fix(worker): report RAM alongside GPU memory (#11167)

* fix(worker): report RAM alongside GPU memory

Assisted-by: Codex:gpt-5

* feat(ui): show worker RAM on node views

Assisted-by: Codex:gpt-5

---------

Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>

Author: localai-org-maint-bot

core/http/react-ui/e2e/nodes-per-node-backend-actions.spec.js

@@ -102,6 +102,9 @@ test.describe('Nodes page — per-node backend actions', () => {
     await mockDistributedNodes(page)
     await openNodeDetail(page)
 
+    await expect(page.locator('.node-detail__metrics')).toContainText('RAM')
+    await expect(page.locator('.node-detail__metrics')).toContainText('3.7 GB / 7.5 GB')
+
     // Negative: the old, ambiguous wording must not be used.
     await expect(page.locator('button[title="Reinstall backend"]')).toHaveCount(0)
     await expect(page.locator('button[title="Reinstall backend"] i.fa-sync-alt')).toHaveCount(0)

core/http/react-ui/e2e/nodes-roster.spec.js

@@ -26,6 +26,23 @@ test.describe('Nodes roster header', () => {
 })
 
 test.describe('Nodes roster panels', () => {
+  test('shows used and total system RAM reported by a worker', async ({ page }) => {
+    await mockCluster(page, [
+      {
+        id: 'n1',
+        name: 'alpha',
+        node_type: 'backend',
+        address: '10.0.0.1:50051',
+        status: 'healthy',
+        total_ram: 8_000_000_000,
+        available_ram: 3_000_000_000,
+      },
+    ])
+
+    await page.goto('/app/nodes')
+    await expect(page.locator('.node-panel').filter({ hasText: 'alpha' })).toContainText('RAM 4.7 GB / 7.5 GB', { timeout: 15_000 })
+  })
+
   test('shows model chips without clicking and filters by type', async ({ page }) => {
     await page.route('**/api/nodes', r => r.fulfill({ status: 200, contentType: 'application/json', body: JSON.stringify([
       { id: 'n1', name: 'alpha', node_type: 'backend', address: '10.0.0.1:50051', status: 'healthy' },

core/http/react-ui/src/components/nodes/NodePanel.jsx

@@ -9,6 +9,7 @@ export default function NodePanel({ node, models = [], onApprove, onDrain, onRes
   const isAgent = node.node_type === 'agent'
   const open = () => navigate(`/app/nodes/${node.id}`)
   const usedVRAM = node.total_vram && node.available_vram != null ? node.total_vram - node.available_vram : null
+  const usedRAM = node.total_ram && node.available_ram != null ? node.total_ram - node.available_ram : null
 
   return (
     <div className="node-panel">
@@ -45,6 +46,9 @@ export default function NodePanel({ node, models = [], onApprove, onDrain, onRes
               {node.total_vram > 0 && (
                 <span className="cell-mono">VRAM {formatVRAM(usedVRAM) || '0'} / {formatVRAM(node.total_vram)}</span>
               )}
+              {node.total_ram > 0 && (
+                <span className="cell-mono">RAM {formatVRAM(usedRAM) || '0'} / {formatVRAM(node.total_ram)}</span>
+              )}
               <span className="cell-mono">{node.in_flight_count || 0} in-flight</span>
             </div>
             <div className="node-panel__models">

core/http/react-ui/src/pages/NodeDetail.jsx

@@ -64,6 +64,7 @@ export default function NodeDetail() {
   const delLabel = async (k) => { try { await nodesApi.deleteLabel(id, k); refresh() } catch (e) { addToast(e.message, 'error') } }
 
   const usedVRAM = node.total_vram && node.available_vram != null ? node.total_vram - node.available_vram : 0
+  const usedRAM = node.total_ram && node.available_ram != null ? node.total_ram - node.available_ram : 0
   // {modelName: replicaCount} of loaded models so the shrink confirm can warn
   // if the new cap is below the actual count of any single model on this node.
   const loadedModelCounts = (() => {
@@ -88,14 +89,20 @@ export default function NodeDetail() {
         }
       />
 
-      {/* Inline metrics row: VRAM / in-flight - no boxes, just labelled values. */}
+      {/* Inline resource and activity metrics - no boxes, just labelled values. */}
       <div className="node-detail__metrics">
         {node.total_vram > 0 && (
           <div>
             <div className="drawer-eyebrow">VRAM</div>
             <span className="cell-mono">{formatVRAM(usedVRAM) || '0'} / {formatVRAM(node.total_vram)}</span>
           </div>
         )}
+        {node.total_ram > 0 && (
+          <div>
+            <div className="drawer-eyebrow">RAM</div>
+            <span className="cell-mono">{formatVRAM(usedRAM) || '0'} / {formatVRAM(node.total_ram)}</span>
+          </div>
+        )}
         {node.total_disk > 0 && (
           <div>
             {/* Free space on the worker's MODELS filesystem. A node can look

core/services/worker/registration.go

@@ -13,6 +13,12 @@ import (
 	"github.com/mudler/xlog"
 )
 
+var (
+	totalAvailableVRAM  = xsysinfo.TotalAvailableVRAM
+	getGPUAggregateInfo = xsysinfo.GetGPUAggregateInfo
+	getSystemRAMInfo    = xsysinfo.GetSystemRAMInfo
+)
+
 // effectiveBasePort returns the port used as base for gRPC backend processes.
 // Priority: Addr port → ServeAddr port → 50051
 func (cfg *Config) effectiveBasePort() int {
@@ -118,7 +124,7 @@ func (cfg *Config) registrationBody() map[string]any {
 	}
 
