feat: add dynamic config reload, unique device identity (UUID), and MQTT support

- New StreamManager for dynamic RTSP/FLV stream lifecycle
- Dynamic worker scaling for inference
- Device UUID generation and persistence
- Telegraf config and NPU monitoring scripts
- .gitignore for build artifacts

.gitignore

@@ -0,0 +1,7 @@
+build/
+logs/
+*.om
+*.onnx
+test.jpg
+fusion_result.json
+device.uuid

cmd/edge-agent/main.go

@@ -21,13 +21,14 @@ func main() {
 	flag.Parse()
 
 	cfg := config.Load(*cfgPath)
-	log.Printf("edge-agent start id=%s cloud=%s", cfg.EdgeID, cfg.CloudURL)
+	log.Printf("edge-agent start id=%s uuid=%s cloud=%s mqtt=%s",
+		cfg.EdgeID, cfg.GetDeviceIdentity(), cfg.CloudURL, cfg.MqttBroker)
 
 	ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
 	defer cancel()
 
-	frames := make(chan stream.Frame, 50)
+	frames := make(chan stream.Frame, 100)
-	events := make(chan event.SuspectedEvent, 200)
+	events := make(chan event.SuspectedEvent, 500)
 
 	var wg sync.WaitGroup
 	run := func(fn func(context.Context)) {
@@ -38,11 +39,23 @@ func main() {
 		}()
 	}
 
-	run(stream.NewIngestor(cfg, frames).Run)
+	// 1. Cloud Communication
-	run(infer.NewClient(cfg, frames, events).Run)
+	mqttMgr := control.NewMqttManager(cfg)
+	run(mqttMgr.Run)
+
+	// 2. Stream Ingestion (Dynamic)
+	run(stream.NewStreamManager(cfg, frames, mqttMgr.GetUpdates()).Run)
+
+	// 3. Inference (Dynamic Workers)
+	run(infer.NewClient(cfg, frames, events, mqttMgr.GetUpdates()).Run)
+
+	// 4. Event Upload
 	run(event.NewUploader(cfg, events).Run)
+
+	// 5. Heartbeat
 	run(control.NewHeartbeat(cfg).Run)
-	run(control.NewConfigAgent(cfg).Run)
+
+	// 6. OTA Update
 	run(control.NewOTAAgent(cfg).Run)
 
 	wg.Wait()

config/edge.yaml

@@ -1,13 +1,21 @@
 edge_id: edge-demo-001
 cloud_url: http://101.36.73.102:8004
+mqtt_broker: "tcp://101.36.73.102:1883"
+mqtt_user: ""
+mqtt_pass: ""
 edge_token: ""
 
-# RTSP 流地址（留空使用演示模式）
+# 视频流配置
 rtsp_urls: []
 
+# 推理配置
 infer_socket: /tmp/edge-infer.sock
 infer_fps: 5
 infer_workers: 3
 conf_threshold: 0.5
 dedup_window_sec: 30
+
+# OTA 配置 (可选: Hawkbit URL 或 Nginx URL)
+ota_url: http://101.36.73.102:8087
+
 version: 1.0.0

config/telegraf.conf

@@ -0,0 +1,32 @@
+[agent]
+  interval = "10s"
+  round_interval = true
+  metric_batch_size = 1000
+  metric_buffer_limit = 10000
+  flush_interval = "10s"
+  hostname = "edge-001" # 启动时替换为实际 edge_id
+
+# 输出到云端 InfluxDB
+[[outputs.influxdb_v2]]
+  urls = ["http://101.36.73.102:18086"]
+  token = "my-super-secret-token"  # 替换为实际 token
+  organization = "tianyan"
+  bucket = "edge_metrics"
+
+# 基础硬件监控
+[[inputs.cpu]]
+  percpu = false
+  totalcpu = true
+
+[[inputs.mem]]
+
+[[inputs.disk]]
+  ignore_fs = ["tmpfs", "devtmpfs", "devfs", "overlay", "aufs", "squashfs"]
+
+[[inputs.net]]
+
+# NPU 自定义监控
+[[inputs.exec]]
+  commands = ["/opt/tianyan-edge/scripts/npu_info.sh"]
+  timeout = "5s"
+  data_format = "influx"

go.mod

@@ -1,7 +1,15 @@
 module tianyan-edge
 
-go 1.22
+go 1.18
 
-require gopkg.in/yaml.v3 v3.0.1
+require (
+	github.com/eclipse/paho.mqtt.golang v1.4.3
+	gopkg.in/yaml.v3 v3.0.1
+)
 
