized} via Assets (objectTypeId=${objectTypeId}). ` +
+        `Tried: local cache, live PAT lookup, service-account AQL. ${cacheHint}. Last service-account total=${total}. ` +
+        `Try POST /api/wxccai/admin/storesCache/refresh to force a resync.`
+    );
+}
+
+/**
+ * Diagnostic helper: run several AQL variants for a given store number and
+ * return each result side-by-side, plus workspace/schema/object-type
+ * introspection and a plain-English diagnosis. Intended for a dev-only debug
+ * endpoint.
+ */
+export async function probeAssetsForStore(rawStoreNumber, { extraVariants = [] } = {}) {
+    const raw = String(rawStoreNumber ?? '').trim();
+    const padded = raw ? raw.padStart(5, '0') : '';
+    const unpadded = raw.replace(/^0+/, '') || raw;
+
+    const objectTypeId = config.jira.assetsStoreObjectTypeId || '109';
+    const schemaId = config.jira.assetsStoreSchemaId;
+    const attribute = config.jira.assetsStoreNumberAttribute || 'Store Number';
+    const attrId = config.jira.assetsStoreNumberAttributeId;
+
+    // A. Workspace discovery — did we get one at all? What are ALL visible workspaces?
+    const workspacesList = await listAssetsWorkspacesRaw();
+    let workspaceIdResolved = null;
+    let workspaceIdError = null;
+    try {
+        workspaceIdResolved = await getAssetsWorkspaceId();
+    } catch (e) {
+        workspaceIdError = e.message;
+    }
+
+    // B. Schemas the token can actually see. If empty, permissions are the issue.
+    let schemasProbe = null;
+    if (workspaceIdResolved) {
+        const schemasResp = await listAssetsSchemas();
+        let visibleSchemas = [];
+        const data = schemasResp.data;
+        const list = Array.isArray(data) ? data
+            : Array.isArray(data?.values) ? data.values
+            : Array.isArray(data?.objectschemas) ? data.objectschemas
+            : Array.isArray(data?.objectSchemas) ? data.objectSchemas
+            : [];
+        visibleSchemas = list.map(s => ({
+            id: s.id ?? null,
+            name: s.name ?? null,
+            objectSchemaKey: s.objectSchemaKey ?? s.key ?? null
+        }));
+        schemasProbe = {
+            httpStatus: schemasResp.status,
+            requestUrl: schemasResp.requestUrl,
+            count: visibleSchemas.length,
+            schemas: visibleSchemas,
+            raw: schemasResp.status === 200 ? undefined : schemasResp.data,
+            headers: schemasResp.headers
+        };
+    }
+
+    // C. Object type detail — is object type 109 visible? What are its attributes actually called?
+    let objectTypeProbe = null;
+    if (workspaceIdResolved) {
+        const { detail, attributes } = await getAssetsObjectType(objectTypeId);
+        const attrList = Array.isArray(attributes.data) ? attributes.data
+            : Array.isArray(attributes.data?.values) ? attributes.data.values
+            : [];
+        objectTypeProbe = {
+            detail: {
+                httpStatus: detail.status,
+                requestUrl: detail.requestUrl,
+                name: detail.data?.name ?? null,
+                objectSchemaId: detail.data?.objectSchemaId ?? null,
+                raw: detail.status === 200 ? { id: detail.data?.id, name: detail.data?.name, objectSchemaId: detail.data?.objectSchemaId, description: detail.data?.description } : detail.data,
+                headers: detail.headers
+            },
+            attributes: {
+                httpStatus: attributes.status,
+                requestUrl: attributes.requestUrl,
+                count: attrList.length,
+                names: attrList.map(a => ({
+                    id: a.id ?? null,
+                    name: a.name ?? null,
+                    type: a.type ?? a.defaultType?.name ?? null,
+                    system: a.system ?? null
+                })),
+                raw: attributes.status === 200 ? undefined : attributes.data,
+                headers: attributes.headers
+            }
+        };
+    }
+
+    // -------- AQL variants --------
+    const variants = [];
+
+    if (schemaId) {
+        variants.push({ label: 'schema_probe', qlQuery: `objectSchemaId = ${schemaId}`, resultPerPage: 5 });
+    }
+    variants.push({ label: 'object_type_probe', qlQuery: `objectTypeId = ${objectTypeId}`, resultPerPage: 5 });
+    variants.push({ label: 'object_type_by_name', qlQuery: `objectType = "Store"`, resultPerPage: 5 });
+
+    if (raw) {
+        variants.push({ label: 'attr_name_padded', qlQuery: `objectTypeId = ${objectTypeId} AND "${attribute}" = "${padded}"`, resultPerPage: 3 });
+        variants.push({ label: 'attr_name_unpadded', qlQuery: `objectTypeId = ${objectTypeId} AND "${attribute}" = "${unpadded}"`, resultPerPage: 3 });
+        variants.push({ label: 'attr_name_like_padded', qlQuery: `objectTypeId = ${objectTypeId} AND "${attribute}" LIKE "${padded}"`, resultPerPage: 3 });
+
+        variants.push({ label: 'name_padded', qlQuery: `objectTypeId = ${objectTypeId} AND Name = "${padded}"`, resultPerPage: 3 });
+        variants.push({ label: 'name_unpadded', qlQuery: `objectTypeId = ${objectTypeId} AND Name = "${unpadded}"`, resultPerPage: 3 });
+        variants.push({ label: 'name_like_padded', qlQuery: `objectTypeId = ${objectTypeId} AND Name LIKE "${padded}"`, resultPerPage: 3 });
+
+        if (schemaId) {
+            variants.push({ label: 'schema_name_padded', qlQuery: `objectSchemaId = ${schemaId} AND Name = "${padded}"`, resultPerPage: 3 });
+            variants.push({ label: 'schema_name_unpadded', qlQuery: `objectSchemaId = ${schemaId} AND Name = "${unpadded}"`, resultPerPage: 3 });
+        }
+
+        if (attrId) {
+            variants.push({ label: 'attr_id_padded', qlQuery: `objectTypeId = ${objectTypeId} AND attribute[${attrId}] = "${padded}"`, resultPerPage: 3 });
+            variants.push({ label: 'attr_id_unpadded', qlQuery: `objectTypeId = ${objectTypeId} AND attribute[${attrId}] = "${unpadded}"`, resultPerPage: 3 });
+        }
+    }
+
+    for (const v of extraVariants) {
+        variants.push({ label: v.label || 'custom', qlQuery: v.qlQuery, resultPerPage: v.resultPerPage ?? 5 });
+    }
+
+    const results = [];
+    for (const v of variants) {
+        const r = await runAssetsAql(v.qlQuery, { resultPerPage: v.resultPerPage, includeAttributes: true });
+        const values = Array.isArray(r.data?.values) ? r.data.values : [];
+        results.push({
+            variant: v.label,
+            qlQuery: v.qlQuery,
+            httpStatus: r.status,
+            statusText: r.statusText,
+            total: r.data?.total ?? values.length,
+            objects: values.map(summarizeAssetsObject),
+            atlassianError: r.status === 200 ? undefined : r.data,
+            headers: r.headers,
+            // Truncated raw body so we can see everything Atlassian sent back
+            // (some tenants surface hints in "hasMoreResults", "objectTypeAttributes", etc.)
