feat: Maestro driver integration + DeviceDriver architecture (#32)

* proto(driver): drop launcher_activity from LaunchRequest

* refactor(driver): rename Driver to DeviceDriver, drop launcherActivity from Launch

* refactor(driver): rename package maestro to sidecar

* feat(driver): add ChromeDriver backed by chromedp

* feat(hierarchy): replace XML parser with TreeNode JSON parser

* refactor(driver): update mock and runner to DeviceDriver, drop launcherActivity

* feat(runner): add platform routing for web vs sidecar

* test(verifier): update hierarchy fixtures from XML to TreeNode JSON

* feat(sidecar): extract readLogcat/readProcMetrics; add MaestroDriverBackend

- Extract readLogcat() and readProcMetrics() as internal package-level
  helpers parameterised by serial
- Add MaestroDriverBackend wiring maestro-client AndroidDriver
- Drop launcherActivity from DriverBackend interface and StubDriverBackend
- Add maestro-utils and micrometer-core as explicit compile deps

* refactor(sidecar): inject DriverBackend into DriverService; drop launcherActivity

- Remove serial and default backend from DriverService constructor
- Make backend a required parameter
- Drop launcherActivity from launch RPC handler
- Create MaestroDriverBackend in Main.kt when platform is android

* test(sidecar): update tests for dropped launcherActivity and required backend

* chore(sidecar): remove unnecessary micrometer-core direct dependency

cmd/sanderling/main.go

@@ -17,7 +17,6 @@ var Version = "dev"
 type testOptions struct {
 	spec     string
 	bundleID string
-	launcherActivity string
 	platform string
 	avd      string
 	duration time.Duration
@@ -45,8 +44,7 @@ func parseTestArgs(args []string, stderr io.Writer) (testOptions, error) {
 	var options testOptions
 	flagSet.StringVar(&options.spec, "spec", "", "path to the TypeScript spec (required)")
 	flagSet.StringVar(&options.bundleID, "bundle-id", "", "target app bundle ID (required)")
-	flagSet.StringVar(&options.launcherActivity, "launcher-activity", "", "optional <pkg>/<activity> to launch (overrides default resolution)")
-	flagSet.StringVar(&options.platform, "platform", "android", "target platform: android (ios deferred)")
+	flagSet.StringVar(&options.platform, "platform", "android", "target platform: android, web")
 	flagSet.StringVar(&options.avd, "avd", "", "Android AVD name to boot if no device is connected")
 	flagSet.DurationVar(&options.duration, "duration", 5*time.Minute, "total test duration")
 	flagSet.Int64Var(&options.seed, "seed", 0, "RNG seed (0 = random)")
@@ -60,8 +58,10 @@ func parseTestArgs(args []string, stderr io.Writer) (testOptions, error) {
 	if options.bundleID == "" {
 		return testOptions{}, errors.New("--bundle-id is required")
 	}
-	if options.platform != "android" {
-		return testOptions{}, fmt.Errorf("unsupported platform: %q (only android in v0.1)", options.platform)
+	switch options.platform {
+	case "android", "ios", "web":
+	default:
+		return testOptions{}, fmt.Errorf("unsupported platform: %q (android, web)", options.platform)
 	}
 	return options, nil
 }

cmd/sanderling/main_test.go

@@ -76,11 +76,11 @@ func TestParseTestArgs_AVDIsOptional(t *testing.T) {
 	}
 }
 
-func TestParseTestArgs_RejectsNonAndroidPlatform(t *testing.T) {
+func TestParseTestArgs_RejectsUnknownPlatform(t *testing.T) {
 	_, err := parseTestArgs([]string{
 		"--spec", "s.ts",
 		"--bundle-id", "com.example",
-		"--platform", "ios",
+		"--platform", "fuchsia",
 		"--avd", "x",
 	}, io.Discard)
 	if err == nil || !strings.Contains(err.Error(), "unsupported platform") {
@@ -88,6 +88,20 @@ func TestParseTestArgs_RejectsNonAndroidPlatform(t *testing.T) {
 	}
 }
 
+func TestParseTestArgs_AcceptsWebPlatform(t *testing.T) {
+	options, err := parseTestArgs([]string{
+		"--spec", "s.ts",
+		"--bundle-id", "http://localhost:3000",
+		"--platform", "web",
+	}, io.Discard)
+	if err != nil {
+		t.Fatalf("unexpected error for web platform: %v", err)
+	}
+	if options.platform != "web" {
+		t.Errorf("expected platform=web, got %q", options.platform)
+	}
+}
+
 func TestRun_HelpPrintsUsage(t *testing.T) {
 	var stdout bytes.Buffer
 	if err := run([]string{"sanderling"}, &stdout, io.Discard); err != nil {

cmd/sanderling/test_run.go

@@ -14,7 +14,9 @@ import (
 
 	"github.com/priyanshujain/sanderling/internal/agent"
 	"github.com/priyanshujain/sanderling/internal/bundler"
-	"github.com/priyanshujain/sanderling/internal/driver/maestro"
+	"github.com/priyanshujain/sanderling/internal/driver"
+	"github.com/priyanshujain/sanderling/internal/driver/chrome"
+	driverSidecar "github.com/priyanshujain/sanderling/internal/driver/sidecar"
 	"github.com/priyanshujain/sanderling/internal/runner"
 	"github.com/priyanshujain/sanderling/internal/sidecar"
 	"github.com/priyanshujain/sanderling/internal/trace"
@@ -28,7 +30,7 @@ const (
 )
 
 func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer) error {
-	if options.platform == "android" {
+	if options.platform == "android" || options.platform == "ios" {
 		if err := ensureDevice(ctx, options.avd, stdout); err != nil {
 			return err
 		}
@@ -36,8 +38,6 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 	aliases := map[string]string{}
 	if specApiPath := resolveSpecAPIPath(options.spec); specApiPath != "" {
 		aliases["@sanderling/spec"] = specApiPath
-		// Also alias published subpath exports so specs importing from
-		// "@sanderling/spec/defaults/properties" resolve to the in-tree source.
 		base := filepath.Dir(specApiPath)
 		aliases["@sanderling/spec/defaults/properties"] = filepath.Join(base, "defaults/properties.ts")
 	}
@@ -54,46 +54,11 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 	}
 	fmt.Fprintf(stdout, "bundled spec: %d bytes (sha256=%s)\n", len(bundle.JavaScript), bundle.SHA256[:12])
 
-	sidecarDirectory := filepath.Join(os.TempDir(), "sanderling-sidecar")
-	jarPath, err := sidecar.Extract(sidecarDirectory)
-	if err != nil {
-		return fmt.Errorf("extract sidecar: %w", err)
-	}
-	fmt.Fprintf(stdout, "sidecar JAR: %s (size=%d)\n", jarPath, sidecar.EmbeddedSize())
-
-	sidecarPort, err := pickFreePort()
+	activeDriver, cleanup, err := buildDriver(ctx, options, stdout)
 	if err != nil {
 		return err
 	}
-	sidecarCommand := exec.CommandContext(ctx, "java", "-jar", jarPath,
-		"--port", strconv.Itoa(sidecarPort),
-		"--platform", options.platform,
-	)
-	sidecarCommand.Stdout = stdout
-	sidecarCommand.Stderr = stdout
-	sidecarCommand.Env = envWithAndroidPlatformTools(os.Environ())
-	if err := sidecarCommand.Start(); err != nil {
-		return fmt.Errorf("spawn sidecar: %w", err)
-	}
-	defer func() {
-		if sidecarCommand.Process != nil {
-			_ = sidecarCommand.Process.Kill()
-		}
-	}()
-	fmt.Fprintf(stdout, "sidecar pid=%d listening on 127.0.0.1:%d\n", sidecarCommand.Process.Pid, sidecarPort)
-
-	driverClient, err := maestro.Dial(fmt.Sprintf("127.0.0.1:%d", sidecarPort))
-	if err != nil {
-		return fmt.Errorf("dial sidecar: %w", err)
-	}
-	defer driverClient.Close()
-	healthCtx, healthCancel := context.WithTimeout(ctx, sidecarStartupTimeout)
-	if err := driverClient.WaitForHealth(healthCtx, 250*time.Millisecond); err != nil {
-		healthCancel()
-		return fmt.Errorf("sidecar health check: %w", err)
-	}
-	healthCancel()
-	fmt.Fprintln(stdout, "sidecar is healthy")
+	defer cleanup()
 
 	listener, err := net.Listen("tcp", "127.0.0.1:0")
 	if err != nil {
@@ -102,6 +67,7 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 	defer listener.Close()
 	agentPort := listener.Addr().(*net.TCPAddr).Port
 
+	if options.platform != "web" {
 		if err := adbReverse(socketName, agentPort); err != nil {
 			return fmt.Errorf("adb reverse: %w", err)
 		}
@@ -111,6 +77,7 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 			}
 		}()
 		fmt.Fprintf(stdout, "forwarded localabstract:%s -> tcp:%d\n", socketName, agentPort)
+	}
 
 	agentServer := agent.NewServer(listener)
 
@@ -126,7 +93,7 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 		acceptChannel <- acceptResult{connection: connection, err: acceptErr}
 	}()
 
-	if err := driverClient.Launch(ctx, options.bundleID, options.launcherActivity, false); err != nil {
+	if err := activeDriver.Launch(ctx, options.bundleID, false); err != nil {
 		return fmt.Errorf("launch app: %w", err)
 	}
 	fmt.Fprintf(stdout, "launched %s; waiting for SDK to connect (%.0fs timeout)\n", options.bundleID, sdkAcceptTimeout.Seconds())
@@ -185,14 +152,14 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 		IdleTimeout:     1 * time.Second,
 		BundleID:        options.bundleID,
 		Connection:      connection,
-		Driver:          driverClient,
+		Driver:          activeDriver,
 		Verifier:        verifierInstance,
 		TraceWriter:     traceWriter,
 		Logger:          newProgressLogger(stdout),
 	})
 
 	terminateCtx, terminateCancel := context.WithTimeout(context.Background(), 5*time.Second)
-	_ = driverClient.Terminate(terminateCtx)
+	_ = activeDriver.Terminate(terminateCtx)
 	terminateCancel()
 
 	if err != nil {
@@ -211,6 +178,62 @@ func runTestPipeline(ctx context.Context, options testOptions, stdout io.Writer)
 	return nil
 }
 