-require gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c // indirect
+require (
+	github.com/gorilla/websocket v1.5.0 // indirect
+	golang.org/x/net v0.8.0 // indirect
+	golang.org/x/sync v0.1.0 // indirect
+	gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c // indirect
+)

go.sum

@@ -1,8 +1,16 @@
+github.com/eclipse/paho.mqtt.golang v1.4.3 h1:2kwcUGn8seMUfWndX0hGbvH8r7crgcJguQNCyp70xik=
+github.com/eclipse/paho.mqtt.golang v1.4.3/go.mod h1:CSYvoAlsMkhYOXh/oKyxa8EcBci6dVkLCbo5tTC1RIE=
+github.com/gorilla/websocket v1.5.0 h1:PPwGk2jz7EePpoHN/+ClbZu8SPxiqlu12wZP/3sWmnc=
+github.com/gorilla/websocket v1.5.0/go.mod h1:YR8l580nyteQvAITg2hZ9XVh4b55+EU/adAjf1fMHhE=
 github.com/kr/pretty v0.2.1 h1:Fmg33tUaq4/8ym9TJN1x7sLJnHVwhP33CNkpYV/7rwI=
 github.com/kr/pretty v0.2.1/go.mod h1:ipq/a2n7PKx3OHsz4KJII5eveXtPO4qwEXGdVfWzfnI=
 github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
 github.com/kr/text v0.1.0 h1:45sCR5RtlFHMR4UwH9sdQ5TC8v0qDQCHnXt+kaKSTVE=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
+golang.org/x/net v0.8.0 h1:Zrh2ngAOFYneWTAIAPethzeaQLuHwhuBkuV6ZiRnUaQ=
+golang.org/x/net v0.8.0/go.mod h1:QVkue5JL9kW//ek3r6jTKnTFis1tRmNAW2P1shuFdJc=
+golang.org/x/sync v0.1.0 h1:wsuoTGHzEhffawBOhz5CYhcrV4IdKZbEyZjBMuTp12o=
+golang.org/x/sync v0.1.0/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c/go.mod h1:JHkPIbrfpd72SG/EVd6muEfDQjcINNoR0C8j2r3qZ4Q=

internal/config/config.go

@@ -1,14 +1,21 @@
 package config
 
 import (
+	"crypto/rand"
+	"encoding/hex"
 	"os"
+	"path/filepath"
 
 	"gopkg.in/yaml.v3"
 )
 
 type Config struct {
+	DeviceUUID      string   `yaml:"device_uuid"`
 	EdgeID          string   `yaml:"edge_id"`
 	CloudURL        string   `yaml:"cloud_url"`
+	MqttBroker      string   `yaml:"mqtt_broker"`
+	MqttUser        string   `yaml:"mqtt_user"`
+	MqttPass        string   `yaml:"mqtt_pass"`
 	EdgeToken       string   `yaml:"edge_token"`
 	RTSPURLs        []string `yaml:"rtsp_urls"`
 	InferSocket     string   `yaml:"infer_socket"`
@@ -16,12 +23,14 @@ type Config struct {
 	InferWorkers    int      `yaml:"infer_workers"`
 	ConfThreshold   float64  `yaml:"conf_threshold"`
 	DedupWindowSec  int      `yaml:"dedup_window_sec"`
+	OTAUrl          string   `yaml:"ota_url"`
 	Version         string   `yaml:"version"`
+	configPath      string   `json:"-"`
 }
 