+            rawBody: r.data && typeof r.data === 'object'
+                ? JSON.parse(JSON.stringify(r.data))
+                : r.data
+        });
+    }
+
+    const diagnosis = buildAssetsProbeDiagnosis({
+        workspaceIdResolved,
+        workspaceIdError,
+        workspacesList,
+        schemasProbe,
+        objectTypeProbe,
+        variantResults: results,
+        configuredAttribute: attribute,
+        configuredAttributeId: attrId,
+        configuredSchemaId: schemaId,
+        configuredObjectTypeId: objectTypeId,
+        jiraEmail: config.jira.email
+    });
+
+    return {
+        input: { raw, padded, unpadded },
+        config: {
+            workspaceId: workspaceIdResolved || `error: ${workspaceIdError}`,
+            objectTypeId,
+            schemaId: schemaId || null,
+            attribute,
+            attrId: attrId || null,
+            storeCustomFieldId: config.jira.storeCustomFieldId || 'customfield_10261',
+            authType: config.jira.authType,
+            jiraEmail: config.jira.email ? maskEmail(config.jira.email) : null
+        },
+        workspace: {
+            resolvedId: workspaceIdResolved,
+            error: workspaceIdError,
+            allVisible: workspacesList.workspaces,
+            httpStatus: workspacesList.httpStatus
+        },
+        schemas: schemasProbe,
+        objectType: objectTypeProbe,
+        variants: results,
+        diagnosis
+    };
+}
+
+function maskEmail(email) {
+    if (!email || !email.includes('@')) return email || null;
+    const [local, domain] = email.split('@');
+    const shown = local.length <= 3 ? local[0] : `${local.slice(0, 3)}…`;
+    return `${shown}@${domain}`;
+}
+
+function buildAssetsProbeDiagnosis({
+    workspaceIdResolved,
+    workspaceIdError,
+    workspacesList,
+    schemasProbe,
+    objectTypeProbe,
+    variantResults,
+    configuredAttribute,
+    // eslint-disable-next-line no-unused-vars
+    configuredAttributeId,
+    configuredSchemaId,
+    configuredObjectTypeId,
+    jiraEmail
+}) {
+    const notes = [];
+    const suggestions = [];
+    let likelyCause = 'unknown';
+
+    const visibleWorkspaces = workspacesList?.workspaces || [];
+    const email = jiraEmail || '(JIRA_EMAIL)';
+
+    if (!workspaceIdResolved) {
+        likelyCause = 'workspace_not_discovered';
+        notes.push(`Could not discover Assets workspace id (${workspaceIdError}).`);
+        suggestions.push('Set JIRA_ASSETS_WORKSPACE_ID explicitly, or ensure /rest/servicedeskapi/assets/workspace is reachable.');
+        return { likelyCause, notes, suggestions };
+    }
+
+    if (visibleWorkspaces.length > 1) {
+        notes.push(`Account can see ${visibleWorkspaces.length} Assets workspaces: ${visibleWorkspaces.map(w => w.workspaceId).join(', ')}. Using ${workspaceIdResolved}.`);
+        suggestions.push('If the Store schema lives in a different workspace, set JIRA_ASSETS_WORKSPACE_ID explicitly.');
+    }
+
+    const schemaHttp = schemasProbe?.httpStatus;
+    const schemaCount = schemasProbe?.count ?? 0;
+
+    if (schemaHttp && schemaHttp !== 200) {
+        likelyCause = 'schema_list_error';
+        notes.push(`GET /objectschema/list returned HTTP ${schemaHttp}. The token cannot list schemas.`);
+        if (schemaHttp === 401 || schemaHttp === 403) {
+            suggestions.push(`Add ${email} to an Object Schema role on the target schema in Jira → Assets → Object schemas → Configure → Roles. In Assets, API-token scopes (read:cmdb-*:jira) are NOT sufficient on their own; the user still needs schema-level role membership.`);
+        }
+        return { likelyCause, notes, suggestions };
+    }
+
+    if (schemaCount === 0) {
+        likelyCause = 'no_schema_visibility';
+        notes.push('GET /objectschema/list returned HTTP 200 with 0 schemas — this account has no visibility to any Assets schema, so every AQL against it returns total=0.');
+        suggestions.push(`Ask a Jira Assets admin to add ${email} to a role on the Store schema in Jira → Assets → Object schemas → Configure → Roles. "Object Schema Viewer" or "Object Schema User" is enough to look up store objects; "Developer" is needed to create/update objects.`);
+        suggestions.push('The four read:cmdb-* / write:cmdb-* scopes on the token are necessary but not sufficient — Assets enforces a separate per-schema role check on top of the OAuth scopes.');
+        return { likelyCause, notes, suggestions };
+    }
+
+    const visibleSchemaIds = (schemasProbe?.schemas || []).map(s => String(s.id));
+    const visibleSchemaSummary = (schemasProbe?.schemas || []).map(s => `${s.id}:${s.name}`).join(', ');
+    notes.push(`Account can see ${schemaCount} schema(s): ${visibleSchemaSummary}.`);
+
+    if (configuredSchemaId && !visibleSchemaIds.includes(String(configuredSchemaId))) {
+        likelyCause = 'schema_not_visible';
+        notes.push(`Configured JIRA_ASSETS_STORE_SCHEMA_ID=${configuredSchemaId} is NOT in the list of schemas this account can see. AQL against schema ${configuredSchemaId} will always return total=0.`);
+        suggestions.push(`Ask a Jira Assets admin to add ${email} to a role on the Store schema (id ${configuredSchemaId}) in Jira → Assets → Object schemas → Configure → Roles. "Object Schema Viewer" is enough for read; "Developer" for writes.`);
+        suggestions.push('Reminder: Jira Assets enforces per-schema role membership on top of OAuth scopes. Granting the token the read:cmdb-* / write:cmdb-* scopes is necessary but NOT sufficient — the underlying user must also be in a role on the schema.');
+        if (visibleSchemaIds.length === 1) {
+            suggestions.push(`Right now the account is only in a role on schema ${visibleSchemaIds[0]} (${visibleSchemaSummary}). Same admin action needs to happen for the Store schema.`);
+        }
+        return { likelyCause, notes, suggestions };
+    }
+
+    const otDetailStatus = objectTypeProbe?.detail?.httpStatus;
+    const otAttrStatus = objectTypeProbe?.attributes?.httpStatus;
+
+    if (otDetailStatus && otDetailStatus !== 200) {
+        if (otDetailStatus === 403) {
+            likelyCause = 'object_type_forbidden';
+            notes.push(`GET /objecttype/${configuredObjectTypeId} returned HTTP 403. The account can list schemas but cannot see this object type — almost always because it is missing an Object Schema role on the Store schema.`);
+            suggestions.push(`Add ${email} to an Object Schema role on the Store schema in Jira → Assets → Object schemas → Configure → Roles.`);
+        } else {
+            likelyCause = 'object_type_not_visible';
+            notes.push(`GET /objecttype/${configuredObjectTypeId} returned HTTP ${otDetailStatus}. The configured objectTypeId is either wrong or not visible to this account.`);
+            suggestions.push(`Verify JIRA_ASSETS_STORE_OBJECT_TYPE_ID matches the actual Store type id in Jira Assets.`);
+        }
+        return { likelyCause, notes, suggestions };
+    }
+
+    if (otDetailStatus === 200) {
+        notes.push(`Object type is visible: ${objectTypeProbe.detail.name} (schema ${objectTypeProbe.detail.objectSchemaId}).`);
+    }
+
+    if (otAttrStatus === 200 && objectTypeProbe.attributes.count > 0) {
+        const attrNames = objectTypeProbe.attributes.names.map(a => a.name).filter(Boolean);
+        const attrMatch = attrNames.find(n => n.toLowerCase() === (configuredAttribute || '').toLowerCase());
+        if (!attrMatch) {
+            likelyCause = 'attribute_name_mismatch';
+            notes.push(`The configured attribute "${configuredAttribute}" is NOT among the object type's attributes. Actual attribute names: ${attrNames.join(', ')}.`);
+            const guess = attrNames.find(n => /store|number|store\s*id/i.test(n));
+            if (guess) suggestions.push(`Set JIRA_ASSETS_STORE_NUMBER_ATTRIBUTE="${guess}" (or use the id form via JIRA_ASSETS_STORE_NUMBER_ATTRIBUTE_ID).`);
+            else suggestions.push('Pick the correct attribute from the list above and set JIRA_ASSETS_STORE_NUMBER_ATTRIBUTE (or ...ATTRIBUTE_ID) accordingly.');
+            return { likelyCause, notes, suggestions };
+        }
+        notes.push(`Attribute "${configuredAttribute}" exists on the object type.`);
+    }
+
+    const anyHits = variantResults.some(v => v.total > 0);
+    if (!anyHits) {
+        likelyCause = 'value_format_mismatch';
+        notes.push('Schema and object type are visible but no AQL variant returned rows. Attribute values are likely stored in a form none of the variants matched.');
+        suggestions.push('Re-run the probe without a storeNumber to sample real Store objects: curl "http://localhost:1866/api/wxccai/debug/assetsProbe" — the object_type_probe row will show up to 5 real Store objects with their actual attribute values, so you can see how Store Number is stored (leading zeros, prefix, etc.).');
+    } else {
+        const winners = variantResults.filter(v => v.total > 0).map(v => v.variant);
+        likelyCause = 'success';
+        notes.push(`These variants returned rows: ${winners.join(', ')}. Lock resolveStoreAssetReference to the first one.`);
+    }
+
+    return { likelyCause, notes, suggestions };
+}
+
+/**
+ * Extract an Assets object id from a variety of AQL response shapes.