+// buildDriver creates the appropriate DeviceDriver for the platform and returns
+// a cleanup function. For web, ChromeDriver is used directly; for android/ios
+// the JVM sidecar is extracted, spawned, and dialed.
+func buildDriver(ctx context.Context, options testOptions, stdout io.Writer) (driver.DeviceDriver, func(), error) {
+	if options.platform == "web" {
+		d := chrome.New()
+		return d, func() { _ = d.Terminate(context.Background()) }, nil
+	}
+
+	sidecarDirectory := filepath.Join(os.TempDir(), "sanderling-sidecar")
+	jarPath, err := sidecar.Extract(sidecarDirectory)
+	if err != nil {
+		return nil, nil, fmt.Errorf("extract sidecar: %w", err)
+	}
+	fmt.Fprintf(stdout, "sidecar JAR: %s (size=%d)\n", jarPath, sidecar.EmbeddedSize())
+
+	sidecarPort, err := pickFreePort()
+	if err != nil {
+		return nil, nil, err
+	}
+	sidecarCommand := exec.CommandContext(ctx, "java", "-jar", jarPath,
+		"--port", strconv.Itoa(sidecarPort),
+		"--platform", options.platform,
+	)
+	sidecarCommand.Stdout = stdout
+	sidecarCommand.Stderr = stdout
+	sidecarCommand.Env = envWithAndroidPlatformTools(os.Environ())
+	if err := sidecarCommand.Start(); err != nil {
+		return nil, nil, fmt.Errorf("spawn sidecar: %w", err)
+	}
+	fmt.Fprintf(stdout, "sidecar pid=%d listening on 127.0.0.1:%d\n", sidecarCommand.Process.Pid, sidecarPort)
+
+	driverClient, err := driverSidecar.Dial(fmt.Sprintf("127.0.0.1:%d", sidecarPort))
+	if err != nil {
+		_ = sidecarCommand.Process.Kill()
+		return nil, nil, fmt.Errorf("dial sidecar: %w", err)
+	}
+	healthCtx, healthCancel := context.WithTimeout(ctx, sidecarStartupTimeout)
+	if err := driverClient.WaitForHealth(healthCtx, 250*time.Millisecond); err != nil {
+		healthCancel()
+		_ = sidecarCommand.Process.Kill()
+		_ = driverClient.Close()
+		return nil, nil, fmt.Errorf("sidecar health check: %w", err)
+	}
+	healthCancel()
+	fmt.Fprintln(stdout, "sidecar is healthy")
+
+	cleanup := func() {
+		_ = driverClient.Close()
+		if sidecarCommand.Process != nil {
+			_ = sidecarCommand.Process.Kill()
+		}
+	}
+	return driverClient, cleanup, nil
+}
+
 // resolveSpecAPIPath returns the path to pkg/spec-api/src/index.ts inside
 // a sanderling source checkout, searched upward from the spec file and the cwd.
 // Returns "" when not found, in which case esbuild resolves @sanderling/spec via

go.mod

@@ -12,11 +12,18 @@ require (
 )
 
 require (
+	github.com/chromedp/cdproto v0.0.0-20260321001828-e3e3800016bc // indirect
+	github.com/chromedp/chromedp v0.15.1 // indirect
+	github.com/chromedp/sysutil v1.1.0 // indirect
 	github.com/dlclark/regexp2 v1.11.4 // indirect
+	github.com/go-json-experiment/json v0.0.0-20260214004413-d219187c3433 // indirect
 	github.com/go-sourcemap/sourcemap v2.1.3+incompatible // indirect
+	github.com/gobwas/httphead v0.1.0 // indirect
+	github.com/gobwas/pool v0.2.1 // indirect
+	github.com/gobwas/ws v1.4.0 // indirect
 	github.com/google/pprof v0.0.0-20230207041349-798e818bf904 // indirect
 	golang.org/x/net v0.49.0 // indirect
-	golang.org/x/sys v0.40.0 // indirect
+	golang.org/x/sys v0.42.0 // indirect
 	golang.org/x/text v0.33.0 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20260120221211-b8f7ae30c516 // indirect
 )

go.sum

@@ -2,6 +2,12 @@ github.com/Masterminds/semver/v3 v3.2.1 h1:RN9w6+7QoMeJVGyfmbcgs28Br8cvmnucEXnY0
 github.com/Masterminds/semver/v3 v3.2.1/go.mod h1:qvl/7zhW3nngYb5+80sSMF+FG2BjYrf8m9wsX0PNOMQ=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
+github.com/chromedp/cdproto v0.0.0-20260321001828-e3e3800016bc h1:wkN/LMi5vc60pBRWx6qpbk/aEvq3/ZVNpnMvsw8PVVU=
+github.com/chromedp/cdproto v0.0.0-20260321001828-e3e3800016bc/go.mod h1:cbyjALe67vDvlvdiG9369P8w5U2w6IshwtyD2f2Tvag=
+github.com/chromedp/chromedp v0.15.1 h1:EJWiPm7BNqDqjYy6U0lTSL5wNH+iNt9GjC3a4gfjNyQ=
+github.com/chromedp/chromedp v0.15.1/go.mod h1:CdTHtUqD/dqaFw/cvFWtTydoEQS44wLBuwbMR9EkOY4=
+github.com/chromedp/sysutil v1.1.0 h1:PUFNv5EcprjqXZD9nJb9b/c9ibAbxiYo4exNWZyipwM=
+github.com/chromedp/sysutil v1.1.0/go.mod h1:WiThHUdltqCNKGc4gaU50XgYjwjYIhKWoHGPTUfWTJ8=
 github.com/dlclark/regexp2 v1.11.4 h1:rPYF9/LECdNymJufQKmri9gV604RvvABwgOA8un7yAo=
 github.com/dlclark/regexp2 v1.11.4/go.mod h1:DHkYz0B9wPfa6wondMfaivmHpzrQ3v9q8cnmRbL6yW8=
 github.com/dop251/goja v0.0.0-20260311135729-065cd970411c h1:OcLmPfx1T1RmZVHHFwWMPaZDdRf0DBMZOFMVWJa7Pdk=
@@ -10,12 +16,20 @@ github.com/evanw/esbuild v0.28.0 h1:V96ghtc5p5JnNUQIUsc5H3kr+AcFcMqOJll2ZmJW6Lo=
 github.com/evanw/esbuild v0.28.0/go.mod h1:D2vIQZqV/vIf/VRHtViaUtViZmG7o+kKmlBfVQuRi48=
 github.com/fsnotify/fsnotify v1.9.0 h1:2Ml+OJNzbYCTzsxtv8vKSFD9PbJjmhYF14k/jKC7S9k=
 github.com/fsnotify/fsnotify v1.9.0/go.mod h1:8jBTzvmWwFyi3Pb8djgCCO5IBqzKJ/Jwo8TRcHyHii0=
+github.com/go-json-experiment/json v0.0.0-20260214004413-d219187c3433 h1:vymEbVwYFP/L05h5TKQxvkXoKxNvTpjxYKdF1Nlwuao=
+github.com/go-json-experiment/json v0.0.0-20260214004413-d219187c3433/go.mod h1:tphK2c80bpPhMOI4v6bIc2xWywPfbqi1Z06+RcrMkDg=
 github.com/go-logr/logr v1.4.3 h1:CjnDlHq8ikf6E492q6eKboGOC0T8CDaOvkHCIg8idEI=
 github.com/go-logr/logr v1.4.3/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ40KvY=
 github.com/go-logr/stdr v1.2.2 h1:hSWxHoqTgW2S2qGc0LTAI563KZ5YKYRhT3MFKZMbjag=
 github.com/go-logr/stdr v1.2.2/go.mod h1:mMo/vtBO5dYbehREoey6XUKy/eSumjCCveDpRre4VKE=
 github.com/go-sourcemap/sourcemap v2.1.3+incompatible h1:W1iEw64niKVGogNgBN3ePyLFfuisuzeidWPMPWmECqU=
 github.com/go-sourcemap/sourcemap v2.1.3+incompatible/go.mod h1:F8jJfvm2KbVjc5NqelyYJmf/v5J0dwNLS2mL4sNA1Jg=
+github.com/gobwas/httphead v0.1.0 h1:exrUm0f4YX0L7EBwZHuCF4GDp8aJfVeBrlLQrs6NqWU=
+github.com/gobwas/httphead v0.1.0/go.mod h1:O/RXo79gxV8G+RqlR/otEwx4Q36zl9rqC5u12GKvMCM=
+github.com/gobwas/pool v0.2.1 h1:xfeeEhW7pwmX8nuLVlqbzVc7udMDrwetjEv+TZIz1og=
+github.com/gobwas/pool v0.2.1/go.mod h1:q8bcK0KcYlCgd9e7WYLm9LpyS+YeLd8JVDW6WezmKEw=
+github.com/gobwas/ws v1.4.0 h1:CTaoG1tojrh4ucGPcoJFiAQUAsEWekEWvLy7GsVNqGs=
+github.com/gobwas/ws v1.4.0/go.mod h1:G3gNqMNtPppf5XUz7O4shetPpcZ1VJ7zt18dlUeakrc=
 github.com/golang/protobuf v1.5.4 h1:i7eJL8qZTpSEXOPTxNKhASYpMn+8e5Q6AdndVa1dWek=
 github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6rSs7xps=
 github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
@@ -41,8 +55,11 @@ golang.org/x/net v0.49.0/go.mod h1:/ysNB2EvaqvesRkuLAyjI1ycPZlQHM3q01F02UY/MV8=
 golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
 golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.40.0 h1:DBZZqJ2Rkml6QMQsZywtnjnnGvHza6BTfYFWY9kjEWQ=
 golang.org/x/sys v0.40.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
+golang.org/x/sys v0.42.0 h1:omrd2nAlyT5ESRdCLYdm3+fMfNFE/+Rf4bDIQImRJeo=
+golang.org/x/sys v0.42.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
 golang.org/x/text v0.33.0 h1:B3njUFyqtHDUI5jMn1YIr5B0IE2U0qck04r6d4KPAxE=
 golang.org/x/text v0.33.0/go.mod h1:LuMebE6+rBincTi9+xWTY8TztLzKHc/9C1uBCG27+q8=
 gonum.org/v1/gonum v0.17.0 h1:VbpOemQlsSMrYmn7T2OUvQ4dqxQXU+ouZFQsZOx50z4=