 func Load(path string) *Config {
 	cfg := &Config{
-		EdgeID:         "edge-demo-001",
+		EdgeID:         "edge-unknown",
 		CloudURL:       "http://localhost:8004",
 		InferSocket:    "/tmp/edge-infer.sock",
 		InferFPS:       5,
@@ -29,12 +38,15 @@ func Load(path string) *Config {
 		ConfThreshold:  0.5,
 		DedupWindowSec: 30,
 		Version:        "1.0.0",
+		configPath:     path,
 	}
-	data, err := os.ReadFile(path)
+
-	if err != nil {
+	cfg.loadUUID()
-		return cfg
+
-	}
+	if data, err := os.ReadFile(path); err == nil {
 		_ = yaml.Unmarshal(data, cfg)
+	}
+
 	if cfg.InferWorkers <= 0 {
 		cfg.InferWorkers = 1
 	}
@@ -43,3 +55,31 @@ func Load(path string) *Config {
 	}
 	return cfg
 }
+
+func (c *Config) loadUUID() {
+	uuidPath := "/opt/tianyan-edge/device.uuid"
+	if data, err := os.ReadFile(uuidPath); err == nil {
+		c.DeviceUUID = string(data)
+	} else {
+		b := make([]byte, 16)
+		rand.Read(b)
+		c.DeviceUUID = hex.EncodeToString(b)
+		os.MkdirAll(filepath.Dir(uuidPath), 0755)
+		os.WriteFile(uuidPath, []byte(c.DeviceUUID), 0644)
+	}
+}
+
+func (c *Config) Save() error {
+	data, err := yaml.Marshal(c)
+	if err != nil {
+		return err
+	}
+	return os.WriteFile(c.configPath, data, 0644)
+}
+
+func (c *Config) GetDeviceIdentity() string {
+	if c.DeviceUUID != "" {
+		return c.DeviceUUID
+	}
+	return c.EdgeID
+}

internal/control/mqtt_manager.go

@@ -0,0 +1,123 @@
+package control
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log"
+	"time"
+
+	mqtt "github.com/eclipse/paho.mqtt.golang"
+	"tianyan-edge/internal/config"
+)
+
+type MqttManager struct {
+	client   mqtt.Client
+	cfg      *config.Config
+	updates  chan *config.Config // Channel to broadcast config updates
+}
+
+func NewMqttManager(cfg *config.Config) *MqttManager {
+	return &MqttManager{
+		cfg:     cfg,
+		updates: make(chan *config.Config, 10),
+	}
+}
+
+func (m *MqttManager) Run(ctx context.Context) {
+	if m.cfg.MqttBroker == "" {
+		log.Println("mqtt: no broker configured, skipping")
+		return
+	}
+
+	opts := mqtt.NewClientOptions()
+	opts.AddBroker(m.cfg.MqttBroker)
+	opts.SetClientID(m.cfg.GetDeviceIdentity())
+	opts.SetUsername(m.cfg.MqttUser)
+	opts.SetPassword(m.cfg.MqttPass)
+	opts.SetAutoReconnect(true)
+	opts.SetMaxReconnectInterval(1 * time.Minute)
+
+	// LWT
+	willTopic := fmt.Sprintf("tianyan/edge/%s/status", m.cfg.GetDeviceIdentity())
+	opts.SetWill(willTopic, "offline", 1, true)
+
+	opts.SetOnConnectHandler(func(c mqtt.Client) {
+		log.Println("mqtt: connected to broker")
+		c.Publish(willTopic, 1, true, "online")
+		configTopic := fmt.Sprintf("tianyan/edge/%s/config", m.cfg.GetDeviceIdentity())
+		token := c.Subscribe(configTopic, 1, m.handleConfigUpdate)
+		token.Wait()
+		if token.Error() != nil {
+			log.Printf("mqtt: subscribe failed: %v", token.Error())
+		} else {
+			log.Printf("mqtt: subscribed to %s", configTopic)
+		}
+	})
+
+	m.client = mqtt.NewClient(opts)
+	if token := m.client.Connect(); token.Wait() && token.Error() != nil {
+		log.Printf("mqtt: connect error: %v", token.Error())
+	}
+
+	<-ctx.Done()
+	m.client.Disconnect(1000)
+	log.Println("mqtt: disconnected")
+}
+
+func (m *MqttManager) handleConfigUpdate(c mqtt.Client, msg mqtt.Message) {
+	log.Printf("mqtt: received config update on %s", msg.Topic())
+	var newCfg map[string]interface{}
+	if err := json.Unmarshal(msg.Payload(), &newCfg); err != nil {
+		log.Printf("mqtt: config parse error: %v", err)
+		return
+	}
+
+	// Update local config struct
+	if v, ok := newCfg["infer_fps"].(float64); ok {
+		m.cfg.InferFPS = int(v)
+	}
+	if v, ok := newCfg["conf_threshold"].(float64); ok {
+		m.cfg.ConfThreshold = v
+	}
+	if v, ok := newCfg["infer_workers"].(float64); ok {
+		m.cfg.InferWorkers = int(v)
+	}
+	if v, ok := newCfg["rtsp_urls"]; ok {
+		if urls, ok := v.([]interface{}); ok {
+			var s []string
+			for _, u := range urls {
+				if str, ok := u.(string); ok {
+					s = append(s, str)
+				}
+			}
+			m.cfg.RTSPURLs = s
+		}
+	}
+
+	// Save to disk
+	if err := m.cfg.Save(); err != nil {
+		log.Printf("mqtt: failed to save config: %v", err)
+	} else {
+		log.Println("mqtt: config saved to disk")
+	}
+
+	// Broadcast update to other components
+	select {
+	case m.updates <- m.cfg:
+	default:
+		log.Println("mqtt: update channel full, dropping update")
+	}
+}
+
+func (m *MqttManager) GetUpdates() <-chan *config.Config {
+	return m.updates
+}
+
+func (m *MqttManager) Publish(topic string, payload interface{}) {
+	if m.client == nil || !m.client.IsConnected() {
+		return
+	}
+	data, _ := json.Marshal(payload)
+	m.client.Publish(topic, 1, false, data)
+}