+ * Tries top-level, then common list keys, then a bounded deep search.
+ */
+function extractObjectIdFromResponse(respData) {
+    if (!respData || typeof respData !== 'object') return null;
+
+    if (respData.id) return respData.id;
+    if (respData.objectId) return respData.objectId;
+
+    const listKeys = ['values', 'objectEntries', 'objects', 'objectList', 'results', 'items'];
+    for (const key of listKeys) {
+        const list = respData[key];
+        if (Array.isArray(list)) {
+            for (const item of list) {
+                if (item && typeof item === 'object') {
+                    if (item.id) return item.id;
+                    if (item.objectId) return item.objectId;
+                    if (item.object && item.object.id) return item.object.id;
+                    if (item.attributes && item.attributes.id) return item.attributes.id;
+                }
+            }
+        }
+    }
+
+    function deepFind(obj, depth = 0) {
+        if (depth > 6 || obj == null || typeof obj !== 'object') return null;
+        if (obj.id && (typeof obj.id === 'string' || typeof obj.id === 'number')) return obj.id;
+        if (obj.objectId && (typeof obj.objectId === 'string' || typeof obj.objectId === 'number')) return obj.objectId;
+        if (Array.isArray(obj)) {
+            for (const el of obj) {
+                const found = deepFind(el, depth + 1);
+                if (found) return found;
+            }
+        } else {
+            for (const k of Object.keys(obj)) {
+                const found = deepFind(obj[k], depth + 1);
+                if (found) return found;
+            }
+        }
+        return null;
+    }
+
+    return deepFind(respData);
+}

src/services/jira/assetsSyncClient.js

@@ -0,0 +1,150 @@
+// Dedicated Jira Assets client that authenticates as a *personal* Atlassian
+// account (email + PAT), not the shared service account used everywhere else.
+//
+// Why this exists:
+//   The service account has schema-level read on the Stores schema but is
+//   silently filtered out for objects in Object Type 109 (Store Address /
+//   Hierarchy). Every direct API path (raw Assets, site gateway, servicedeskapi
+//   CMDB proxy) is either 403 or returns total=0. A personal account with the
+//   right Assets role sees the objects fine, so this client is used *only* by
+//   the store cache sync — never for anything that mutates state on Atlassian's
+//   side.
+//
+// Credentials come from config.jira.assetsSyncEmail / assetsSyncToken. Those
+// are populated from env vars (which are typically exported from macOS
+// Keychain via bin/load-assets-sync-secret.sh). If either is missing this
+// module is inert and the caller falls back to the service account path.
+import axios from 'axios';
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+
+const ASSETS_HOST = 'https://api.atlassian.com';
+
+let cachedAuthHeader = null;
+let warnedMissing = false;
+
+function buildAuthHeader() {
+    const email = config.jira.assetsSyncEmail;
+    const token = config.jira.assetsSyncToken;
+    if (!email || !token) return null;
+    const encoded = Buffer.from(`${email}:${token}`).toString('base64');
+    return `Basic ${encoded}`;
+}
+
+function getAuthHeader() {
+    if (cachedAuthHeader) return cachedAuthHeader;
+    cachedAuthHeader = buildAuthHeader();
+    if (!cachedAuthHeader && !warnedMissing) {
+        logger.warn('Assets sync credentials not configured (ASSETS_SYNC_EMAIL / ASSETS_SYNC_TOKEN). Stores cache sync is disabled; resolveStoreAssetReference will fall back to the service-account AQL path.');
+        warnedMissing = true;
+    }
+    return cachedAuthHeader;
+}
+
+/**
+ * True iff both an email and a PAT are configured for the personal-account
+ * Assets sync path.
+ */
+export function isAssetsSyncConfigured() {
+    return !!getAuthHeader();
+}
+
+/**
+ * Best-effort masked identity for logging / status endpoints. Never returns
+ * the token.
+ */
+export function describeSyncIdentity() {
+    const email = config.jira.assetsSyncEmail || null;
+    return {
+        configured: isAssetsSyncConfigured(),
+        email: email
+            ? (email.includes('@')
+                ? `${email.slice(0, 3)}…@${email.split('@')[1]}`
+                : `${email.slice(0, 3)}…`)
+            : null,
+    };
+}
+
+/**
+ * Low-level AQL POST via the personal PAT. Never throws on non-2xx — always
+ * returns `{ status, statusText, data, headers, requestUrl, requestBody, error }`.
+ *
+ * `qlQuery`  raw AQL string, e.g. `objectTypeId = 109`
+ * `opts`     { startAt, maxResults, includeAttributes, extraBody, timeoutMs }
+ *
+ * Endpoint reference: Atlassian Assets REST API v1 `POST /object/aql`
+ *   - Query params: `startAt` (default 0), `maxResults` (default 25, cap
+ *     varies by tenant but 500 is safe), `includeAttributes` (default true)
+ *   - Body: `{ "qlQuery": "..." }`
+ *   - Response: `{ startAt, maxResults, total, isLast, values: [...] }`
+ *
+ * The old `page` / `resultPerPage` params are for a different endpoint and
+ * are silently ignored here — always use startAt/maxResults on this one.
+ */
+export async function assetsAql(qlQuery, opts = {}) {
+    const auth = getAuthHeader();
+    if (!auth) {
+        return {
+            status: 'not_configured',
+            statusText: 'ASSETS_SYNC_EMAIL / ASSETS_SYNC_TOKEN not set',
+            data: null,
+            error: 'assets sync credentials not configured',
+        };
+    }
+
+    const workspaceId = config.jira.assetsWorkspaceId;
+    if (!workspaceId) {
+        return {
+            status: 'not_configured',
+            statusText: 'JIRA_ASSETS_WORKSPACE_ID not set',
+            data: null,
+            error: 'assets workspace id not configured',
+        };
+    }
+
+    const {
+        startAt = 0,
+        maxResults = 500,
+        includeAttributes = true,
+        extraBody = {},
+        timeoutMs = 30000,
+    } = opts;
+
+    const qp = new URLSearchParams({
+        startAt: String(startAt),
+        maxResults: String(maxResults),
+        includeAttributes: String(includeAttributes),
+    });
+    const url = `${ASSETS_HOST}/jsm/assets/workspace/${workspaceId}/v1/object/aql?${qp.toString()}`;
+    const body = { qlQuery, ...extraBody };
+
+    try {
+        const resp = await axios.post(url, body, {
+            headers: {
+                Authorization: auth,
+                'Content-Type': 'application/json',
+                Accept: 'application/json',
+            },
+            timeout: timeoutMs,
+            validateStatus: () => true,
+        });
+        return {
+            status: resp.status,
+            statusText: resp.statusText,
+            data: resp.data ?? null,
+            headers: resp.headers,
+            requestUrl: url,
+            requestBody: body,
+        };
+    } catch (err) {
+        return {
+            status: err.response?.status || 'network',
+            statusText: err.response?.statusText || err.code || 'error',
+            data: err.response?.data || null,
+            headers: err.response?.headers || {},
+            requestUrl: url,
+            requestBody: body,
+            error: err.message,
+        };
+    }
+}

src/services/jira/attachments.js

@@ -0,0 +1,154 @@
+// Jira attachments + Webex CC transcript conversion.
+// Downloads happen via downloadClient (own retry policy); uploads use jiraClient.
+import FormData from 'form-data';
+import logger from '../../utilities/logger.js';
+import { jiraClient, downloadClient } from './client.js';
+
+/**
+ * Shared helper: POST a Buffer as multipart attachment to the core Jira
+ * attachments endpoint.
+ *   - Uses jiraClient (correct base + auth).
+ *   - Spreads form.getHeaders() so boundary is set.
+ *   - Cleans any charset from Content-Type (prevents 415).
+ *   - Retry loop only around the API call (download is caller's job).
+ *   - Fail-fast on 401/403 (scope/perms).