internal/driver/chrome/driver.go

@@ -0,0 +1,282 @@
+package chrome
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/chromedp/cdproto/input"
+	"github.com/chromedp/cdproto/network"
+	"github.com/chromedp/cdproto/runtime"
+	"github.com/chromedp/chromedp"
+
+	"github.com/priyanshujain/sanderling/internal/driver"
+)
+
+// Driver implements DeviceDriver via chromedp for web platform testing.
+type Driver struct {
+	allocCtx    context.Context
+	allocCancel context.CancelFunc
+	tabCtx      context.Context
+	tabCancel   context.CancelFunc
+
+	logsMu sync.Mutex
+	logs   []driver.LogEntry
+}
+
+// New creates a new ChromeDriver. Call Terminate when done.
+func New() *Driver {
+	allocCtx, allocCancel := chromedp.NewExecAllocator(context.Background(),
+		append(chromedp.DefaultExecAllocatorOptions[:],
+			chromedp.Flag("headless", true),
+			chromedp.Flag("disable-gpu", true),
+		)...,
+	)
+	tabCtx, tabCancel := chromedp.NewContext(allocCtx)
+
+	d := &Driver{
+		allocCtx:    allocCtx,
+		allocCancel: allocCancel,
+		tabCtx:      tabCtx,
+		tabCancel:   tabCancel,
+	}
+
+	chromedp.ListenTarget(tabCtx, func(ev any) {
+		e, ok := ev.(*runtime.EventConsoleAPICalled)
+		if !ok {
+			return
+		}
+		var parts []string
+		for _, arg := range e.Args {
+			if arg.Value != nil {
+				var s string
+				if err := json.Unmarshal(arg.Value, &s); err == nil {
+					parts = append(parts, s)
+				} else {
+					parts = append(parts, string(arg.Value))
+				}
+			}
+		}
+		level := strings.ToUpper(string(e.Type))
+		if level == "LOG" {
+			level = "I"
+		}
+		d.logsMu.Lock()
+		d.logs = append(d.logs, driver.LogEntry{
+			UnixMillis: int64(e.Timestamp.Time().UnixMilli()),
+			Level:      level,
+			Tag:        "console",
+			Message:    strings.Join(parts, " "),
+		})
+		d.logsMu.Unlock()
+	})
+
+	return d
+}
+
+func (d *Driver) Launch(ctx context.Context, bundleID string, clearState bool) error {
+	if clearState {
+		if err := chromedp.Run(d.tabCtx, network.ClearBrowserCookies()); err != nil {
+			return fmt.Errorf("clear cookies: %w", err)
+		}
+		if err := chromedp.Run(d.tabCtx, chromedp.ActionFunc(func(ctx context.Context) error {
+			_, exp, err := runtime.Evaluate(`localStorage.clear(); sessionStorage.clear();`).Do(ctx)
+			if exp != nil {
+				return fmt.Errorf("clear storage: %s", exp.Text)
+			}
+			return err
+		})); err != nil {
+			return fmt.Errorf("clear storage: %w", err)
+		}
+	}
+	return chromedp.Run(d.tabCtx, chromedp.Navigate(bundleID))
+}
+
+func (d *Driver) Terminate(_ context.Context) error {
+	d.tabCancel()
+	d.allocCancel()
+	return nil
+}
+
+func (d *Driver) Tap(_ context.Context, x, y int) error {
+	return chromedp.Run(d.tabCtx,
+		chromedp.MouseClickXY(float64(x), float64(y)),
+	)
+}
+
+func (d *Driver) TapSelector(_ context.Context, selector string) error {
+	return chromedp.Run(d.tabCtx,
+		chromedp.Click(selector, chromedp.NodeVisible),
+	)
+}
+
+func (d *Driver) InputText(_ context.Context, text string) error {
+	return chromedp.Run(d.tabCtx,
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			return input.DispatchKeyEvent(input.KeyChar).WithText(text).Do(ctx)
+		}),
+	)
+}
+
+func (d *Driver) Swipe(_ context.Context, fromX, fromY, toX, toY int, duration time.Duration) error {
+	millis := max(duration.Milliseconds(), 50)
+	script := fmt.Sprintf(`
+(function() {
+  const el = document.elementFromPoint(%d, %d);
+  if (!el) return;
+  const steps = Math.max(1, Math.floor(%d / 16));
+  const dx = (%d - %d) / steps;
+  const dy = (%d - %d) / steps;
+  el.dispatchEvent(new PointerEvent('pointerdown', {clientX: %d, clientY: %d, bubbles: true}));
+  for (let i = 1; i <= steps; i++) {
+    el.dispatchEvent(new PointerEvent('pointermove', {clientX: %d + dx*i, clientY: %d + dy*i, bubbles: true}));
+  }
+  el.dispatchEvent(new PointerEvent('pointerup', {clientX: %d, clientY: %d, bubbles: true}));
+})();`,
+		fromX, fromY,
+		millis,
+		toX, fromX, toY, fromY,
+		fromX, fromY,
+		fromX, fromY,
+		toX, toY,
+	)
+	return chromedp.Run(d.tabCtx, chromedp.Evaluate(script, nil))
+}
+
+func (d *Driver) PressKey(_ context.Context, key string) error {
+	k, ok := keyMap[key]
+	if !ok {
+		return fmt.Errorf("unsupported key: %q", key)
+	}
+	return chromedp.Run(d.tabCtx, chromedp.KeyEvent(k))
+}
+
+var keyMap = map[string]string{
+	"back":  "\b",
+	"home":  "\x00",
+	"enter": "\r",
+	"tab":   "\t",
+	"up":    "\x26",
+	"down":  "\x28",
+	"left":  "\x25",
+	"right": "\x27",
+}
+
+func (d *Driver) Hierarchy(_ context.Context) (string, error) {
+	script := `
+(function buildTree(el) {
+  const rect = el.getBoundingClientRect();
+  const attrs = {};
+  const bounds = '[' + Math.round(rect.left) + ',' + Math.round(rect.top) + ',' +
+    Math.round(rect.right) + ',' + Math.round(rect.bottom) + ']';
+  if (rect.width > 0 || rect.height > 0) attrs.bounds = bounds;
+  const text = (el.textContent || '').trim().slice(0, 200);
+  if (text) attrs.text = text;
+  if (el.id) attrs['resource-id'] = el.id;
+  const label = el.getAttribute('aria-label') || el.getAttribute('alt') || el.getAttribute('title') || '';
+  if (label) attrs['content-desc'] = label;
+  if (el.tagName) attrs['class'] = el.tagName.toLowerCase();
+  const isClickable = !!(el.onclick || el.tagName === 'A' || el.tagName === 'BUTTON' ||
+    el.tagName === 'INPUT' || el.tagName === 'SELECT' ||
+    el.getAttribute('role') === 'button' || el.getAttribute('onclick'));
+  const children = [];
+  for (const child of el.children) {
+    children.push(buildTree(child));
+  }
+  return {
+    attributes: attrs,
+    children: children,
+    clickable: isClickable || null,
+    enabled: (!el.disabled) || null,
+    focused: document.activeElement === el || null,
+    checked: el.checked || null,
+    selected: el.selected || null,
+  };
+})(document.body);`
+
+	var result any
+	if err := chromedp.Run(d.tabCtx, chromedp.Evaluate(script, &result)); err != nil {
+		return "", fmt.Errorf("hierarchy: %w", err)
+	}
+	bytes, err := json.Marshal(result)
+	if err != nil {
+		return "", fmt.Errorf("hierarchy marshal: %w", err)
+	}
+	return string(bytes), nil
+}
+
+func (d *Driver) Screenshot(_ context.Context) (driver.Image, error) {
+	var buf []byte
+	if err := chromedp.Run(d.tabCtx, chromedp.CaptureScreenshot(&buf)); err != nil {
+		return driver.Image{}, fmt.Errorf("screenshot: %w", err)
+	}
+	w, h := pngDimensions(buf)
+	return driver.Image{PNG: buf, Width: w, Height: h}, nil
+}
+
+func (d *Driver) RecentLogs(_ context.Context, since time.Time, minLevel string) ([]driver.LogEntry, error) {
+	sinceMillis := since.UnixMilli()
+	d.logsMu.Lock()
+	defer d.logsMu.Unlock()
+	var result []driver.LogEntry
+	for _, entry := range d.logs {
+		if entry.UnixMillis < sinceMillis {
+			continue
+		}
+		if minLevel != "" && !meetsLevel(entry.Level, minLevel) {
+			continue
+		}
+		result = append(result, entry)
+	}
+	return result, nil
+}
+
+func (d *Driver) WaitForIdle(_ context.Context, _ time.Duration) error {
+	return chromedp.Run(d.tabCtx, chromedp.WaitReady("body", chromedp.ByQuery))
+}
+
+func (d *Driver) Health(_ context.Context) (driver.Health, error) {
+	select {
+	case <-d.tabCtx.Done():
+		return driver.Health{Ready: false, Version: "chrome", Platform: "web"}, nil
+	default:
+		return driver.Health{Ready: true, Version: "chrome", Platform: "web"}, nil
+	}
+}
+
+func (d *Driver) Metrics(_ context.Context, _ string) (driver.Metrics, error) {
+	var result map[string]any
+	script := `
+(function() {
+  const mem = performance.memory || {};
+  return {heap: mem.usedJSHeapSize || 0, totalMem: mem.totalJSHeapSize || 0};
+})()`
+	if err := chromedp.Run(d.tabCtx, chromedp.Evaluate(script, &result)); err != nil {
+		return driver.Metrics{}, nil
+	}
+	heap, _ := result["heap"].(float64)
+	total, _ := result["totalMem"].(float64)
+	return driver.Metrics{
+		HeapBytes:        int64(heap),
+		TotalMemoryBytes: int64(total),
+	}, nil
+}
+
+func meetsLevel(level, minLevel string) bool {
+	order := map[string]int{"V": 0, "D": 1, "I": 2, "W": 3, "E": 4, "F": 5}
+	return order[level] >= order[minLevel]
+}
+
+func pngDimensions(png []byte) (int, int) {
+	if len(png) < 24 {
+		return 0, 0
+	}
+	w := int(png[16])<<24 | int(png[17])<<16 | int(png[18])<<8 | int(png[19])
+	h := int(png[20])<<24 | int(png[21])<<16 | int(png[22])<<8 | int(png[23])
+	return w, h
+}
+
+var _ driver.DeviceDriver = (*Driver)(nil)

internal/driver/driver.go

@@ -5,15 +5,12 @@ import (
 	"time"
 )
 
-// Driver abstracts the platform-specific UI automation backend. v0.1 surface
-// matches proto/driverpb/driver.proto. The Maestro sidecar implementation
-// lives under driver/maestro; tests use driver/mock.
-type Driver interface {
-	// Launch asks the backend to bring the target app to the foreground.
-	// launcherActivity is an optional "<pkg>/<activity>" component that
-	// overrides the backend's default launcher resolution — needed for
-	// apps that declare multiple MAIN+LAUNCHER activities.
-	Launch(ctx context.Context, bundleID, launcherActivity string, clearState bool) error
+// DeviceDriver abstracts the platform-specific UI automation backend. v0.1
+// surface matches proto/driverpb/driver.proto. The sidecar implementation
+// lives under driver/sidecar; the web implementation under driver/chrome;
+// tests use driver/mock.
+type DeviceDriver interface {
+	Launch(ctx context.Context, bundleID string, clearState bool) error
 	Terminate(ctx context.Context) error
 
 	Tap(ctx context.Context, x, y int) error
@@ -24,7 +21,7 @@ type Driver interface {
 