internal/infer/client.go

@@ -50,27 +50,71 @@ type Client struct {
 	frames    <-chan stream.Frame
 	events    chan<- event.SuspectedEvent
 	dedup     *dedupState
+	updates   <-chan *config.Config
 }
 
-func NewClient(cfg *config.Config, frames <-chan stream.Frame, events chan<- event.SuspectedEvent) *Client {
+func NewClient(cfg *config.Config, frames <-chan stream.Frame, events chan<- event.SuspectedEvent, updates <-chan *config.Config) *Client {
 	return &Client{
 		cfg:     cfg,
 		frames:  frames,
 		events:  events,
 		dedup:   &dedupState{seen: make(map[string]float64)},
+		updates: updates,
 	}
 }
 
 func (c *Client) Run(ctx context.Context) {
 	var wg sync.WaitGroup
-	for i := 0; i < c.cfg.InferWorkers; i++ {
+
+	// Worker management loop
+	numWorkers := c.cfg.InferWorkers
+	activeWorkers := make([]context.CancelFunc, numWorkers)
+
+	for i := 0; i < numWorkers; i++ {
+		workerCtx, cancel := context.WithCancel(ctx)
+		activeWorkers[i] = cancel
 		wg.Add(1)
-		go func(workerID int) {
+		go func(id int, wCtx context.Context) {
 			defer wg.Done()
-			c.workerLoop(ctx, workerID)
+			c.workerLoop(wCtx, id)
-		}(i)
+		}(i, workerCtx)
+	}
+
+	for {
+		select {
+		case <-ctx.Done():
+			for _, cancel := range activeWorkers {
+				cancel()
 			}
 			wg.Wait()
+			return
+		case newCfg := <-c.updates:
+			c.cfg = newCfg
+			target := c.cfg.InferWorkers
+			if target < 0 {
+				target = 0
+			}
+
+			// Scale up
+			for i := len(activeWorkers); i < target; i++ {
+				workerCtx, cancel := context.WithCancel(ctx)
+				activeWorkers = append(activeWorkers, cancel)
+				wg.Add(1)
+				go func(id int, wCtx context.Context) {
+					defer wg.Done()
+					c.workerLoop(wCtx, id)
+				}(i, workerCtx)
+			}
+			// Scale down
+			if target < len(activeWorkers) {
+				for i := target; i < len(activeWorkers); i++ {
+					activeWorkers[i]()
+				}
+				activeWorkers = activeWorkers[:target]
+			}
+			log.Printf("infer: workers scaled to %d", target)
+		}
+	}
 }
 
 func (c *Client) workerLoop(ctx context.Context, workerID int) {
@@ -89,10 +133,9 @@ func (c *Client) workerLoop(ctx context.Context, workerID int) {
 		log.Printf("infer[w%d]: socket connected", workerID)
 		if !c.loop(ctx, conn, workerID) {
 			conn.Close()
-			return
+			time.Sleep(1 * time.Second)
 		}
 		conn.Close()
-		time.Sleep(500 * time.Millisecond)
 	}
 }
 