+ */
+async function attachBufferToJira(jiraKey, fileBuffer, fileName) {
+    for (let attempt = 1; attempt <= 3; attempt++) {
+        try {
+            const form = new FormData();
+            form.append('file', fileBuffer, fileName);
+
+            const formHeaders = form.getHeaders();
+            if (formHeaders['content-type']) {
+                formHeaders['content-type'] = formHeaders['content-type'].replace(/;\s*charset=[^;]*/i, '');
+            }
+
+            const uploadPath = `/rest/api/3/issue/${jiraKey}/attachments`;
+
+            await jiraClient.post(uploadPath, form, {
+                headers: {
+                    'X-Atlassian-Token': 'no-check',
+                    ...formHeaders
+                },
+                timeout: 15000
+            });
+
+            logger.info('File attached successfully', { jiraKey, fileName, attempt });
+            return;
+        } catch (err) {
+            const status = err.response?.status;
+            logger.error('Jira file attach attempt failed', {
+                jiraKey,
+                fileName,
+                attempt,
+                status,
+                responseData: err.response?.data,
+                responseHeaders: err.response?.headers ? Object.fromEntries(
+                    Object.entries(err.response.headers).filter(([k]) => !k.toLowerCase().includes('auth'))
+                ) : undefined
+            });
+            if (status === 401 || status === 403) {
+                throw err;
+            }
+            if (attempt === 3) throw err;
+            await new Promise(r => setTimeout(r, attempt * 1500));
+        }
+    }
+}
+
+/**
+ * Download a file from the given URL (S3 pre-signed) *once* and attach using
+ * the core Jira attachments API. Download happens outside the retry loop
+ * because the signed URL expires (~1800s).
+ */
+export async function attachFileToJira(jiraKey, fileUrl, fileName) {
+    let fileBuffer;
+    try {
+        const dl = await downloadClient.get(fileUrl, {
+            responseType: 'arraybuffer'
+        });
+        fileBuffer = Buffer.from(dl.data);
+    } catch (dlErr) {
+        logger.error('Failed to download file from S3 for attachment (URL likely expired on replay)', {
+            jiraKey,
+            fileName,
+            url: fileUrl,
+            status: dlErr.response?.status,
+            message: dlErr.message
+        });
+        throw dlErr;
+    }
+
+    await attachBufferToJira(jiraKey, fileBuffer, fileName);
+}
+
+/**
+ * Convert a Webex-style JSON transcript to human-readable text.
+ * Returns null if the shape isn't recognized.
+ */
+function formatTranscriptToHumanReadable(data) {
+    if (!data || !Array.isArray(data.responseContents)) return null;
+    const lines = [];
+    lines.push(`Transcript`);
+    if (data.interactionId) lines.push(`Interaction ID: ${data.interactionId}`);
+    if (data.languageCode) lines.push(`Language: ${data.languageCode}`);
+    lines.push('');
+    for (const entry of data.responseContents) {
+        const res = entry.recognitionResult;
+        if (!res || !res.alternatives || !res.alternatives[0]) continue;
+        const role = (res.role || 'UNKNOWN').toUpperCase();
+        const alt = res.alternatives[0];
+        const transcript = (alt.transcript || '').trim();
+        if (!transcript) continue;
+        let ts = '';
+        const words = alt.words || [];
+        if (words.length > 0) {
+            const start = words[0].start_time || {};
+            const totalSec = (start.seconds || 0) + Math.floor((start.nanos || 0) / 1e9);
+            const min = Math.floor(totalSec / 60);
+            const sec = Math.floor(totalSec % 60);
+            ts = `[${String(min).padStart(2, '0')}:${String(sec).padStart(2, '0')}] `;
+        }
+        lines.push(`${ts}${role}: ${transcript}`);
+    }
+    return lines.join('\n');
+}
+
+async function fetchAndConvertTranscript(url) {
+    try {
+        const resp = await downloadClient.get(url, { timeout: 10000 });
+        return formatTranscriptToHumanReadable(resp.data);
+    } catch (e) {
+        logger.warn(`Failed to fetch/convert transcript: ${e.message}`);
+        return null;
+    }
+}
+
+/**
+ * Download the JSON transcript, convert to human-readable text, and attach as
+ * `<base>-readable.txt`. Failures are swallowed so they don't mark the
+ * original JSON attach as failed.
+ */
+export async function attachReadableTranscript(jiraKey, transcriptUrl, originalFileName = null) {
+    const readable = await fetchAndConvertTranscript(transcriptUrl);
+    if (!readable) {
+        logger.warn('Readable transcript conversion yielded no content (check transcript JSON shape or URL)', { jiraKey });
+        return false;
+    }
+    const base = (originalFileName || `transcript-${jiraKey}`).replace(/\.json$/i, '');
+    const fileName = `${base}-readable.txt`;
+
+    try {
+        await attachBufferToJira(jiraKey, Buffer.from(readable, 'utf8'), fileName);
+        return true;
+    } catch (err) {
+        logger.error('Readable transcript attach failed (non-fatal)', {
+            jiraKey,
+            fileName,
+            error: err.response?.data?.message || err.message,
+            status: err.response?.status
+        });
+        return false;
+    }
+}

src/services/jira/client.js

@@ -0,0 +1,84 @@
+// Foundational Jira module: shared axios instances + tiny ADF helper.
+// Nothing else in services/jira/* should import axios or axios-retry directly.
+import axios from 'axios';
+import axiosRetry from 'axios-retry';
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+
+// Reusable Jira client with flexible auth. No default Content-Type on the
+// instance because callers need both JSON bodies AND multipart/form-data
+// (attachments). Content-Type is set explicitly per-request when needed.
+const createJiraClient = () => {
+    const headers = {};
+    const effectiveBase = config.jira.baseUrl || '(not configured)';
+    logger.debug(`[jira] baseUrl=${effectiveBase} authType=${config.jira.authType}`);
+
+    if (config.jira.authType === 'bearer') {
+        headers.Authorization = `Bearer ${config.jira.apiToken}`;
+        logger.info('Using Jira Bearer Token authentication');
+    } else {
+        const authStr = `${config.jira.email}:${config.jira.apiToken}`;
+        headers.Authorization = `Basic ${Buffer.from(authStr).toString('base64')}`;
+        logger.info('Using Jira Basic Auth');
+    }
+
+    const client = axios.create({
+        baseURL: config.jira.baseUrl,
+        headers,
+    });
+
+    // Retry policy scoped to this client only. Never mutates the default
+    // axios instance — anything else that needs retries uses its own client.
+    axiosRetry(client, {
+        retries: 3,
+        retryDelay: (retryCount) => Math.pow(2, retryCount) * 1000,
+        retryCondition: (error) => {
+            return axiosRetry.isNetworkOrIdempotentRequestError(error) ||
+                error.response?.status === 429 ||
+                error.response?.status >= 500;
+        }
+    });
+
+    return client;
+};
+
+export const jiraClient = createJiraClient();
+
+// Dedicated instance for pre-signed S3 downloads (audio + transcript files
+// from Webex CC). Separate from jiraClient because:
+//   1. No baseURL — always pass the full pre-signed URL.
+//   2. No Authorization header — the S3 URL is already signed.
+//   3. We want retries — S3 pre-signed downloads are the flakiest thing
+//      in the pipeline (transient 5xx, TLS resets, TCP timeouts).
+export const downloadClient = axios.create({ timeout: 20000 });
+axiosRetry(downloadClient, {
+    retries: 3,
+    retryDelay: (retryCount) => Math.pow(2, retryCount) * 1000,
+    retryCondition: (error) => {
+        return axiosRetry.isNetworkOrIdempotentRequestError(error) ||
+            error.response?.status === 429 ||
+            error.response?.status >= 500;
+    }
+});
+
+/**
+ * Convert a plain string to a minimal ADF document.
+ * ADF is what /rest/api/3/... expects for description/comment bodies.
+ * Blank lines split paragraphs; single newlines become hardBreak nodes.
+ */
+export function plainTextToAdf(text) {
+    const safe = (text ?? '').toString();
+    if (!safe) {
+        return { version: 1, type: 'doc', content: [] };
+    }
+    const paragraphs = safe.split(/\n{2,}/).map(block => {
+        const parts = block.split('\n');
+        const content = [];
+        parts.forEach((line, idx) => {
+            if (line.length) content.push({ type: 'text', text: line });
+            if (idx < parts.length - 1) content.push({ type: 'hardBreak' });
+        });
+        return { type: 'paragraph', content };
+    });
+    return { version: 1, type: 'doc', content: paragraphs };
+}

src/services/jira/comments.js

@@ -0,0 +1,121 @@
+// Jira issue comments: fetch (normalized), add plain, and post structured
+// Webex CC summary blocks. All comment ADF construction lives here.