 	Hierarchy(ctx context.Context) (string, error)
 	Screenshot(ctx context.Context) (Image, error)
-	// RecentLogs returns logcat entries at or after `since`, filtered to
+	// RecentLogs returns log entries at or after `since`, filtered to
 	// `minLevel` or above. An empty minLevel defaults to "E".
 	RecentLogs(ctx context.Context, since time.Time, minLevel string) ([]LogEntry, error)
 

internal/driver/mock/mock.go

@@ -29,7 +29,6 @@ const (
 type Action struct {
 	Kind       ActionKind
 	BundleID   string
-	LauncherActivity string
 	ClearState bool
 	X, Y       int
 	FromX, FromY int
@@ -89,11 +88,11 @@ func (d *Driver) failure(kind ActionKind) error {
 	return d.Failures[kind]
 }
 
-func (d *Driver) Launch(ctx context.Context, bundleID, launcherActivity string, clearState bool) error {
+func (d *Driver) Launch(_ context.Context, bundleID string, clearState bool) error {
 	if err := d.failure(ActionLaunch); err != nil {
 		return err
 	}
-	d.record(Action{Kind: ActionLaunch, BundleID: bundleID, LauncherActivity: launcherActivity, ClearState: clearState})
+	d.record(Action{Kind: ActionLaunch, BundleID: bundleID, ClearState: clearState})
 	return nil
 }
 
@@ -220,4 +219,4 @@ func (d *Driver) Metrics(ctx context.Context, bundleID string) (driver.Metrics,
 	return d.MetricsData, nil
 }
 
-var _ driver.Driver = (*Driver)(nil)
+var _ driver.DeviceDriver = (*Driver)(nil)

internal/driver/mock/mock_test.go

@@ -29,7 +29,7 @@ func TestRecordsAllActionsInOrder(t *testing.T) {
 	mock := New()
 	ctx := context.Background()
 
-	if err := mock.Launch(ctx, "com.example", "com.example/.MainActivity", true); err != nil {
+	if err := mock.Launch(ctx, "com.example", true); err != nil {
 		t.Fatal(err)
 	}
 	if err := mock.Tap(ctx, 100, 200); err != nil {
@@ -137,5 +137,5 @@ func TestActionsReturnsCopy(t *testing.T) {
 }
 
 func TestSatisfiesDriverInterface(t *testing.T) {
-	var _ driver.Driver = New()
+	var _ driver.DeviceDriver = New()
 }

internal/driver/sidecar/client.go

@@ -1,4 +1,4 @@
-package maestro
+package sidecar
 
 import (
 	"context"
@@ -48,11 +48,10 @@ func (c *Client) WaitForHealth(ctx context.Context, pollInterval time.Duration)
 	}
 }
 
-func (c *Client) Launch(ctx context.Context, bundleID, launcherActivity string, clearState bool) error {
+func (c *Client) Launch(ctx context.Context, bundleID string, clearState bool) error {
 	_, err := c.stub.Launch(ctx, &driverpb.LaunchRequest{
 		BundleId:   bundleID,
 		ClearState: clearState,
-		LauncherActivity: launcherActivity,
 	})
 	return err
 }
@@ -165,4 +164,4 @@ func (c *Client) Metrics(ctx context.Context, bundleID string) (driver.Metrics,
 	}, nil
 }
 
-var _ driver.Driver = (*Client)(nil)
+var _ driver.DeviceDriver = (*Client)(nil)

internal/driver/sidecar/client_test.go

@@ -1,4 +1,4 @@
-package maestro
+package sidecar
 
 import (
 	"context"
@@ -22,7 +22,6 @@ type fakeServer struct {
 	healthReadyAfterCall int
 
 	launchedBundleID string
-	launcherActivity string
 	clearState       bool
 	terminateCalls   int
 	taps             []int32
@@ -55,7 +54,6 @@ func (s *fakeServer) Launch(_ context.Context, request *driverpb.LaunchRequest)
 	s.mutex.Lock()
 	defer s.mutex.Unlock()
 	s.launchedBundleID = request.GetBundleId()
-	s.launcherActivity = request.GetLauncherActivity()
 	s.clearState = request.GetClearState()
 	return &driverpb.Empty{}, nil
 }
@@ -193,10 +191,10 @@ func TestClient_LaunchAndTerminate(t *testing.T) {
 	client, _ := Dial(state.address)
 	defer client.Close()
 
-	if err := client.Launch(context.Background(), "com.example", "com.example/.MainActivity", true); err != nil {
+	if err := client.Launch(context.Background(), "com.example", true); err != nil {
 		t.Fatal(err)
 	}
-	if state.fake.launchedBundleID != "com.example" || !state.fake.clearState || state.fake.launcherActivity != "com.example/.MainActivity" {
+	if state.fake.launchedBundleID != "com.example" || !state.fake.clearState {
 		t.Errorf("launch payload wrong: %+v", state.fake)
 	}
 	if err := client.Terminate(context.Background()); err != nil {

internal/hierarchy/hierarchy.go

@@ -1,5 +1,5 @@
-// Package hierarchy parses the XML produced by `uiautomator dump` and
-// resolves selectors against it.
+// Package hierarchy parses the TreeNode JSON produced by the Maestro sidecar
+// and resolves selectors against it.
 //
 // Selector grammar (v0.1):
 //
@@ -10,7 +10,7 @@
 package hierarchy
 
 import (
-	"encoding/xml"
+	"encoding/json"
 	"fmt"
 	"regexp"
 	"strconv"
@@ -36,7 +36,7 @@ func (b Bounds) Width() int { return b.Right - b.Left }
 // Height returns the bounds' height.
 func (b Bounds) Height() int { return b.Bottom - b.Top }
 
-// Element is a flattened view of one uiautomator node.
+// Element is a flattened view of one hierarchy node.
 type Element struct {
 	ResourceID  string `json:"resourceId,omitempty"`
 	Text        string `json:"text,omitempty"`
@@ -56,38 +56,82 @@ type Tree struct {
 	Elements []*Element `json:"elements"`
 }
 
-// Parse parses a uiautomator-style XML dump.
-func Parse(xmlText string) (*Tree, error) {
-	xmlText = strings.TrimSpace(xmlText)
-	if xmlText == "" {
+// treeNodeJSON mirrors the Maestro TreeNode JSON structure.
+type treeNodeJSON struct {
+	Attributes map[string]string `json:"attributes"`
+	Children   []treeNodeJSON    `json:"children"`
+	Clickable  *bool             `json:"clickable"`
+	Enabled    *bool             `json:"enabled"`
+	Focused    *bool             `json:"focused"`
+	Checked    *bool             `json:"checked"`
+	Selected   *bool             `json:"selected"`
+}
+
+// Parse parses a Maestro TreeNode JSON hierarchy.
+func Parse(text string) (*Tree, error) {
+	text = strings.TrimSpace(text)
+	if text == "" {
 		return &Tree{}, nil
 	}
-	decoder := xml.NewDecoder(strings.NewReader(xmlText))
+	var root treeNodeJSON
+	if err := json.Unmarshal([]byte(text), &root); err != nil {
+		return nil, fmt.Errorf("hierarchy: %w", err)
+	}
 	tree := &Tree{}
-	for {
-		token, err := decoder.Token()
-		if err != nil {
-			if err.Error() == "EOF" {
-				break
-			}
-			break
-		}
-		start, ok := token.(xml.StartElement)
-		if !ok {
-			continue
-		}
-		if start.Name.Local != "node" {
-			continue
-		}
-		element, parseErr := elementFromStart(start)
-		if parseErr != nil {
-			return nil, parseErr
-		}
-		tree.Elements = append(tree.Elements, element)
-	}
+	walkNode(&root, tree)
 	return tree, nil
 }
 
+func walkNode(node *treeNodeJSON, tree *Tree) {
+	element := elementFromNode(node)
+	tree.Elements = append(tree.Elements, element)
+	for i := range node.Children {
+		walkNode(&node.Children[i], tree)
+	}
+}
+
+func elementFromNode(node *treeNodeJSON) *Element {
+	attrs := node.Attributes
+	element := &Element{}
+
+	element.ResourceID = attrs["resource-id"]
+	if element.ResourceID == "" {
+		element.ResourceID = attrs["identifier"]
+	}
+	element.Text = attrs["text"]
+	element.Description = attrs["content-desc"]
+	if element.Description == "" {
+		element.Description = attrs["accessibilityText"]
+	}
+	element.Class = attrs["class"]
+	element.Package = attrs["package"]
+
+	if node.Clickable != nil {
+		element.Clickable = *node.Clickable
+	}
+	if node.Enabled != nil {
+		element.Enabled = *node.Enabled
+	}
+	if node.Focused != nil {
+		element.Focused = *node.Focused
+	}
+	if node.Checked != nil {
+		element.Checked = *node.Checked
+	}
+	if node.Selected != nil {
+		element.Selected = *node.Selected
+	}
+
+	if b, ok := attrs["bounds"]; ok && b != "" {
+		bounds, err := parseBounds(b)
+		if err == nil {
+			element.Bounds = bounds
+		}
+	}
+
+	return element
+}
+
 // Find returns the first element matching the selector, or nil.
 func (t *Tree) Find(selector string) *Element {
 	kind, value, ok := parseSelector(selector)
@@ -131,11 +175,7 @@ func match(element *Element, kind, value string) bool {
 		if element.ResourceID == value {
 			return true
 		}
-		// ResourceID looks like "<package>:id/<suffix>".
-		if strings.HasSuffix(element.ResourceID, ":id/"+value) {
-			return true
-		}
-		return false
+		return strings.HasSuffix(element.ResourceID, ":id/"+value)
 	case "text":
 		return element.Text == value
 	case "desc":
@@ -147,59 +187,21 @@ func match(element *Element, kind, value string) bool {
 	}
 }
 
-func elementFromStart(start xml.StartElement) (*Element, error) {
-	element := &Element{}
-	var boundsText string
-	for _, attribute := range start.Attr {
-		switch attribute.Name.Local {
-		case "resource-id":
-			element.ResourceID = attribute.Value
-		case "text":
-			element.Text = attribute.Value
-		case "content-desc":
-			element.Description = attribute.Value
-		case "class":
-			element.Class = attribute.Value
-		case "package":
-			element.Package = attribute.Value
-		case "clickable":
-			element.Clickable = attribute.Value == "true"
-		case "enabled":
-			element.Enabled = attribute.Value == "true"
-		case "checked":
-			element.Checked = attribute.Value == "true"
-		case "focused":
-			element.Focused = attribute.Value == "true"
-		case "selected":
-			element.Selected = attribute.Value == "true"
-		case "bounds":
-			boundsText = attribute.Value
-		}
-	}
-	if boundsText != "" {
-		bounds, err := parseBounds(boundsText)
-		if err != nil {
-			return nil, fmt.Errorf("bounds %q: %w", boundsText, err)
-		}
-		element.Bounds = bounds
-	}
-	return element, nil
-}
-
-var boundsPattern = regexp.MustCompile(`^\[(-?\d+),(-?\d+)\]\[(-?\d+),(-?\d+)\]$`)
+// boundsPattern matches "[l,t,r,b]" (4-value Maestro format).
+var boundsPattern = regexp.MustCompile(`^\[(-?\d+),(-?\d+),(-?\d+),(-?\d+)\]$`)
 
 func parseBounds(text string) (Bounds, error) {
-	match := boundsPattern.FindStringSubmatch(text)
-	if match == nil {
-		return Bounds{}, fmt.Errorf("not in [L,T][R,B] form")
+	m := boundsPattern.FindStringSubmatch(text)
+	if m == nil {
+		return Bounds{}, fmt.Errorf("bounds %q: not in [L,T,R,B] form", text)
 	}
-	coordinates := make([]int, 4)
-	for index := range 4 {
-		value, err := strconv.Atoi(match[index+1])
+	coords := make([]int, 4)
+	for i := range 4 {
+		v, err := strconv.Atoi(m[i+1])
 		if err != nil {
 			return Bounds{}, err
 		}
-		coordinates[index] = value
+		coords[i] = v
 	}
-	return Bounds{Left: coordinates[0], Top: coordinates[1], Right: coordinates[2], Bottom: coordinates[3]}, nil
+	return Bounds{Left: coords[0], Top: coords[1], Right: coords[2], Bottom: coords[3]}, nil
 }

internal/hierarchy/hierarchy_test.go

@@ -2,14 +2,30 @@ package hierarchy
 
 import "testing"
 