@@ -103,12 +146,10 @@ func (c *Client) loop(ctx context.Context, conn net.Conn, workerID int) bool {
 			return false
 		case f := <-c.frames:
 			if err := c.send(conn, f); err != nil {
-				log.Printf("infer[w%d]: send error: %v", workerID, err)
 				return true
 			}
 			res, err := c.recv(conn)
 			if err != nil {
-				log.Printf("infer[w%d]: recv error: %v", workerID, err)
 				return true
 			}
 			c.emitEvents(f, res)

internal/stream/ingestor.go

@@ -1,7 +1,6 @@
 package stream
 
 import (
-	"bytes"
 	"context"
 	"fmt"
 	"io"
@@ -10,6 +9,8 @@ import (
 	"sync"
 	"time"
 
+	"bytes"
+
 	"tianyan-edge/internal/config"
 )
 
@@ -21,63 +22,109 @@ type Frame struct {
 	TS       float64
 }
 
-type Ingestor struct {
+type StreamManager struct {
 	cfg       *config.Config
 	frames    chan<- Frame
+	processes map[string]*exec.Cmd
+	mu        sync.Mutex
+	updates   <-chan *config.Config
 }
 
-func NewIngestor(cfg *config.Config, frames chan<- Frame) *Ingestor {
+func NewStreamManager(cfg *config.Config, frames chan<- Frame, updates <-chan *config.Config) *StreamManager {
-	return &Ingestor{cfg: cfg, frames: frames}
+	return &StreamManager{
+		cfg:       cfg,
+		frames:    frames,
+		processes: make(map[string]*exec.Cmd),
+		updates:   updates,
+	}
 }
 
-func (ing *Ingestor) Run(ctx context.Context) {
+func (sm *StreamManager) Run(ctx context.Context) {
-	if len(ing.cfg.RTSPURLs) == 0 {
+	// Initial start
-		log.Println("stream: no RTSP URLs configured, idle")
+	sm.applyStreams(ctx, sm.cfg.RTSPURLs)
-		<-ctx.Done()
+
+	for {
+		select {
+		case <-ctx.Done():
+			sm.stopAll()
 			return
+		case newCfg := <-sm.updates:
+			log.Printf("stream: applying new config, urls=%d", len(newCfg.RTSPURLs))
+			sm.applyStreams(ctx, newCfg.RTSPURLs)
 		}
-	var wg sync.WaitGroup
-	for i, url := range ing.cfg.RTSPURLs {
-		wg.Add(1)
-		go func(idx int, u string) {
-			defer wg.Done()
-			ing.streamLoop(ctx, idx, u)
-		}(i, url)
 	}
-	wg.Wait()
 }
 
-func (ing *Ingestor) streamLoop(ctx context.Context, idx int, url string) {
+func (sm *StreamManager) applyStreams(ctx context.Context, urls []string) {
-	deviceID := fmt.Sprintf("cam-%03d", idx)
+	sm.mu.Lock()
-	fps := ing.cfg.InferFPS
+	defer sm.mu.Unlock()
-	if fps <= 0 {
+
-		fps = 5
+	// Identify URLs to remove (present in sm.processes but not in new urls)
+	toRemove := map[string]*exec.Cmd{}
+	for url, cmd := range sm.processes {
+		found := false
+		for _, u := range urls {
+			if u == url {
+				found = true
+				break
 			}
+		}
+		if !found {
+			toRemove[url] = cmd
+		}
+	}
+
+	// Stop removed streams
+	for url, cmd := range toRemove {
+		log.Printf("stream: stopping %s", url)
+		cmd.Process.Kill()
+		cmd.Wait()
+		delete(sm.processes, url)
+	}
+
+	// Start new streams
+	for i, url := range urls {
+		if _, exists := sm.processes[url]; !exists {
+			go sm.streamLoop(ctx, i, url)
+		}
+	}
+}
+
+func (sm *StreamManager) streamLoop(ctx context.Context, idx int, url string) {
+	deviceID := fmt.Sprintf("cam-%03d", idx)
 	for {
 		select {
 		case <-ctx.Done():
 			return
 		default:
 		}
+
 		cmd := exec.CommandContext(ctx, "ffmpeg",
-			"-rtsp_transport", "tcp", "-i", url,
+			"-i", url,
-			"-vf", fmt.Sprintf("fps=%d", fps),
+			"-vf", fmt.Sprintf("fps=%d", sm.cfg.InferFPS),
 			"-f", "image2pipe", "-vcodec", "mjpeg", "-q:v", "5", "-",
 		)
 		stdout, err := cmd.StdoutPipe()
 		if err != nil {
-			log.Printf("stream[%d] stdout pipe failed: %v, retry 10s", idx, err)
+			log.Printf("stream[%d] stdout pipe failed: %v", idx, err)
-			time.Sleep(10 * time.Second)
+			time.Sleep(5 * time.Second)
 			continue
 		}
+
+		sm.mu.Lock()
+		sm.processes[url] = cmd
+		sm.mu.Unlock()
+
 		if err := cmd.Start(); err != nil {
-			log.Printf("stream[%d] ffmpeg start failed: %v, retry 10s", idx, err)
+			log.Printf("stream[%d] start failed: %v", idx, err)
-			time.Sleep(10 * time.Second)
+			time.Sleep(5 * time.Second)
 			continue
 		}
+
 		log.Printf("stream[%d] connected  url=%s", idx, url)
-		ing.readFrames(ctx, stdout, idx, deviceID, url)
+		sm.readFrames(ctx, stdout, idx, deviceID, url)
 		cmd.Wait()
+
 		select {
 		case <-ctx.Done():
 			return
@@ -88,7 +135,7 @@ func (ing *Ingestor) streamLoop(ctx context.Context, idx int, url string) {
 	}
 }
 