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+import { adfToPlainText } from '../../utilities/adfToPlainText.js';
+import { jiraClient, plainTextToAdf } from './client.js';
+
+/**
+ * Fetch public comments for an issue, normalized to
+ *   { author, body (plain text), created, createdIso }.
+ * Uses the core /rest/api/3/issue/{key}/comment endpoint (servicedeskapi's
+ * variant can require different auth/perms with the current cloudId + Basic
+ * auth setup).
+ */
+export async function fetchPublicComments(key) {
+    const url = `/rest/api/3/issue/${key}/comment`;
+    try {
+        const response = await jiraClient.get(url);
+        const values = response.data?.values || [];
+        return values.map(comment => ({
+            author: comment.author?.displayName || comment.author?.name || 'Unknown',
+            body: adfToPlainText(comment.body),
+            created: comment.created,
+            createdIso: typeof comment.created === 'string' ? comment.created : (comment.created?.iso8601 || comment.created || null)
+        }));
+    } catch (error) {
+        logger.warn('Failed to fetch public comments:', error.message);
+        return [];
+    }
+}
+
+/**
+ * Add a comment to an issue. `text` is plain; converted to ADF here.
+ * If `internal: true`, restricts visibility to `config.jira.commentVisibilityRole`
+ * (same behavior as postWebexSummaryComment).
+ */
+export async function addComment(key, text, { internal = false } = {}) {
+    if (!key || !/^[A-Z]+-\d+$/.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+    if (!text || !String(text).trim()) {
+        throw new Error('Comment text is required');
+    }
+
+    const payload = { body: plainTextToAdf(String(text)) };
+    if (internal) {
+        payload.visibility = { type: 'role', value: config.jira.commentVisibilityRole || 'Service Desk Team' };
+    }
+
+    try {
+        const { data } = await jiraClient.post(`/rest/api/3/issue/${key}/comment`, payload, {
+            headers: { 'Content-Type': 'application/json' }
+        });
+        logger.info('Comment posted', { key, commentId: data?.id, internal });
+        return { key, commentId: data?.id, internal };
+    } catch (err) {
+        logger.error('addComment failed', { key, status: err.response?.status, details: err.response?.data });
+        const e = new Error(err.response?.data?.errorMessages?.join('; ') || err.message);
+        e.status = err.response?.status;
+        e.details = err.response?.data;
+        throw e;
+    }
+}
+
+/**
+ * Build a clean ADF comment from the Webex CC AI summaries object and post it
+ * to the Jira issue. Restricted visibility (via role) so only the configured
+ * role sees the summary + attachment references.
+ */
+export async function postWebexSummaryComment(jiraKey, summaries, attachedFiles = []) {
+    const items = [
+        { type: "listItem", content: [{ type: "paragraph", content: [{ type: "text", text: `Initial Contact Reason: ${summaries.intialContactReason || 'N/A'}` }] }] },
+        { type: "listItem", content: [{ type: "paragraph", content: [{ type: "text", text: `Additional Context: ${summaries.additionalContext || 'N/A'}` }] }] },
+        { type: "listItem", content: [{ type: "paragraph", content: [{ type: "text", text: `Key Actions Taken: ${summaries.keyActionsTake || 'N/A'}` }] }] },
+        { type: "listItem", content: [{ type: "paragraph", content: [{ type: "text", text: `Next Steps: ${summaries.nextSteps || 'N/A'}` }] }] },
+        { type: "listItem", content: [{ type: "paragraph", content: [{ type: "text", text: `Resolution: ${summaries.resolution || 'N/A'}` }] }] }
+    ];
+
+    const content = [
+        {
+            type: "heading",
+            attrs: { level: 3 },
+            content: [{ type: "text", text: "Webex Contact Center Summary" }]
+        },
+        { type: "bulletList", content: items },
+        {
+            type: "paragraph",
+            content: [{ type: "text", text: `Posted via Webex Integration — ${new Date().toISOString()}` }]
+        }
+    ];
+
+    if (Array.isArray(attachedFiles) && attachedFiles.length > 0) {
+        content.push({
+            type: "paragraph",
+            content: [{
+                type: "text",
+                text: `Attached files (internal): ${attachedFiles.join(', ')}`
+            }]
+        });
+    }
+
+    const commentPayload = {
+        body: {
+            version: 1,
+            type: "doc",
+            content
+        },
+        visibility: {
+            type: "role",
+            value: config.jira.commentVisibilityRole || 'Service Desk Team'
+        }
+    };
+
+    await jiraClient.post(
+        `/rest/api/3/issue/${jiraKey}/comment`,
+        commentPayload,
+        { headers: { 'Content-Type': 'application/json' } }
+    );
+
+    logger.info('Clean summary comment posted (restricted)', { jiraKey, attachedFiles });
+}

src/services/jira/issues.js

@@ -0,0 +1,290 @@
+// Jira issue lifecycle: fetch, search-by-reporter, status, update, transitions,
+// close. All project-agnostic (works for any project the token can see) and
+// uses the core /rest/api/3/issue/... endpoints.
+import logger from '../../utilities/logger.js';
+import config from '../../config/index.js';
+import { jiraClient, plainTextToAdf } from './client.js';
+import { fetchPublicComments } from './comments.js';
+
+const KEY_RE = /^[A-Z]+-\d+$/;
+
+export async function fetchJiraIssue(key) {
+    if (!key || !KEY_RE.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+
+    const url = `/rest/api/3/issue/${key}?fields=summary,status,assignee,created,updated,priority`;
+    try {
+        const response = await jiraClient.get(url);
+        return response.data;
+    } catch (error) {
+        logger.error('Fetch Jira issue failed:', error.response?.data || error.message);
+        throw new Error(`Issue fetch failed: ${error.message}`);
+    }
+}
+
+export async function fetchPlainDescription(key) {
+    const payload = { expression: "issue.description.plainText", context: { issue: { key } } };
+    try {
+        const response = await jiraClient.post('/rest/api/3/expression/evaluate', payload, {
+            headers: { 'Content-Type': 'application/json' }
+        });
+        return response.data.value || 'No description available.';
+    } catch (error) {
+        logger.warn('Failed to fetch plain description:', error.message);
+        return 'No description available.';
+    }
+}
+
+// Kept for future use (email → accountId lookup). Not currently called; the
+// reporter search uses email directly in JQL.
+// eslint-disable-next-line no-unused-vars
+async function getAccountIdFromEmail(email) {
+    if (!email) {
+        throw new Error('Email is required');
+    }
+
+    const url = `/rest/api/3/user/search?query=${encodeURIComponent(email)}&maxResults=10`;
+    try {
+        const response = await jiraClient.get(url);
+        const users = response.data || [];
+
+        if (users.length === 0) {
+            throw new Error(`No users found matching "${email}"`);
+        }
+
+        let user = users.find(u => u.emailAddress?.toLowerCase() === email.toLowerCase());
+        if (!user && users.length > 0) {
+            user = users[0];
+        }
+
+        if (!user?.accountId) {
+            throw new Error(`No usable accountId found for "${email}"`);
+        }
+
+        logger.info(`Using accountId ${user.accountId} for email "${email}"`);
+        return user.accountId;
+    } catch (err) {
+        throw new Error(`User lookup failed: ${err.message}`);
+    }
+}
+
+/**
+ * Search open tickets reported by a given email (across CS/SS/SUPPORT
+ * projects). Enriches each result with plain-text description + last 6
+ * public comments so downstream (Grok) has full context in one round trip.
+ */
+export async function searchOpenTicketsByReporterEmail(email) {
+    if (!email || typeof email !== 'string' || email.trim() === '') {
+        throw new Error("Email is required");
+    }
+
+    const jql = `project in (CS, SS, SUPPORT) 
+                 AND reporter = "${email}" 
+                 AND statusCategory != Done 
+                 ORDER BY updated DESC`;
+
+    try {
+        const response = await jiraClient.post('/rest/api/3/search/jql', {
+            jql: jql,
+            maxResults: 8,
+            fields: ["key", "summary", "status", "updated", "description"],
+            expand: "comments"
+        }, {
+            headers: { 'Content-Type': 'application/json' }
+        });
+
+        const issues = response.data.issues || [];
+
+        const enrichedIssues = await Promise.all(
+            issues.map(async (issue) => {
+                const key = issue.key;
+                try {
+                    const [plainDesc, publicComments] = await Promise.all([
+                        fetchPlainDescription(key).catch(() => "No description available."),
+                        fetchPublicComments(key).catch(() => [])
+                    ]);
+
+                    issue.enrichedNotes = {
+                        description: plainDesc,
+                        publicComments: publicComments.slice(-6)
+                    };
+                } catch (err) {
+                    logger.warn(`Failed to enrich notes for ${key}:`, err.message);
+                    issue.enrichedNotes = { description: "Notes unavailable.", publicComments: [] };
+                }
+                return issue;
+            })
+        );
+
+        return enrichedIssues;
+    } catch (error) {
+        logger.error('Jira reporter search failed:', error.response?.data || error.message);
+        throw new Error(`Failed to search tickets reported by ${email}: ${error.message}`);
+    }
+}
+
+/**
+ * Compact, purpose-built status view — no Grok, no comment enrichment.