-const sampleDump = `<?xml version='1.0' encoding='UTF-8' standalone='yes' ?>
-<hierarchy rotation="0">
-  <node class="android.widget.LinearLayout" package="app" bounds="[0,0][1080,2340]">
-    <node resource-id="app:id/title" text="Hello" bounds="[10,20][200,60]" clickable="false" enabled="true"/>
-    <node resource-id="app:id/row" text="Alice" content-desc="row" bounds="[0,100][1080,200]" clickable="true" enabled="true"/>
-    <node resource-id="app:id/row" text="Bob" content-desc="row" bounds="[0,200][1080,300]" clickable="true" enabled="true"/>
-  </node>
-</hierarchy>`
+// sampleDump is a Maestro TreeNode JSON equivalent of the old XML fixture.
+const sampleDump = `{
+  "attributes": {"class": "android.widget.LinearLayout", "package": "app", "bounds": "[0,0,1080,2340]"},
+  "children": [
+    {
+      "attributes": {"resource-id": "app:id/title", "text": "Hello", "bounds": "[10,20,200,60]"},
+      "children": [],
+      "clickable": false,
+      "enabled": true
+    },
+    {
+      "attributes": {"resource-id": "app:id/row", "text": "Alice", "content-desc": "row", "bounds": "[0,100,1080,200]"},
+      "children": [],
+      "clickable": true,
+      "enabled": true
+    },
+    {
+      "attributes": {"resource-id": "app:id/row", "text": "Bob", "content-desc": "row", "bounds": "[0,200,1080,300]"},
+      "children": [],
+      "clickable": true,
+      "enabled": true
+    }
+  ]
+}`
 
 func TestParseCountsNodes(t *testing.T) {
 	tree, err := Parse(sampleDump)
@@ -75,14 +91,76 @@ func TestUnknownSelector(t *testing.T) {
 }
 
 func TestDescPrefix(t *testing.T) {
-	xml := `<hierarchy>
-		<node content-desc="customer_row_abc-123" bounds="[0,0][100,100]"/>
-		<node content-desc="customer_row_def-456" bounds="[0,100][100,200]"/>
-		<node content-desc="supplier_row_xyz" bounds="[0,200][100,300]"/>
-	</hierarchy>`
-	tree, _ := Parse(xml)
+	input := `{
+	  "attributes": {},
+	  "children": [
+	    {"attributes": {"content-desc": "customer_row_abc-123", "bounds": "[0,0,100,100]"}, "children": []},
+	    {"attributes": {"content-desc": "customer_row_def-456", "bounds": "[0,100,100,200]"}, "children": []},
+	    {"attributes": {"content-desc": "supplier_row_xyz", "bounds": "[0,200,100,300]"}, "children": []}
+	  ]
+	}`
+	tree, _ := Parse(input)
 	rows := tree.FindAll("descPrefix:customer_row_")
 	if len(rows) != 2 {
 		t.Fatalf("want 2 customer rows, got %d", len(rows))
 	}
 }
+
+func TestBoolFieldsFromNode(t *testing.T) {
+	input := `{
+	  "attributes": {"resource-id": "x", "bounds": "[0,0,100,100]"},
+	  "children": [],
+	  "clickable": true,
+	  "enabled": false,
+	  "focused": true,
+	  "checked": true,
+	  "selected": false
+	}`
+	tree, err := Parse(input)
+	if err != nil {
+		t.Fatalf("Parse: %v", err)
+	}
+	if len(tree.Elements) != 1 {
+		t.Fatalf("want 1 element, got %d", len(tree.Elements))
+	}
+	el := tree.Elements[0]
+	if !el.Clickable {
+		t.Error("expected clickable=true")
+	}
+	if el.Enabled {
+		t.Error("expected enabled=false")
+	}
+	if !el.Focused {
+		t.Error("expected focused=true")
+	}
+	if !el.Checked {
+		t.Error("expected checked=true")
+	}
+	if el.Selected {
+		t.Error("expected selected=false")
+	}
+}
+
+func TestIdentifierFallback(t *testing.T) {
+	input := `{
+	  "attributes": {"identifier": "my-button", "bounds": "[0,0,100,100]"},
+	  "children": []
+	}`
+	tree, _ := Parse(input)
+	el := tree.Find("id:my-button")
+	if el == nil {
+		t.Fatal("expected match via identifier fallback")
+	}
+}
+
+func TestAccessibilityTextFallback(t *testing.T) {
+	input := `{
+	  "attributes": {"accessibilityText": "Close dialog", "bounds": "[0,0,50,50]"},
+	  "children": []
+	}`
+	tree, _ := Parse(input)
+	el := tree.Find("desc:Close dialog")
+	if el == nil {
+		t.Fatal("expected match via accessibilityText fallback")
+	}
+}

internal/runner/runner.go

@@ -25,7 +25,7 @@ type Options struct {
 
 	BundleID    string
 	Connection  *agent.Conn
-	Driver      driver.Driver
+	Driver      driver.DeviceDriver
 	Verifier    *verifier.Verifier
 	TraceWriter *trace.Writer
 	Logger      *slog.Logger
@@ -272,7 +272,7 @@ func screenFromSnapshot(snapshots map[string]json.RawMessage) (string, error) {
 	return screen, nil
 }
 
-func applyAction(ctx context.Context, drv driver.Driver, action verifier.Action, tree *hierarchy.Tree) error {
+func applyAction(ctx context.Context, drv driver.DeviceDriver, action verifier.Action, tree *hierarchy.Tree) error {
 	switch action.Kind {
 	case verifier.ActionKindTap:
 		x, y, ok := resolveCoordinates(action, tree)
@@ -326,7 +326,7 @@ func applyAction(ctx context.Context, drv driver.Driver, action verifier.Action,
 // collectLogs pulls recent error-level log entries from the driver since the
 // previous fetch. A failure is warned-on but not fatal: log capture is a
 // best-effort observability channel, not a correctness dependency.
-func collectLogs(ctx context.Context, drv driver.Driver, since time.Time) []verifier.LogEntry {
+func collectLogs(ctx context.Context, drv driver.DeviceDriver, since time.Time) []verifier.LogEntry {
 	entries, err := drv.RecentLogs(ctx, since, "E")
 	if err != nil {
 		return nil
@@ -374,7 +374,7 @@ func resolveCoordinates(action verifier.Action, tree *hierarchy.Tree) (int, int,
 	return 0, 0, false
 }
 
-func fetchHierarchy(ctx context.Context, drv driver.Driver) (*hierarchy.Tree, error) {
+func fetchHierarchy(ctx context.Context, drv driver.DeviceDriver) (*hierarchy.Tree, error) {
 	xmlText, err := drv.Hierarchy(ctx)
 	if err != nil {
 		return nil, err

internal/runner/runner_test.go

@@ -321,10 +321,7 @@ func TestRunner_StampsHierarchyResolvedBoundsAndResiduals(t *testing.T) {
 	state.startSDK(t)
 	state.acceptConnection(t)
 
-	state.mock.HierarchyJSON = `<?xml version="1.0"?>
-<hierarchy>
-  <node resource-id="com.fixture:id/next" clickable="true" enabled="true" bounds="[40,80][240,160]"/>
-</hierarchy>`
+	state.mock.HierarchyJSON = `{"attributes":{"resource-id":"com.fixture:id/next","bounds":"[40,80,240,160]"},"children":[],"clickable":true,"enabled":true}`
 
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()

internal/verifier/ax_integration_test.go

@@ -9,12 +9,13 @@ import (
 
 // TestStateAxFindWorks verifies that a Parse+PushSnapshot+extract round trip
 // actually lets the spec resolve selectors through state.ax.find.
+// Reads /tmp/live-dump.json (Maestro TreeNode JSON format); skipped if absent.
 func TestStateAxFindWorks(t *testing.T) {
-	xmlText, err := os.ReadFile("/tmp/live-dump.xml")
+	jsonText, err := os.ReadFile("/tmp/live-dump.json")
 	if err != nil {
-		t.Skip("live-dump.xml not present")
+		t.Skip("live-dump.json not present")
 	}
-	tree, err := hierarchy.Parse(string(xmlText))
+	tree, err := hierarchy.Parse(string(jsonText))
 	if err != nil {
 		t.Fatal(err)
 	}

internal/verifier/hierarchies_test.go

@@ -1,19 +1,19 @@
 package verifier
 
-const listHierarchyXML = `<?xml version="1.0" encoding="UTF-8"?>
-<hierarchy rotation="0">
-  <node index="0" class="android.widget.FrameLayout" package="test.app" bounds="[0,0][1080,2400]">
-    <node index="0" class="android.widget.TextView" text="Items" bounds="[100,200][900,300]" />
-    <node index="1" class="android.view.View" content-desc="primary_action" clickable="true" enabled="true" bounds="[64,2200][1016,2320]" />
-    <node index="2" class="android.view.View" content-desc="secondary_action" clickable="true" enabled="true" bounds="[980,80][1060,160]" />
-  </node>
-</hierarchy>`
+const listHierarchyJSON = `{
+  "attributes": {"class": "android.widget.FrameLayout", "package": "test.app", "bounds": "[0,0,1080,2400]"},
+  "children": [
+    {"attributes": {"class": "android.widget.TextView", "text": "Items", "bounds": "[100,200,900,300]"}, "children": []},
+    {"attributes": {"content-desc": "primary_action", "bounds": "[64,2200,1016,2320]"}, "children": [], "clickable": true, "enabled": true},
+    {"attributes": {"content-desc": "secondary_action", "bounds": "[980,80,1060,160]"}, "children": [], "clickable": true, "enabled": true}
+  ]
+}`
 