-func (ing *Ingestor) readFrames(ctx context.Context, r io.Reader, idx int, deviceID, url string) {
+func (sm *StreamManager) readFrames(ctx context.Context, r io.Reader, idx int, deviceID, url string) {
 	soi, eoi := []byte{0xFF, 0xD8}, []byte{0xFF, 0xD9}
 	buf, tmp := make([]byte, 0, 1<<20), make([]byte, 32768)
 	for {
@@ -116,7 +163,7 @@ func (ing *Ingestor) readFrames(ctx context.Context, r io.Reader, idx int, devic
 				copy(frame, buf[:end])
 				buf = buf[end:]
 				select {
-				case ing.frames <- Frame{StreamID: idx, DeviceID: deviceID, URL: url,
+				case sm.frames <- Frame{StreamID: idx, DeviceID: deviceID, URL: url,
 					JPEG: frame, TS: float64(time.Now().UnixMilli()) / 1000.0}:
 				default:
 				}
@@ -127,3 +174,13 @@ func (ing *Ingestor) readFrames(ctx context.Context, r io.Reader, idx int, devic
 		}
 	}
 }
+
+func (sm *StreamManager) stopAll() {
+	sm.mu.Lock()
+	defer sm.mu.Unlock()
+	for url, cmd := range sm.processes {
+		log.Printf("stream: stopping all %s", url)
+		cmd.Process.Kill()
+		delete(sm.processes, url)
+	}
+}

scripts/npu_info.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+# 采集 NPU 状态并输出为 Influx Line Protocol
+
+COMMON_OUTPUT=$(/usr/local/sbin/npu-smi info -t common -i 0 2>/dev/null)
+MEM_OUTPUT=$(/usr/local/sbin/npu-smi info -t memory -i 0 2>/dev/null)
+
+TEMP=$(echo "$COMMON_OUTPUT" | grep "Temperature(C)" | awk '{print $NF}')
+UTIL=$(echo "$COMMON_OUTPUT" | grep "Aicore Usage Rate" | awk '{print $NF}')
+MEM_TOTAL_MB=$(echo "$MEM_OUTPUT" | grep "Capacity(MB)" | awk '{print $NF}')
+MEM_RATE=$(echo "$COMMON_OUTPUT" | grep "Memory Usage Rate" | awk '{print $NF}')
+
+# 计算实际使用量 (使用 awk 代替 bc)
+if [ -n "$MEM_TOTAL_MB" ] && [ -n "$MEM_RATE" ]; then
+    MEM_USED_MB=$(awk "BEGIN {printf \"%d\", $MEM_TOTAL_MB * $MEM_RATE / 100}")
+else
+    MEM_TOTAL_MB=0
+    MEM_USED_MB=0
+    UTIL=0
+    TEMP=0
+fi
+
+# 默认值保护
+TEMP=${TEMP:-0}
+UTIL=${UTIL:-0}
+
+echo "npu_status,device=0 temp=${TEMP},utilization=${UTIL},memory_used=${MEM_USED_MB:-0},memory_total=${MEM_TOTAL_MB:-0}"