+ * Use this when a caller just wants "where is this ticket right now?".
+ */
+export async function getTicketStatus(key) {
+    if (!key || !KEY_RE.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+
+    const url = `/rest/api/3/issue/${key}?fields=summary,status,assignee,reporter,priority,resolution,created,updated,labels`;
+    try {
+        const { data } = await jiraClient.get(url);
+        const f = data.fields || {};
+        return {
+            key: data.key,
+            summary: f.summary || null,
+            status: f.status?.name || null,
+            statusCategory: f.status?.statusCategory?.key || null,
+            assignee: f.assignee?.displayName || f.assignee?.emailAddress || null,
+            reporter: f.reporter?.displayName || f.reporter?.emailAddress || null,
+            priority: f.priority?.name || null,
+            resolution: f.resolution?.name || null,
+            labels: f.labels || [],
+            created: f.created || null,
+            updated: f.updated || null
+        };
+    } catch (err) {
+        logger.error('getTicketStatus failed', { key, status: err.response?.status, details: err.response?.data });
+        const e = new Error(`Failed to fetch status for ${key}: ${err.message}`);
+        e.status = err.response?.status;
+        e.details = err.response?.data;
+        throw e;
+    }
+}
+
+/**
+ * Partial issue update. Accepts flat, friendly fields:
+ *   { summary, description, priority, labels, assigneeAccountId, additional }
+ * `additional` is merged raw into the `fields` object (e.g. customfield_*).
+ * `description` is a plain string; converted to ADF here.
+ */
+export async function updateTicket(key, updates = {}) {
+    if (!key || !KEY_RE.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+
+    const fields = {};
+    if (updates.summary !== undefined) fields.summary = String(updates.summary);
+    if (updates.description !== undefined) fields.description = plainTextToAdf(updates.description);
+    if (updates.priority) fields.priority = { name: String(updates.priority) };
+    if (Array.isArray(updates.labels)) fields.labels = updates.labels.map(String);
+    if (updates.assigneeAccountId) fields.assignee = { accountId: String(updates.assigneeAccountId) };
+    if (updates.additional && typeof updates.additional === 'object') Object.assign(fields, updates.additional);
+
+    if (Object.keys(fields).length === 0) {
+        throw new Error('updateTicket called with no updatable fields');
+    }
+
+    try {
+        await jiraClient.put(`/rest/api/3/issue/${key}`, { fields }, {
+            headers: { 'Content-Type': 'application/json' }
+        });
+        logger.info('Ticket updated', { key, fieldKeys: Object.keys(fields) });
+        return { key, updated: Object.keys(fields) };
+    } catch (err) {
+        logger.error('updateTicket failed', { key, status: err.response?.status, details: err.response?.data });
+        const e = new Error(err.response?.data?.errorMessages?.join('; ') || err.message);
+        e.status = err.response?.status;
+        e.details = err.response?.data;
+        throw e;
+    }
+}
+
+/**
+ * Fetch available workflow transitions for an issue. Useful for both the
+ * client picking a transition manually and for closeTicket() below.
+ */
+export async function getTransitions(key) {
+    if (!key || !KEY_RE.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+    try {
+        const { data } = await jiraClient.get(`/rest/api/3/issue/${key}/transitions`);
+        return (data.transitions || []).map(t => ({
+            id: t.id,
+            name: t.name,
+            to: { id: t.to?.id, name: t.to?.name, statusCategory: t.to?.statusCategory?.key },
+            hasScreen: !!t.hasScreen
+        }));
+    } catch (err) {
+        logger.error('getTransitions failed', { key, status: err.response?.status, details: err.response?.data });
+        throw new Error(`Failed to fetch transitions for ${key}: ${err.message}`);
+    }
+}
+
+/**
+ * Execute a specific transition. Optionally set a resolution and/or append a
+ * comment in the same call (both are fields the transition screen can accept).
+ */
+export async function transitionTicket(key, transitionId, { resolution, comment, internal = false, additionalFields } = {}) {
+    if (!key || !KEY_RE.test(key)) {
+        throw new Error(`Invalid ticket key: "${key}"`);
+    }
+    if (!transitionId) throw new Error('transitionId is required');
+
+    const payload = { transition: { id: String(transitionId) } };
+
+    const fields = { ...(additionalFields || {}) };
+    if (resolution) fields.resolution = { name: String(resolution) };
+    if (Object.keys(fields).length) payload.fields = fields;
+
+    if (comment) {
+        const commentEntry = { add: { body: plainTextToAdf(String(comment)) } };
+        if (internal) {
+            commentEntry.add.visibility = { type: 'role', value: config.jira.commentVisibilityRole || 'Service Desk Team' };
+        }
+        payload.update = { comment: [commentEntry] };
+    }
+
+    try {
+        await jiraClient.post(`/rest/api/3/issue/${key}/transitions`, payload, {
+            headers: { 'Content-Type': 'application/json' }
+        });
+        logger.info('Ticket transitioned', { key, transitionId, resolution });
+        return { key, transitionId, resolution: resolution || null };
+    } catch (err) {
+        logger.error('transitionTicket failed', {
+            key, transitionId, status: err.response?.status, details: err.response?.data
+        });
+        const e = new Error(err.response?.data?.errorMessages?.join('; ') || err.message);
+        e.status = err.response?.status;
+        e.details = err.response?.data;
+        throw e;
+    }
+}
+
+/**
+ * Convenience: find a "closing" transition and execute it.
+ * Prefers explicit `transitionName` if provided, otherwise picks the first
+ * transition whose target status is in category "done" (Jira's canonical
+ * category for closed/resolved/completed), falling back to a name-based match.
+ */
+export async function closeTicket(key, { transitionName, resolution = 'Done', comment, internal = false } = {}) {
+    const transitions = await getTransitions(key);
+    if (transitions.length === 0) {
+        throw new Error(`No workflow transitions available for ${key} (check assignee/permissions)`);
+    }
+
+    let chosen = null;
+    if (transitionName) {
+        chosen = transitions.find(t => t.name.toLowerCase() === transitionName.toLowerCase());
+        if (!chosen) {
+            throw new Error(`Transition "${transitionName}" not available for ${key}. Available: ${transitions.map(t => t.name).join(', ')}`);
+        }
+    } else {
+        chosen = transitions.find(t => t.to?.statusCategory === 'done')
+            || transitions.find(t => /done|closed|resolved|complete/i.test(t.name));
+        if (!chosen) {
+            throw new Error(`Could not find a closing transition for ${key}. Available: ${transitions.map(t => t.name).join(', ')}`);
+        }
+    }
+
+    return transitionTicket(key, chosen.id, { resolution, comment, internal });
+}

src/services/jira/jsmRequests.js

@@ -0,0 +1,149 @@
+// Jira Service Management (JSM) request creation.
+// Uses /rest/servicedeskapi/request to create proper JSM customer requests
+// with request types. This is separate from the corporate-project ticket
+// flow (see issues.js). Depends on assets.js for store resolution.
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+import { jiraClient } from './client.js';
+import { resolveStoreAssetReference } from './assets.js';
+
+const REQUEST_TYPE_MAP = {
+    // Point of Sale
+    'Register Not functioning properly': 269,
+    'Unable to login': 275,
+    'Business report issue': 267,
+    'Broken device / hardware': 266,
+
+    // Hardware
+    'Broken Device / Hardware': 266,
+    'Report Missing Hardware': 273,
+    'Request Additional Hardware': 274,
+
+    // Technology
+    'Business Report Issue': 267,
+    'Report an Issue with Sterling Application': 272,
+    'Omni Turn Off / On': 268,
+    'Report a Traffic Counter Issue': 271,
+    'Report a Technology issue': 270,
+    'UKG Pro / Workforce Management Issues': 426,
+    'Store Transportation Request': 493,
+};
+
+// SubTypes that require a storeNumber. Every currently-supported subType maps
+// to a request type whose Store Number field is `required: true` in JSM (see
+// ss-fields-*.json). Kept as an explicit set so a future subType that does NOT
+// require Store Number can be added by simply omitting it from this set.