-const formHierarchyXML = `<?xml version="1.0" encoding="UTF-8"?>
-<hierarchy rotation="0">
-  <node index="0" class="android.widget.FrameLayout" package="test.app" bounds="[0,0][1080,2400]">
-    <node index="0" class="android.view.View" content-desc="text_field" clickable="true" enabled="true" bounds="[64,320][1016,440]" />
-    <node index="1" class="android.view.View" content-desc="primary_action" clickable="true" enabled="true" bounds="[64,2200][1016,2320]" />
-    <node index="2" class="android.view.View" content-desc="secondary_action" clickable="true" enabled="true" bounds="[32,80][112,160]" />
-  </node>
-</hierarchy>`
+const formHierarchyJSON = `{
+  "attributes": {"class": "android.widget.FrameLayout", "package": "test.app", "bounds": "[0,0,1080,2400]"},
+  "children": [
+    {"attributes": {"content-desc": "text_field", "bounds": "[64,320,1016,440]"}, "children": [], "clickable": true, "enabled": true},
+    {"attributes": {"content-desc": "primary_action", "bounds": "[64,2200,1016,2320]"}, "children": [], "clickable": true, "enabled": true},
+    {"attributes": {"content-desc": "secondary_action", "bounds": "[32,80,112,160]"}, "children": [], "clickable": true, "enabled": true}
+  ]
+}`

internal/verifier/spec_integration_test.go

@@ -65,7 +65,7 @@ func TestIntegrationSpecFiresInputActions(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	tree, err := hierarchy.Parse(formHierarchyXML)
+	tree, err := hierarchy.Parse(formHierarchyJSON)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -105,7 +105,7 @@ func TestIntegrationSpecPropertiesEvaluate(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	tree, err := hierarchy.Parse(listHierarchyXML)
+	tree, err := hierarchy.Parse(listHierarchyJSON)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -134,21 +134,21 @@ func TestIntegrationSpecPropertiesEvaluate(t *testing.T) {
 func TestIntegrationSpecActionsFireOnEachRoute(t *testing.T) {
 	cases := []struct {
 		name       string
-		xml        string
+		hierarchy  string
 		snapshots  Snapshots
 		expectKind ActionKind
 		expectOns  []string
 	}{
 		{
 			name:       "list",
-			xml:        listHierarchyXML,
+			hierarchy:  listHierarchyJSON,
 			snapshots:  listSnapshots(),
 			expectKind: ActionKindTap,
 			expectOns:  []string{"desc:primary_action", "desc:secondary_action"},
 		},
 		{
 			name:       "form",
-			xml:        formHierarchyXML,
+			hierarchy:  formHierarchyJSON,
 			snapshots:  formSnapshots(),
 			expectKind: ActionKindTap,
 			expectOns:  []string{"desc:primary_action", "desc:secondary_action"},
@@ -161,7 +161,7 @@ func TestIntegrationSpecActionsFireOnEachRoute(t *testing.T) {
 			if err := v.Load(bundleIntegrationSpec(t)); err != nil {
 				t.Fatal(err)
 			}
-			tree, err := hierarchy.Parse(tc.xml)
+			tree, err := hierarchy.Parse(tc.hierarchy)
 			if err != nil {
 				t.Fatal(err)
 			}

proto/driverpb/driver.pb.go

@@ -61,10 +61,6 @@ type LaunchRequest struct {
 	state         protoimpl.MessageState `protogen:"open.v1"`
 	BundleId      string                 `protobuf:"bytes,1,opt,name=bundle_id,json=bundleId,proto3" json:"bundle_id,omitempty"`
 	ClearState    bool                   `protobuf:"varint,2,opt,name=clear_state,json=clearState,proto3" json:"clear_state,omitempty"`
-	// Optional "<pkg>/<activity>" component to pass to `am start -n`. Needed
-	// when an app declares multiple MAIN+LAUNCHER activities where the first
-	// one reported by `pm dump` is an alias whose target class is disabled.
-	LauncherActivity string `protobuf:"bytes,3,opt,name=launcher_activity,json=launcherActivity,proto3" json:"launcher_activity,omitempty"`
 	unknownFields protoimpl.UnknownFields
 	sizeCache     protoimpl.SizeCache
 }
@@ -113,13 +109,6 @@ func (x *LaunchRequest) GetClearState() bool {
 	return false
 }
 
-func (x *LaunchRequest) GetLauncherActivity() string {
-	if x != nil {
-		return x.LauncherActivity
-	}
-	return ""
-}
-
 type Point struct {
 	state         protoimpl.MessageState `protogen:"open.v1"`
 	X             int32                  `protobuf:"varint,1,opt,name=x,proto3" json:"x,omitempty"`
@@ -850,12 +839,11 @@ var File_driverpb_driver_proto protoreflect.FileDescriptor
 const file_driverpb_driver_proto_rawDesc = "" +
 	"\n" +
 	"\x15driverpb/driver.proto\x12\x14sanderling.driver.v1\"\a\n" +
-	"\x05Empty\"z\n" +
+	"\x05Empty\"M\n" +
 	"\rLaunchRequest\x12\x1b\n" +
 	"\tbundle_id\x18\x01 \x01(\tR\bbundleId\x12\x1f\n" +
 	"\vclear_state\x18\x02 \x01(\bR\n" +
-	"clearState\x12+\n" +
-	"\x11launcher_activity\x18\x03 \x01(\tR\x10launcherActivity\"#\n" +
+	"clearState\"#\n" +
 	"\x05Point\x12\f\n" +
 	"\x01x\x18\x01 \x01(\x05R\x01x\x12\f\n" +
 	"\x01y\x18\x02 \x01(\x05R\x01y\"\x1c\n" +

proto/driverpb/driver.proto

@@ -30,10 +30,6 @@ message Empty {}
 message LaunchRequest {
   string bundle_id = 1;
   bool clear_state = 2;
-  // Optional "<pkg>/<activity>" component to pass to `am start -n`. Needed
-  // when an app declares multiple MAIN+LAUNCHER activities where the first
-  // one reported by `pm dump` is an alias whose target class is disabled.
-  string launcher_activity = 3;
 }
 
 message Point {

sidecar/build.gradle.kts

@@ -25,6 +25,7 @@ val maestroVersion = "1.40.0"
 
 dependencies {
     implementation("dev.mobile:maestro-client:$maestroVersion")
+    implementation("dev.mobile:maestro-utils:$maestroVersion")
 
     implementation("io.grpc:grpc-netty-shaded:$grpcVersion")
     implementation("io.grpc:grpc-protobuf:$grpcVersion")

sidecar/src/main/kotlin/dev/sanderling/sidecar/DriverBackend.kt

@@ -1,7 +1,7 @@
 package dev.sanderling.sidecar
 
 interface DriverBackend {
-    fun launch(bundleId: String, launcherActivity: String, clearState: Boolean)
+    fun launch(bundleId: String, clearState: Boolean)
     fun terminate(bundleId: String)
     fun tap(x: Int, y: Int)
     fun tapSelector(selector: String)
@@ -29,12 +29,96 @@ data class LogLine(
     val message: String,
 )
 
-/**
- * StubDriverBackend records calls but takes no real device action. Real
- * Maestro integration arrives in a follow-up; v0.1 wires the gRPC plumbing
- * end-to-end so the Go side can be exercised against a running sidecar
- * even before Maestro is plugged in.
- */
+internal fun readLogcat(serial: String?, sinceUnixMillis: Long, minLevel: String): List<LogLine> {
+    val level = if (minLevel.isEmpty()) "E" else minLevel
+    val since = if (sinceUnixMillis > 0) StubDriverBackend.formatAdbLogcatTimestamp(sinceUnixMillis) else null
+    val arguments = mutableListOf("logcat", "-d", "*:$level")
+    if (since != null) {
+        arguments.add("-T")
+        arguments.add(since)
+    }
+    return try {
+        val process = ProcessBuilder(adbCmd(serial) + arguments).redirectErrorStream(false).start()
+        val output = process.inputStream.bufferedReader().readText()
+        process.waitFor()
+        StubDriverBackend.parseLogcatOutput(output)
+    } catch (cause: Exception) {
+        println("adb logcat failed: $cause")
+        emptyList()
+    }
+}
+
+internal fun readProcMetrics(serial: String?, bundleId: String): MetricsSample {
+    if (bundleId.isEmpty()) return MetricsSample(0.0, 0L, 0L)
+    return try {
+        val pid = adbOutput(serial, listOf("shell", "pidof", bundleId))
+            .trim().split(Regex("\\s+")).firstOrNull()?.toIntOrNull()
+            ?: return MetricsSample(0.0, 0L, 0L)
+        val cpu = sampleCpuTwice(serial, pid)
+        val (rssBytes, vmSizeBytes) = sampleProcessMemory(serial, pid)
+        MetricsSample(cpu, rssBytes, vmSizeBytes)
+    } catch (cause: Exception) {
+        println("metrics capture failed: $cause")
+        MetricsSample(0.0, 0L, 0L)
+    }
+}
+
+private fun adbCmd(serial: String?): List<String> =
+    if (serial == null) listOf("adb") else listOf("adb", "-s", serial)
+
+private fun adbOutput(serial: String?, arguments: List<String>): String {
+    return try {
+        val process = ProcessBuilder(adbCmd(serial) + arguments).redirectErrorStream(false).start()
+        val output = process.inputStream.bufferedReader().readText()
+        process.waitFor()
+        output
+    } catch (cause: Exception) {
+        ""
+    }
+}
+
+private fun sampleCpuTwice(serial: String?, pid: Int): Double {
+    val sleepArg = "0.050"
+    val command = "cat /proc/$pid/stat; sleep $sleepArg; cat /proc/$pid/stat"
+    val output = adbOutput(serial, listOf("shell", command))
+    val lines = output.lines().filter { it.isNotBlank() }
+    if (lines.size < 2) return 0.0
+    val first = parseCpuTicks(lines[0]) ?: return 0.0
+    val second = parseCpuTicks(lines[1]) ?: return 0.0
+    val clockHz = adbOutput(serial, listOf("shell", "getconf", "CLK_TCK")).trim().toLongOrNull() ?: 100L
+    val deltaCpuNanos = (second - first) * 1_000_000_000.0 / clockHz.coerceAtLeast(1L)
+    return (deltaCpuNanos / 50_000_000.0) * 100.0
+}
+
+private fun parseCpuTicks(statLine: String): Long? {
+    val afterComm = statLine.substringAfterLast(')').trim()
+    val fields = afterComm.split(Regex("\\s+"))
+    if (fields.size < 13) return null
+    val utime = fields[11].toLongOrNull() ?: return null
+    val stime = fields[12].toLongOrNull() ?: return null
+    return utime + stime
+}
+
+private fun sampleProcessMemory(serial: String?, pid: Int): Pair<Long, Long> {
+    val status = adbOutput(serial, listOf("shell", "cat", "/proc/$pid/status"))
+    var rssKb = 0L
+    var vmSizeKb = 0L
+    for (raw in status.lineSequence()) {
+        val line = raw.trim()
+        when {
+            line.startsWith("VmRSS:") -> rssKb = parseKb(line) ?: rssKb
+            line.startsWith("VmSize:") -> vmSizeKb = parseKb(line) ?: vmSizeKb
+        }
+    }
+    return Pair(rssKb * 1024L, vmSizeKb * 1024L)
+}
+
+private fun parseKb(line: String): Long? {
+    val parts = line.split(Regex("\\s+"))
+    if (parts.size < 2) return null
+    return parts[1].toLongOrNull()
+}
+
 class StubDriverBackend(private val platform: String) : DriverBackend {
     @Volatile var launchCount: Int = 0
         private set
@@ -47,38 +131,13 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
     @Volatile var lastInputText: String? = null
         private set
 