 	// Detect GPU info for VRAM-aware scheduling
-	totalVRAM, err := xsysinfo.TotalAvailableVRAM()
+	totalVRAM, err := totalAvailableVRAM()
 	if err != nil {
 		xlog.Debug("Failed to detect worker VRAM; registering without GPU capacity", "error", err)
 	}
@@ -179,15 +185,14 @@ func (cfg *Config) registrationBody() map[string]any {
 		body["vram_budget"] = cfg.VRAMBudget
 	}
 
-	// If no GPU detected, report system RAM so the scheduler/UI has capacity info
-	if totalVRAM == 0 {
-		ramInfo, err := xsysinfo.GetSystemRAMInfo()
-		if err != nil {
-			xlog.Debug("Failed to detect worker RAM for registration", "error", err)
-		} else {
-			body["total_ram"] = ramInfo.Total
-			body["available_ram"] = ramInfo.Available
-		}
+	// Report system RAM independently from VRAM so both discrete-GPU and
+	// unified-memory workers expose the capacity visible to the host.
+	ramInfo, err := getSystemRAMInfo()
+	if err != nil {
+		xlog.Debug("Failed to detect worker RAM for registration", "error", err)
+	} else {
+		body["total_ram"] = ramInfo.Total
+		body["available_ram"] = ramInfo.Available
 	}
 	if cfg.RegistrationToken != "" {
 		body["token"] = cfg.RegistrationToken
@@ -220,20 +225,17 @@ func (cfg *Config) registrationBody() map[string]any {
 // free capacity.
 func (cfg *Config) heartbeatBody() map[string]any {
 	body := map[string]any{}
-	aggregate := xsysinfo.GetGPUAggregateInfo()
+	aggregate := getGPUAggregateInfo()
 	if aggregate.TotalVRAM > 0 {
 		body["available_vram"] = aggregate.FreeVRAM
 	}
 
-	// CPU-only workers (or workers that lost GPU visibility momentarily):
-	// report system RAM so the scheduler still has capacity info.
-	if aggregate.TotalVRAM == 0 {
-		ramInfo, err := xsysinfo.GetSystemRAMInfo()
-		if err != nil {
-			xlog.Debug("Failed to detect worker RAM for heartbeat", "error", err)
-		} else {
-			body["available_ram"] = ramInfo.Available
-		}
+	// RAM availability changes independently from VRAM on discrete-GPU nodes.
+	ramInfo, err := getSystemRAMInfo()
+	if err != nil {
+		xlog.Debug("Failed to detect worker RAM for heartbeat", "error", err)
+	} else {
+		body["available_ram"] = ramInfo.Available
 	}
 
 	// Free disk changes far faster than VRAM under staging traffic (each

core/services/worker/registration_test.go

@@ -8,6 +8,17 @@ import (
 )
 
 var _ = Describe("Worker registration body", func() {
+	BeforeEach(func() {
+		originalTotalAvailableVRAM := totalAvailableVRAM
+		originalGetGPUAggregateInfo := getGPUAggregateInfo
+		originalGetSystemRAMInfo := getSystemRAMInfo
+		DeferCleanup(func() {
+			totalAvailableVRAM = originalTotalAvailableVRAM
+			getGPUAggregateInfo = originalGetGPUAggregateInfo
+			getSystemRAMInfo = originalGetSystemRAMInfo
+		})
+	})
+
 	It("includes the VRAM budget in the registration body when set", func() {
 		cfg := &Config{VRAMBudget: "80%"}
 		body := cfg.registrationBody()
@@ -30,4 +41,33 @@ var _ = Describe("Worker registration body", func() {
 		body := cfg.registrationBody()
 		Expect(body["total_vram"]).To(Equal(rawTotal))
 	})
+
+	It("reports RAM when GPU memory is visible", func() {
+		totalAvailableVRAM = func() (uint64, error) {
+			return 24_000, nil
+		}
+		getSystemRAMInfo = func() (*xsysinfo.SystemRAMInfo, error) {
+			return &xsysinfo.SystemRAMInfo{Total: 64_000, Available: 48_000}, nil
+		}
+
+		body := (&Config{}).registrationBody()
+
+		Expect(body["total_vram"]).To(Equal(uint64(24_000)))
+		Expect(body["total_ram"]).To(Equal(uint64(64_000)))
+		Expect(body["available_ram"]).To(Equal(uint64(48_000)))
+	})
+
+	It("reports current RAM in heartbeats when GPU memory is visible", func() {
+		getGPUAggregateInfo = func() xsysinfo.GPUAggregateInfo {
+			return xsysinfo.GPUAggregateInfo{TotalVRAM: 24_000, FreeVRAM: 12_000}
+		}
+		getSystemRAMInfo = func() (*xsysinfo.SystemRAMInfo, error) {
+			return &xsysinfo.SystemRAMInfo{Total: 64_000, Available: 40_000}, nil
+		}
+
+		body := (&Config{}).heartbeatBody()
+
+		Expect(body["available_vram"]).To(Equal(uint64(12_000)))
+		Expect(body["available_ram"]).To(Equal(uint64(40_000)))
+	})
 })

docs/content/features/distributed-mode.md

@@ -372,7 +372,10 @@ usage is reported back to the frontend:
 share one physical RAM between CPU and GPU. LocalAI detects them via
 `/sys/devices/soc0/family` and `/sys/devices/soc0/soc_id` (no `nvidia-smi`
 required) and reports system-RAM figures as VRAM. Free VRAM therefore tracks
-`MemAvailable` in `/proc/meminfo`.
+`MemAvailable` in `/proc/meminfo`. Workers report RAM metrics independently
+from VRAM on every registration and heartbeat. On unified-memory nodes, the
+available RAM and available VRAM values should therefore track each other
+closely; on discrete-GPU nodes they can change independently.
 
 ### Node Labels