+const SUBTYPES_REQUIRING_STORE_NUMBER = new Set(Object.keys(REQUEST_TYPE_MAP));
+
+export function getSupportedSSSubTypes() {
+    return Object.keys(REQUEST_TYPE_MAP);
+}
+
+export { REQUEST_TYPE_MAP };
+
+/**
+ * Create a Store Support ticket (JSM request) using the Service Desk API.
+ * @param {Object} params
+ * @param {string} params.subType - Exact key from REQUEST_TYPE_MAP (e.g. "Register Not functioning properly")
+ * @param {string} [params.onBehalfOf] - email or accountId (becomes raiseOnBehalfOf)
+ * @param {string} params.summary
+ * @param {string} [params.description]
+ * @param {string|number} [params.storeNumber]
+ * @param {Object} [params.additional] - extra customfield_* values merged into requestFieldValues
+ */
+export async function createSSRequest(params = {}) {
+    const {
+        subType,
+        onBehalfOf,
+        summary,
+        description,
+        storeNumber,
+        additional = {}
+    } = params;
+
+    if (!subType || !summary) {
+        const err = new Error('subType and summary are required');
+        err.status = 400;
+        throw err;
+    }
+
+    const requestTypeId = REQUEST_TYPE_MAP[subType];
+    if (!requestTypeId) {
+        const err = new Error(`Unknown subType: "${subType}". Must be one of the supported values.`);
+        err.status = 400;
+        throw err;
+    }
+
+    // Fail-fast: every current subType requires Store Number. Catching this
+    // client-side gives a clean API error instead of forwarding to Jira and
+    // getting back an opaque "Please provide a value for required field
+    // 'Store Number'" that references Jira internals.
+    const normalizedStoreNumber = storeNumber != null && String(storeNumber).trim() !== ''
+        ? String(storeNumber).trim()
+        : null;
+    if (SUBTYPES_REQUIRING_STORE_NUMBER.has(subType) && !normalizedStoreNumber) {
+        const err = new Error(`storeNumber is required for subType "${subType}"`);
+        err.status = 400;
+        throw err;
+    }
+
+    const serviceDeskId = config.jira.serviceDeskId || '170';
+    const storeCustomField = config.jira.storeCustomFieldId || 'customfield_10261';
+
+    const requestFieldValues = {
+        summary,
+        description: description || summary,
+        ...additional
+    };
+
+    if (normalizedStoreNumber) {
+        const storeRef = await resolveStoreAssetReference(normalizedStoreNumber);
+        if (storeRef) {
+            requestFieldValues[storeCustomField] = storeRef;
+        }
+    }
+
+    const payload = {
+        serviceDeskId: String(serviceDeskId),
+        requestTypeId: String(requestTypeId),
+        requestFieldValues
+    };
+
+    if (onBehalfOf) {
+        payload.raiseOnBehalfOf = onBehalfOf;
+    }
+
+    try {
+        const response = await jiraClient.post(
+            '/rest/servicedeskapi/request',
+            payload,
+            {
+                headers: {
+                    'Content-Type': 'application/json',
+                    'Accept': 'application/json'
+                }
+            }
+        );
+
+        const data = response.data;
+        logger.info('SS ticket created successfully', {
+            issueKey: data?.issueKey,
+            subType,
+            storeNumber: String(storeNumber || '').padStart(5, '0')
+        });
+        return data;
+    } catch (err) {
+        const errData = err.response?.data || {};
+        const message = errData.errorMessage || errData.message || err.message || 'Unknown error creating SS request';
+        logger.error('Failed to create SS request', {
+            subType,
+            storeNumber,
+            status: err.response?.status,
+            details: errData
+        });
+        const error = new Error(message);
+        error.status = err.response?.status;
+        error.details = errData;
+        throw error;
+    }
+}

src/services/jira/storesCache.js

@@ -0,0 +1,328 @@
+// In-memory + on-disk cache of `storeNumber → Assets objectId`.
+//
+// Populated via assetsSyncClient (personal PAT). This is the primary lookup
+// path used by resolveStoreAssetReference — the raw service-account AQL call
+// is only a fallback for the (currently-blocked) scenario where the cache is
+// unavailable and the service account has been granted enough permission to
+// query the store type directly.
+//
+// Storage:
+//   In-memory: Map<paddedStoreNumber, entry>
+//   On-disk:   data/stores.json (see config.storesCache.path). Gitignored.
+//
+// Lifecycle:
+//   init()              — called at boot; loads disk cache and kicks off
+//                         background refresh if missing/stale, then schedules
+//                         periodic refresh via setInterval.
+//   refresh({force})    — full paginated resync via personal PAT.
+//   get(storeNumber)    — sync lookup, returns entry or null.
+//   status()            — for the admin endpoint.
+//
+// Store objects live in ObjectType 109 in schema 68. The `label` on those
+// objects is the padded store number (verified against real objects in the
+// tenant); the `Store Number` attribute is a secondary source we fall back to.
+import fs from 'node:fs/promises';
+import path from 'node:path';
+import config from '../../config/index.js';
+import logger from '../../utilities/logger.js';
+import { assetsAql, isAssetsSyncConfigured, describeSyncIdentity } from './assetsSyncClient.js';
+
+const state = {
+    cache: new Map(),
+    lastSyncAt: null,
+    lastError: null,
+    syncing: false,
+    scheduleTimer: null,
+    loadedFromDisk: false,
+};
+
+function normalizeStoreNumber(raw) {
+    if (raw === null || raw === undefined) return null;
+    const s = String(raw).trim();
+    if (!s) return null;
+    if (!/^\d+$/.test(s)) return null;
+    return s.padStart(5, '0');
+}
+
+function extractStoreNumberFromObject(obj, storeNumberAttrName) {
+    // Primary: the label is the padded store number on type 109 objects.
+    const label = obj?.label || obj?.name;
+    if (label && /^\d+$/.test(String(label).trim())) {
+        return String(label).trim().padStart(5, '0');
+    }
+    // Secondary: check the named attribute.
+    const attrs = Array.isArray(obj?.attributes) ? obj.attributes : [];
+    for (const attr of attrs) {
+        const meta = attr?.objectTypeAttribute || attr?.typeAttribute || {};
+        if (meta?.name === storeNumberAttrName) {
+            const vals = Array.isArray(attr?.objectAttributeValues) ? attr.objectAttributeValues : [];
+            for (const v of vals) {
+                const cand = v?.displayValue ?? v?.value ?? v?.searchValue;
+                if (cand && /^\d+$/.test(String(cand).trim())) {
+                    return String(cand).trim().padStart(5, '0');
+                }
+            }
+        }
+    }
+    return null;
+}
+
+/**
+ * Sync (in-memory) lookup. Returns null on miss; never throws.
+ */
+export function get(storeNumber) {
+    const n = normalizeStoreNumber(storeNumber);
+    if (!n) return null;
+    return state.cache.get(n) || null;
+}
+
+/**
+ * Snapshot for the admin status endpoint. Safe to serialize to JSON.