-    override fun launch(bundleId: String, launcherActivity: String, clearState: Boolean) {
+    override fun launch(bundleId: String, clearState: Boolean) {
         launchCount++
         lastBundleId = bundleId
         if (clearState) {
             runAdb(listOf("shell", "pm", "clear", bundleId))
         }
-        val component = when {
-            launcherActivity.isEmpty() -> "$bundleId/${resolveLauncherActivity(bundleId)}"
-            launcherActivity.contains('/') -> launcherActivity
-            else -> "$bundleId/$launcherActivity"
-        }
-        runAdb(listOf("shell", "am", "start", "-W", "-n", component))
-    }
-
-    private fun resolveLauncherActivity(bundleId: String): String {
-        val output = captureAdb(
-            listOf(
-                "shell", "cmd", "package", "resolve-activity", "--brief",
-                "-a", "android.intent.action.MAIN",
-                "-c", "android.intent.category.LAUNCHER",
-                bundleId,
-            ),
-        )
-        return parseResolvedActivity(bundleId, output)
-            ?: throw IllegalStateException("could not resolve launcher activity for $bundleId: $output")
-    }
-
-    private fun captureAdb(arguments: List<String>): String {
-        val process = ProcessBuilder(listOf("adb") + arguments).redirectErrorStream(true).start()
-        val output = process.inputStream.bufferedReader().readText()
-        process.waitFor()
-        return output
+        runAdb(listOf("shell", "am", "start", "-W", "-n", "$bundleId/.MainActivity"))
     }
 
     companion object {
@@ -87,9 +146,6 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
         internal fun isAnimationCountIdle(grepOutput: String): Boolean =
             (grepOutput.trim().toIntOrNull() ?: 0) == 0
 
-        // parseResolvedActivity extracts the activity name from the output of
-        // `cmd package resolve-activity --brief`. The brief output is two
-        // lines: metadata, then `<pkg>/<activity>`.
         internal fun parseResolvedActivity(bundleId: String, output: String): String? {
             val prefix = "$bundleId/"
             for (line in output.lines()) {
@@ -101,8 +157,6 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
             return null
         }
 
-        // Hard cap on KEYCODE_DEL events per clear. Guards against a pathological
-        // hierarchy that reports an enormous text length for the focused field.
         internal const val MAX_CLEAR_DELETES: Int = 1024
 
         internal fun buildClearKeyevents(textLength: Int): List<String> {
@@ -113,11 +167,6 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
             return args
         }
 
-        // `adb shell input text` runs through a remote sh, so shell metacharacters
-        // in the payload would be interpreted by the device shell. Substitute
-        // spaces with %s (input's escape) and backslash-escape characters sh
-        // would otherwise expand. Keep this list conservative; anything not
-        // listed passes through literally.
         internal fun escapeForAdbInputText(text: String): String {
             val sb = StringBuilder(text.length)
             for (ch in text) {
@@ -131,9 +180,6 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
             return sb.toString()
         }
 
-        // Matches a uiautomator-dump <node ...> tag where `focused="true"` is
-        // present. Captures only the tag's attribute string so we can pull
-        // `text="..."` out of it without building a full XML tree.
         private val FOCUSED_NODE = Regex(
             "<node\\b([^>]*\\bfocused=\"true\"[^>]*)/?>",
         )
@@ -169,10 +215,6 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
             return "$seconds.${millis.toString().padStart(3, '0')}"
         }
 
-        // Logcat default threadtime format:
-        //   MM-dd HH:mm:ss.SSS  PID  TID L TAG: message
-        // The leading date is the local year-inferred date; we convert to a
-        // unix-millis best-effort using the current year.
         private val LOGCAT_LINE = Regex(
             "^(\\d{2})-(\\d{2}) (\\d{2}):(\\d{2}):(\\d{2})\\.(\\d{3})" +
                 "\\s+\\d+\\s+\\d+\\s+([VDIWEFS])\\s+([^:]+?):\\s?(.*)$",
@@ -215,16 +257,10 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
 
     override fun tapSelector(selector: String) {
         lastTapSelector = selector
-        // v0.1: selector resolution lives in Maestro proper; the stub
-        // records the selector so logs/traces show what was requested.
     }
 
     override fun inputText(text: String) {
         lastInputText = text
-        // `adb shell input text` types keystrokes at the caret, so repeated
-        // calls append. Clear the focused field first so the caller sees a
-        // pure replace: read the current value's length from the hierarchy,
-        // then move-end + N backspaces before typing.
         clearFocusedField()
         runAdb(listOf("shell", "input", "text", escapeForAdbInputText(text)))
     }
@@ -271,31 +307,14 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
         runAdb(listOf("shell", "input", "keyevent", keyCode))
     }
 
-    override fun recentLogs(sinceUnixMillis: Long, minLevel: String): List<LogLine> {
-        val level = if (minLevel.isEmpty()) "E" else minLevel
-        val since = if (sinceUnixMillis > 0) formatAdbLogcatTimestamp(sinceUnixMillis) else null
-        val arguments = mutableListOf("logcat", "-d", "*:$level")
-        if (since != null) {
-            arguments.add("-T")
-            arguments.add(since)
-        }
-        return try {
-            val process = ProcessBuilder(listOf("adb") + arguments).redirectErrorStream(false).start()
-            val output = process.inputStream.bufferedReader().readText()
-            process.waitFor()
-            parseLogcatOutput(output)
-        } catch (cause: Exception) {
-            println("adb logcat failed: $cause")
-            emptyList()
-        }
-    }
+    override fun recentLogs(sinceUnixMillis: Long, minLevel: String): List<LogLine> =
+        readLogcat(null, sinceUnixMillis, minLevel)
 
     data class SwipeRecord(val fromX: Int, val fromY: Int, val toX: Int, val toY: Int, val durationMillis: Long)
 
     private fun runAdb(arguments: List<String>) {
         try {
             val command = ProcessBuilder(listOf("adb") + arguments).redirectErrorStream(true).start()
-            // Drain output before waiting so a large write doesn't block the child.
             command.inputStream.bufferedReader().readText()
             command.waitFor()
         } catch (cause: Exception) {
@@ -345,129 +364,138 @@ class StubDriverBackend(private val platform: String) : DriverBackend {
 
     private fun isDeviceIdle(): Boolean {
         return try {
-            val output = captureAdb(
-                listOf("shell", "dumpsys window -a | grep -c mAnimating=true"),
-            )
+            val output = adbOutput(null, listOf("shell", "dumpsys window -a | grep -c mAnimating=true"))
             isAnimationCountIdle(output)
         } catch (cause: Exception) {
             false
         }
     }
 
-
     override fun healthy(): Boolean = true
 
-    // Stateful CPU delta tracker. Reading /proc/<pid>/stat gives cumulative
-    // utime+stime in jiffies; CPU % over a step is (delta_ticks / delta_wall)
-    // * 100. First call blocks briefly for a real reading instead of 0.
-    private data class CpuSample(val pid: Int, val ticks: Long, val wallNanos: Long)
-    @Volatile private var lastCpuSample: CpuSample? = null
-    private val clockTicksPerSecond: Long by lazy { resolveClockTicksPerSecond() }
-    private val cpuFirstSampleSleepMillis: Long = 50L
+    override fun metrics(bundleId: String): MetricsSample = readProcMetrics(null, bundleId)
+}
 
-    override fun metrics(bundleId: String): MetricsSample {
-        if (bundleId.isEmpty()) return MetricsSample(0.0, 0L, 0L)
-        return try {
-            val pid = runAdbOutput(listOf("shell", "pidof", bundleId)).trim().split(Regex("\\s+")).firstOrNull()?.toIntOrNull()
-                ?: run {
-                    lastCpuSample = null
-                    return MetricsSample(0.0, 0L, 0L)
+class MaestroDriverBackend(private val serial: String?) : DriverBackend {
+    private val dadb: dadb.Dadb
+    private val driver: maestro.drivers.AndroidDriver
+
+    init {
+        dadb = buildDadb(serial)
+        driver = maestro.drivers.AndroidDriver(dadb, 7001, "localhost")
+        driver.open()
     }
-            val cpu = sampleCpuPercent(pid)
-            val (rssBytes, vmSizeBytes) = sampleProcessMemory(pid)
-            MetricsSample(cpu, rssBytes, vmSizeBytes)
-        } catch (cause: Exception) {
-            println("metrics capture failed: $cause")
-            MetricsSample(0.0, 0L, 0L)
+
+    override fun launch(bundleId: String, clearState: Boolean) {
+        if (clearState) driver.clearAppState(bundleId)
+        driver.launchApp(bundleId, emptyMap(), java.util.UUID.randomUUID())
+    }
+
+    override fun terminate(bundleId: String) = driver.stopApp(bundleId)
+
+    override fun tap(x: Int, y: Int) = driver.tap(maestro.Point(x, y))
+
+    override fun tapSelector(selector: String) {
+        val root = driver.contentDescriptor(false)
+        val bounds = findBoundsBySelector(root, selector) ?: return
+        driver.tap(maestro.Point((bounds[0] + bounds[2]) / 2, (bounds[1] + bounds[3]) / 2))
+    }
+
+    override fun inputText(text: String) = driver.inputText(text)
+
+    override fun swipe(fromX: Int, fromY: Int, toX: Int, toY: Int, durationMillis: Long) =
+        driver.swipe(maestro.Point(fromX, fromY), maestro.Point(toX, toY), maxOf(durationMillis, 250L))
+
+    override fun pressKey(key: String) {
+        StubDriverBackend.KEY_MAP[key]?.let { keyCode ->
+            keyCodeToMaestro(keyCode)?.let { driver.pressKey(it) }
         }
     }
 
