feat(metrics): single chart engine + full-width stacked layout

- One engine: go-analyze/charts (grafana theme) for all live metrics
- Server chart: CPU temp, CPU load%, mem load%, power W, fan RPMs
- GPU charts: temp, load%, mem%, power W — one card per GPU, added dynamically
- Charts 1400x280px SVG, rendered at width:100% in single-column layout
- Add CPU load (from /proc/stat) and mem load (from /proc/meminfo) to LiveMetricSample
- Add GPU mem utilization to GPUMetricRow (nvidia-smi utilization.memory)
- Document charting architecture in bible-local/architecture/charting.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

audit/internal/platform/gpu_metrics.go

@@ -13,18 +13,19 @@ import (
 
 // GPUMetricRow is one telemetry sample from nvidia-smi during a stress test.
 type GPUMetricRow struct {
-	ElapsedSec float64
-	GPUIndex   int
-	TempC      float64
-	UsagePct   float64
-	PowerW     float64
-	ClockMHz   float64
+	ElapsedSec  float64 `json:"elapsed_sec"`
+	GPUIndex    int     `json:"index"`
+	TempC       float64 `json:"temp_c"`
+	UsagePct    float64 `json:"usage_pct"`
+	MemUsagePct float64 `json:"mem_usage_pct"`
+	PowerW      float64 `json:"power_w"`
+	ClockMHz    float64 `json:"clock_mhz"`
 }
 
 // sampleGPUMetrics runs nvidia-smi once and returns current metrics for each GPU.
 func sampleGPUMetrics(gpuIndices []int) ([]GPUMetricRow, error) {
 	args := []string{
-		"--query-gpu=index,temperature.gpu,utilization.gpu,power.draw,clocks.current.graphics",
+		"--query-gpu=index,temperature.gpu,utilization.gpu,utilization.memory,power.draw,clocks.current.graphics",
 		"--format=csv,noheader,nounits",
 	}
 	if len(gpuIndices) > 0 {
@@ -45,16 +46,17 @@ func sampleGPUMetrics(gpuIndices []int) ([]GPUMetricRow, error) {
 			continue
 		}
 		parts := strings.Split(line, ", ")
-		if len(parts) < 5 {
+		if len(parts) < 6 {
 			continue
 		}
 		idx, _ := strconv.Atoi(strings.TrimSpace(parts[0]))
 		rows = append(rows, GPUMetricRow{
-			GPUIndex: idx,
-			TempC:    parseGPUFloat(parts[1]),
-			UsagePct: parseGPUFloat(parts[2]),
-			PowerW:   parseGPUFloat(parts[3]),
-			ClockMHz: parseGPUFloat(parts[4]),
+			GPUIndex:    idx,
+			TempC:       parseGPUFloat(parts[1]),
+			UsagePct:    parseGPUFloat(parts[2]),
+			MemUsagePct: parseGPUFloat(parts[3]),
+			PowerW:      parseGPUFloat(parts[4]),
+			ClockMHz:    parseGPUFloat(parts[5]),
 		})
 	}
 	return rows, nil

audit/internal/platform/live_metrics.go

@@ -1,15 +1,23 @@
 package platform
 
-import "time"
+import (
+	"bufio"
+	"os"
+	"strconv"
+	"strings"
+	"time"
+)
 
 // LiveMetricSample is a single point-in-time snapshot of server metrics
 // collected for the web UI metrics page.
 type LiveMetricSample struct {
-	Timestamp time.Time       `json:"ts"`
-	Fans      []FanReading    `json:"fans"`
-	Temps     []TempReading   `json:"temps"`
-	PowerW    float64         `json:"power_w"`
-	GPUs      []GPUMetricRow  `json:"gpus"`
+	Timestamp  time.Time      `json:"ts"`
+	Fans       []FanReading   `json:"fans"`
+	Temps      []TempReading  `json:"temps"`
+	PowerW     float64        `json:"power_w"`
+	CPULoadPct float64        `json:"cpu_load_pct"`
+	MemLoadPct float64        `json:"mem_load_pct"`
+	GPUs       []GPUMetricRow `json:"gpus"`
 }
 
 // TempReading is a named temperature sensor value.
@@ -41,5 +49,91 @@ func SampleLiveMetrics() LiveMetricSample {
 	// System power — returns 0 if unavailable
 	s.PowerW = sampleSystemPower()
 
+	// CPU load — from /proc/stat
+	s.CPULoadPct = sampleCPULoadPct()
+
+	// Memory load — from /proc/meminfo
+	s.MemLoadPct = sampleMemLoadPct()
+
 	return s
 }
+
+// sampleCPULoadPct reads two /proc/stat snapshots 200ms apart and returns
+// the overall CPU utilisation percentage.
+var cpuStatPrev [2]uint64 // [total, idle]
+
+func sampleCPULoadPct() float64 {
+	total, idle := readCPUStat()
+	if total == 0 {
+		return 0
+	}
+	prevTotal, prevIdle := cpuStatPrev[0], cpuStatPrev[1]
+	cpuStatPrev = [2]uint64{total, idle}
+	if prevTotal == 0 {
+		return 0
+	}
+	dt := float64(total - prevTotal)
+	di := float64(idle - prevIdle)
+	if dt <= 0 {
+		return 0
+	}
+	pct := (1 - di/dt) * 100
+	if pct < 0 {
+		return 0
+	}
+	if pct > 100 {
+		return 100
+	}
+	return pct
+}
+
+func readCPUStat() (total, idle uint64) {
+	f, err := os.Open("/proc/stat")
+	if err != nil {
+		return 0, 0
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		line := sc.Text()
+		if !strings.HasPrefix(line, "cpu ") {
+			continue
+		}
+		fields := strings.Fields(line)[1:] // skip "cpu"
+		var vals [10]uint64
+		for i := 0; i < len(fields) && i < 10; i++ {
+			vals[i], _ = strconv.ParseUint(fields[i], 10, 64)
+		}
+		// idle = idle + iowait
+		idle = vals[3] + vals[4]
+		for _, v := range vals {
+			total += v
+		}
+		return total, idle
+	}
+	return 0, 0
+}
+
+func sampleMemLoadPct() float64 {
+	f, err := os.Open("/proc/meminfo")
+	if err != nil {
+		return 0
+	}
+	defer f.Close()
+	vals := map[string]uint64{}
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		fields := strings.Fields(sc.Text())
+		if len(fields) >= 2 {
+			v, _ := strconv.ParseUint(fields[1], 10, 64)
+			vals[strings.TrimSuffix(fields[0], ":")] = v
+		}
+	}
+	total := vals["MemTotal"]
+	avail := vals["MemAvailable"]
+	if total == 0 {
+		return 0
+	}
+	used := total - avail
+	return float64(used) / float64(total) * 100
+}