+ */
+export function status() {
+    const lastSyncEpoch = state.lastSyncAt ? new Date(state.lastSyncAt).getTime() : null;
+    return {
+        storeCount: state.cache.size,
+        lastSyncAt: state.lastSyncAt,
+        ageSeconds: lastSyncEpoch ? Math.floor((Date.now() - lastSyncEpoch) / 1000) : null,
+        syncing: state.syncing,
+        lastError: state.lastError,
+        loadedFromDisk: state.loadedFromDisk,
+        assetsSyncConfigured: isAssetsSyncConfigured(),
+        syncIdentity: describeSyncIdentity(),
+        cachePath: config.storesCache.path,
+        refreshIntervalHours: config.storesCache.refreshIntervalHours,
+        staleAfterHours: config.storesCache.staleAfterHours,
+    };
+}
+
+async function loadFromDisk() {
+    const cachePath = config.storesCache.path;
+    try {
+        const raw = await fs.readFile(cachePath, 'utf8');
+        const parsed = JSON.parse(raw);
+        if (parsed && typeof parsed === 'object' && parsed.entries) {
+            state.cache = new Map(Object.entries(parsed.entries));
+            state.lastSyncAt = parsed.lastSyncAt || null;
+            state.loadedFromDisk = true;
+            logger.info('Loaded stores cache from disk', {
+                path: cachePath,
+                storeCount: state.cache.size,
+                lastSyncAt: state.lastSyncAt,
+            });
+            return true;
+        }
+        logger.warn('Stores cache file present but shape unexpected; ignoring', { path: cachePath });
+    } catch (e) {
+        if (e.code === 'ENOENT') {
+            logger.info('No stores cache on disk yet; will populate on first sync', { path: cachePath });
+        } else {
+            logger.warn('Failed to load stores cache from disk', { path: cachePath, error: e.message });
+        }
+    }
+    return false;
+}
+
+async function saveToDisk() {
+    const cachePath = config.storesCache.path;
+    const dir = path.dirname(cachePath);
+    try {
+        await fs.mkdir(dir, { recursive: true });
+        const payload = {
+            version: 1,
+            lastSyncAt: state.lastSyncAt,
+            storeCount: state.cache.size,
+            entries: Object.fromEntries(state.cache),
+        };
+        // Atomic-ish write: to tmp then rename, avoids readers seeing a half-written file.
+        const tmp = `${cachePath}.tmp`;
+        await fs.writeFile(tmp, JSON.stringify(payload, null, 2), 'utf8');
+        await fs.rename(tmp, cachePath);
+        logger.debug('Persisted stores cache to disk', { path: cachePath, storeCount: state.cache.size });
+    } catch (e) {
+        logger.warn('Failed to persist stores cache to disk', { path: cachePath, error: e.message });
+    }
+}
+
+/**
+ * Full paginated resync of all objects in the configured store object type
+ * via the personal PAT. Rebuilds the in-memory cache atomically (only swaps
+ * once every page has been read successfully) and then writes to disk.
+ *
+ * Never called reentrantly — if a refresh is already in flight, returns
+ * { skipped: true, reason: 'already_syncing' }.
+ */
+export async function refresh({ force = false } = {}) {
+    if (state.syncing) {
+        logger.info('Stores cache refresh already in progress; ignoring duplicate call');
+        return { skipped: true, reason: 'already_syncing' };
+    }
+    if (!isAssetsSyncConfigured()) {
+        return { skipped: true, reason: 'not_configured' };
+    }
+
+    state.syncing = true;
+    state.lastError = null;
+    const started = Date.now();
+    const objectTypeId = config.jira.assetsStoreObjectTypeId || '109';
+    const storeNumberAttr = config.jira.assetsStoreNumberAttribute || 'Store Number';
+    const pageSize = config.storesCache.pageSize || 500;
+    const maxPages = config.storesCache.maxPages || 200; // safety cap; 200*500 = 100k stores
+
+    try {
+        const newCache = new Map();
+        // Atlassian Assets `/object/aql` paginates via startAt + maxResults +
+        // isLast. Some responses also cap `total` at 1000 as a hint, so we
+        // trust isLast (or an empty/short page) rather than the reported total.
+        let startAt = 0;
+        let page = 0;
+        let seenTotal = 0;
+        let expectedTotal = null;
+        let orphaned = 0;
+
+        while (page < maxPages) {
+            const r = await assetsAql(`objectTypeId = ${objectTypeId}`, {
+                startAt,
+                maxResults: pageSize,
+                includeAttributes: true,
+            });
+
+            if (r.status !== 200) {
+                const bodySnippet = typeof r.data === 'object'
+                    ? JSON.stringify(r.data).slice(0, 400)
+                    : String(r.data).slice(0, 400);
+                throw new Error(
+                    `Assets AQL page ${page + 1} (startAt=${startAt}) returned HTTP ${r.status} (${r.statusText}). Body: ${bodySnippet}`
+                );
+            }
+
+            const values = Array.isArray(r.data?.values) ? r.data.values : [];
+            const returnedMaxResults = typeof r.data?.maxResults === 'number' ? r.data.maxResults : pageSize;
+            if (expectedTotal === null && typeof r.data?.total === 'number') {
+                expectedTotal = r.data.total;
+            }
+
+            for (const obj of values) {
+                const storeNumber = extractStoreNumberFromObject(obj, storeNumberAttr);
+                if (!storeNumber || !obj?.id) {
+                    orphaned++;
+                    continue;
+                }
+                newCache.set(storeNumber, {
+                    objectId: String(obj.id),
+                    objectKey: obj.objectKey || null,
+                    label: obj.label || obj.name || null,
+                    syncedAt: new Date().toISOString(),
+                });
+            }
+            seenTotal += values.length;
+            page++;
+
+            const isLast = r.data?.isLast === true
+                || values.length === 0
+                || values.length < returnedMaxResults;
+
+            logger.debug('Stores cache sync page', {
+                page,
+                startAt,
+                returnedMaxResults,
+                rowsThisPage: values.length,
+                seenTotal,
+                expectedTotal,
+                mappedSoFar: newCache.size,
+                orphanedSoFar: orphaned,
+                isLast,
+            });
+
+            if (isLast) break;
+            startAt += values.length;
+        }
+
+        if (page >= maxPages) {
+            logger.warn('Stores cache sync hit page cap; some stores may be missing', {
+                maxPages,
+                pageSize,
+                seenTotal,
+            });
+        }
+
+        // Only swap once we've read every page successfully.
+        state.cache = newCache;
+        state.lastSyncAt = new Date().toISOString();
+        await saveToDisk();
+
+        const durationMs = Date.now() - started;
+        logger.info('Stores cache refresh complete', {
+            storeCount: state.cache.size,
+            orphaned,
+            pagesRead: page,
+            durationMs,
+            forced: force,
+        });
+
+        return {
+            ok: true,
+            storeCount: state.cache.size,
+            orphaned,
+            pagesRead: page,
+            durationMs,
+        };
+    } catch (e) {
+        state.lastError = { message: e.message, at: new Date().toISOString() };
+        logger.error('Stores cache refresh failed', { error: e.message });
+        throw e;
+    } finally {
+        state.syncing = false;
+    }
+}
+
+/**
+ * Load the disk cache, kick off a background refresh if missing or stale, and
+ * schedule periodic refreshes. Idempotent — safe to call more than once.
+ *
+ * Never throws. Individual failures are logged and surfaced through status().
+ */
+export async function init() {
+    if (!config.storesCache.enabled) {
+        logger.info('Stores cache disabled via config; skipping init');
+        return;
+    }
+
+    await loadFromDisk();
+
+    const staleAfterMs = (config.storesCache.staleAfterHours || 48) * 3600 * 1000;
+    const shouldRefreshNow = !state.lastSyncAt
+        || (Date.now() - new Date(state.lastSyncAt).getTime()) > staleAfterMs;
+
+    if (shouldRefreshNow && isAssetsSyncConfigured()) {
+        logger.info('Stores cache is missing or stale; starting background refresh');
+        refresh().catch(e => logger.error('Initial stores cache refresh failed', { error: e.message }));
+    } else if (shouldRefreshNow) {
+        logger.warn('Stores cache is missing or stale but assets sync is not configured; skipping initial refresh');
+    }
+
+    const intervalHours = config.storesCache.refreshIntervalHours;
+    if (intervalHours > 0) {
+        if (state.scheduleTimer) clearInterval(state.scheduleTimer);
+        state.scheduleTimer = setInterval(() => {
+            if (!isAssetsSyncConfigured()) return;
+            refresh().catch(e => logger.error('Scheduled stores cache refresh failed', { error: e.message }));
+        }, intervalHours * 3600 * 1000);
+        state.scheduleTimer.unref?.(); // don't block process exit for tests
+        logger.info('Stores cache: periodic refresh scheduled', { intervalHours });
+    }
+}
+
+/**
+ * Test / shutdown hook. Clears the schedule timer and the in-memory map.
+ * Does NOT delete the on-disk cache.
+ */
+export function shutdown() {
+    if (state.scheduleTimer) clearInterval(state.scheduleTimer);
+    state.scheduleTimer = null;
+    state.cache = new Map();
+    state.lastSyncAt = null;
+    state.lastError = null;
+    state.syncing = false;
+    state.loadedFromDisk = false;
+}