-    private fun sampleCpuPercent(pid: Int): Double {
-        val previous = lastCpuSample
-        if (previous != null && previous.pid == pid) {
-            val ticks = readCpuTicks(pid) ?: return 0.0
-            val now = System.nanoTime()
-            lastCpuSample = CpuSample(pid, ticks, now)
-            return cpuPercentFromDelta(ticks - previous.ticks, now - previous.wallNanos)
-        }
-        // No baseline for this PID: one adb round-trip with a device-side sleep
-        // so the first step gets a real reading instead of 0.
-        val pair = readCpuTicksPair(pid, cpuFirstSampleSleepMillis) ?: return 0.0
-        lastCpuSample = CpuSample(pid, pair.second, System.nanoTime())
-        return cpuPercentFromDelta(pair.second - pair.first, cpuFirstSampleSleepMillis * 1_000_000L)
+    override fun screenshot(): Triple<ByteArray, Int, Int> {
+        val buf = okio.Buffer()
+        driver.takeScreenshot(buf, false)
+        val bytes = buf.readByteArray()
+        return Triple(bytes, pngWidth(bytes), pngHeight(bytes))
     }
 
-    private fun readCpuTicks(pid: Int): Long? {
-        val stat = runAdbOutput(listOf("shell", "cat", "/proc/$pid/stat")).trim()
-        if (stat.isEmpty()) return null
-        return parseCpuTicks(stat)
+    override fun hierarchy(): String =
+        com.fasterxml.jackson.module.kotlin.jacksonObjectMapper().writeValueAsString(driver.contentDescriptor(false))
+
+    override fun recentLogs(sinceUnixMillis: Long, minLevel: String) =
+        readLogcat(serial, sinceUnixMillis, minLevel)
+
+    override fun waitForIdle(durationMillis: Long) {
+        driver.waitForAppToSettle(null, null, durationMillis.toInt())
     }
 
-    private fun readCpuTicksPair(pid: Int, sleepMillis: Long): Pair<Long, Long>? {
-        // "sleep 0.050" — toybox sleep accepts fractional seconds on modern Android.
-        val sleepArg = "0.${"%03d".format(sleepMillis)}"
-        val command = "cat /proc/$pid/stat; sleep $sleepArg; cat /proc/$pid/stat"
-        val output = runAdbOutput(listOf("shell", command))
-        val lines = output.lines().filter { it.isNotBlank() }
-        if (lines.size < 2) return null
-        val first = parseCpuTicks(lines[0]) ?: return null
-        val second = parseCpuTicks(lines[1]) ?: return null
-        return Pair(first, second)
-    }
+    override fun healthy() = runCatching { driver.contentDescriptor(false); true }.getOrElse { false }
 
-    private fun parseCpuTicks(statLine: String): Long? {
-        // /proc/<pid>/stat format: pid (comm) state ppid ... utime stime ...
-        // comm is parenthesized and may contain spaces; rsplit on ')' to skip it.
-        val afterComm = statLine.substringAfterLast(')').trim()
-        val fields = afterComm.split(Regex("\\s+"))
-        // After the ')' we are at the "state" field (index 0 in afterComm).
-        // utime is proc(14) = afterComm[11], stime is proc(15) = afterComm[12].
-        if (fields.size < 13) return null
-        val utime = fields[11].toLongOrNull() ?: return null
-        val stime = fields[12].toLongOrNull() ?: return null
-        return utime + stime
-    }
+    override fun metrics(bundleId: String) = readProcMetrics(serial, bundleId)
+}
 
-    private fun cpuPercentFromDelta(deltaTicks: Long, deltaWallNanos: Long): Double {
-        if (deltaTicks < 0 || deltaWallNanos <= 0) return 0.0
-        val tickHz = clockTicksPerSecond.coerceAtLeast(1L)
-        val deltaCpuNanos = deltaTicks * 1_000_000_000.0 / tickHz
-        return (deltaCpuNanos / deltaWallNanos) * 100.0
+private fun buildDadb(serial: String?): dadb.Dadb {
+    return if (serial == null) {
+        dadb.Dadb.create("localhost", 5555)
+    } else {
+        dadb.Dadb.create(serial.substringBefore(":"), serial.substringAfter(":").toIntOrNull() ?: 5555)
     }
+}
 
-    private fun resolveClockTicksPerSecond(): Long {
-        val output = runAdbOutput(listOf("shell", "getconf", "CLK_TCK")).trim()
-        return output.toLongOrNull() ?: 100L
-    }
+private fun findBoundsBySelector(root: maestro.TreeNode, selector: String): IntArray? {
+    val colon = selector.indexOf(':')
+    if (colon < 0) return null
+    val kind = selector.substring(0, colon)
+    val value = selector.substring(colon + 1)
+    return findBoundsInTree(root, kind, value)
+}
 
-    private fun sampleProcessMemory(pid: Int): Pair<Long, Long> {
-        val status = runAdbOutput(listOf("shell", "cat", "/proc/$pid/status"))
-        var rssKb = 0L
-        var vmSizeKb = 0L
-        for (raw in status.lineSequence()) {
-            val line = raw.trim()
-            when {
-                line.startsWith("VmRSS:") -> rssKb = parseKb(line) ?: rssKb
-                line.startsWith("VmSize:") -> vmSizeKb = parseKb(line) ?: vmSizeKb
+private fun findBoundsInTree(node: maestro.TreeNode, kind: String, value: String): IntArray? {
+    val attrs = node.attributes
+    val matches = when (kind) {
+        "id" -> attrs["resource-id"]?.let { it == value || it.endsWith(":id/$value") } == true
+        "text" -> attrs["text"] == value
+        "desc" -> attrs["content-desc"] == value
+        "descPrefix" -> attrs["content-desc"]?.startsWith(value) == true
+        else -> false
     }
+    if (matches) {
+        attrs["bounds"]?.let { b -> parseBounds(b)?.let { return it } }
     }
-        return Pair(rssKb * 1024L, vmSizeKb * 1024L)
+    for (child in node.children) {
+        findBoundsInTree(child, kind, value)?.let { return it }
     }
+    return null
+}
+
+private fun parseBounds(s: String): IntArray? {
+    val pattern = Regex("^\\[(-?\\d+),(-?\\d+),(-?\\d+),(-?\\d+)\\]$")
+    val m = pattern.matchEntire(s) ?: return null
+    return IntArray(4) { m.groupValues[it + 1].toInt() }
+}
+
+private fun pngWidth(bytes: ByteArray): Int {
+    if (bytes.size < 24) return 0
+    return (bytes[16].toInt() and 0xFF shl 24) or (bytes[17].toInt() and 0xFF shl 16) or
+        (bytes[18].toInt() and 0xFF shl 8) or (bytes[19].toInt() and 0xFF)
+}
 
-    private fun parseKb(line: String): Long? {
-        val parts = line.split(Regex("\\s+"))
-        if (parts.size < 2) return null
-        return parts[1].toLongOrNull()
-    }
+private fun pngHeight(bytes: ByteArray): Int {
+    if (bytes.size < 24) return 0
+    return (bytes[20].toInt() and 0xFF shl 24) or (bytes[21].toInt() and 0xFF shl 16) or
+        (bytes[22].toInt() and 0xFF shl 8) or (bytes[23].toInt() and 0xFF)
+}
 
-    private fun runAdbOutput(arguments: List<String>): String {
-        return try {
-            val process = ProcessBuilder(listOf("adb") + arguments).redirectErrorStream(false).start()
-            val output = process.inputStream.bufferedReader().readText()
-            process.waitFor()
-            output
-        } catch (cause: Exception) {
-            ""
-        }
+private fun keyCodeToMaestro(adbKeyCode: String): maestro.KeyCode? {
+    return when (adbKeyCode) {
+        "KEYCODE_BACK" -> maestro.KeyCode.BACK
+        "KEYCODE_HOME" -> maestro.KeyCode.HOME
+        "KEYCODE_ENTER" -> maestro.KeyCode.ENTER
+        "KEYCODE_TAB" -> maestro.KeyCode.TAB
+        "KEYCODE_DPAD_UP" -> maestro.KeyCode.REMOTE_UP
+        "KEYCODE_DPAD_DOWN" -> maestro.KeyCode.REMOTE_DOWN
+        "KEYCODE_DPAD_LEFT" -> maestro.KeyCode.REMOTE_LEFT
+        "KEYCODE_DPAD_RIGHT" -> maestro.KeyCode.REMOTE_RIGHT
+        else -> null
     }
 }

sidecar/src/main/kotlin/dev/sanderling/sidecar/DriverService.kt

@@ -23,15 +23,14 @@ import java.util.concurrent.atomic.AtomicReference
 
 class DriverService(
     private val platform: String = "android",
-    private val serial: String? = null,
-    private val backend: DriverBackend = StubDriverBackend(platform),
+    private val backend: DriverBackend,
 ) : DriverGrpc.DriverImplBase() {
 
     private val launchedBundleId = AtomicReference<String?>(null)
 
     override fun launch(request: LaunchRequest, responseObserver: StreamObserver<Empty>) {
         runRpc(responseObserver) {
-            backend.launch(request.bundleId, request.launcherActivity, request.clearState)
+            backend.launch(request.bundleId, request.clearState)
             launchedBundleId.set(request.bundleId)
             Empty.getDefaultInstance()
         }

sidecar/src/main/kotlin/dev/sanderling/sidecar/Main.kt

@@ -45,7 +45,12 @@ fun main(arguments: Array<String>) {
         if (index >= 0 && index + 1 < arguments.size) arguments[index + 1] else null
     }
 
-    val service = DriverService(platform = platform, serial = serial)
+    val backend: DriverBackend = if (platform == "android") {
+        MaestroDriverBackend(serial)
+    } else {
+        StubDriverBackend(platform)
+    }
+    val service = DriverService(platform = platform, backend = backend)
     val server = SidecarServer(port, service)
     val boundPort = server.start()
     println("sanderling-sidecar listening on 127.0.0.1:$boundPort platform=$platform")

sidecar/src/test/kotlin/dev/sanderling/sidecar/DriverServiceTest.kt

@@ -26,7 +26,7 @@ class DriverServiceTest {
 
     private fun newClient(backend: DriverBackend): DriverGrpc.DriverBlockingStub {
         val serverName = InProcessServerBuilder.generateName()
-        val service = DriverService(platform = "android", serial = null, backend = backend)
+        val service = DriverService(platform = "android", backend = backend)
         grpcCleanup.register(
             InProcessServerBuilder.forName(serverName).directExecutor().addService(service).build().start()
         )
@@ -43,7 +43,6 @@ class DriverServiceTest {
         client.launch(
             LaunchRequest.newBuilder()
                 .setBundleId("com.example")
-                .setLauncherActivity(".MainActivity")
                 .setClearState(true)
                 .build(),
         )

sidecar/src/test/kotlin/dev/sanderling/sidecar/SidecarServerTest.kt

@@ -6,7 +6,7 @@ import kotlin.test.assertTrue
 class SidecarServerTest {
     @Test
     fun startBindsEphemeralPortAndStopReleasesIt() {
-        val server = SidecarServer(port = 0, service = DriverService())
+        val server = SidecarServer(port = 0, service = DriverService(backend = StubDriverBackend("android")))
         val boundPort = server.start()
         try {
             assertTrue(boundPort > 0, "expected ephemeral port, got $boundPort")