ui-svelte: fix histogram calculation (#695)

- Fix the histogram calculation to use server provided generation
tokens/second.
- Move histogram to Activities page where it can exist with the rest of
the token metrics

Fixes #681

.github/workflows/ui-tests.yml

@@ -19,9 +19,6 @@ jobs:
 
   run-tests:
     runs-on: ubuntu-latest
-    defaults:
-      run:
-        working-directory: ui-svelte
     steps:
     - uses: actions/checkout@v4
 
@@ -32,11 +29,5 @@ jobs:
         cache: 'npm'
         cache-dependency-path: ui-svelte/package-lock.json
 
-    - name: Install dependencies
-      run: npm ci
-
-    - name: Type check
-      run: npm run check
-
-    - name: Run tests
-      run: npm test
+    - name: Run UI tests
+      run: make test-ui

.github/workflows/unified-docker.yml

@@ -36,6 +36,11 @@ on:
         type: boolean
         required: false
         default: true
+      push_to_ghcr:
+        description: "Push images to ghcr.io"
+        type: boolean
+        required: false
+        default: true
 
 permissions:
   contents: read
@@ -116,7 +121,7 @@ jobs:
           docker/unified/build-image.sh --${{ matrix.backend }}
 
       - name: Push to GitHub Container Registry
-        if: ${{ !env.ACT }}
+        if: ${{ !env.ACT && (github.event_name == 'schedule' || inputs.push_to_ghcr == true) }}
         run: |
           BASE_TAG="ghcr.io/mostlygeek/llama-swap:unified-${{ matrix.backend }}"
           DATE_TAG=$(date -u +%Y-%m-%d)

AGENTS.md

@@ -24,6 +24,7 @@ llama-swap is a light weight, transparent proxy server that provides automatic m
 - Run `gofmt -l .` before committing to verify formatting. Fix any reported files with `gofmt -w <file>`.
 - Use `make test-dev` after running new tests for a quick over all test run. This runs `go test` and `staticcheck`. Fix any static checking errors. Use this only when changes are made to any code under the `proxy/` directory
 - Use `make test-all` before completing work. This includes long running concurrency tests.
+- Use `make test-ui` after making changes to the UI in ui-svelte/
 
 ### Commit message example format:
 

Makefile

@@ -97,6 +97,9 @@ wol-proxy: $(BUILD_DIR)
 	@echo "Building wol-proxy"
 	go build -o $(BUILD_DIR)/wol-proxy-$(GOOS)-$(GOARCH)-$(shell date +%Y-%m-%d) cmd/wol-proxy/wol-proxy.go
 
+test-ui:
+	cd ui-svelte && npm ci && npm run check && npm test
+
 # Phony targets
-.PHONY: all clean ui mac windows simple-responder simple-responder-windows test test-all test-dev wol-proxy
+.PHONY: all clean ui mac windows simple-responder simple-responder-windows test test-all test-dev test-ui wol-proxy
 .PHONE: linux linux-arm64 linux-amd64

docker/unified/Dockerfile

@@ -42,6 +42,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     build-essential cmake git python3 python3-pip libssl-dev \
     curl ca-certificates ccache make wget software-properties-common \
     libvulkan-dev glslang-tools spirv-tools vulkan-validationlayers glslc \
+    spirv-headers \
     && rm -rf /var/lib/apt/lists/*
 
 WORKDIR /build
@@ -148,7 +149,7 @@ ARG IK_LLAMA_COMMIT_HASH=unknown
 ARG RUN_UID=0
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    python3-numpy python3-sentencepiece \
+    python3-numpy python3-sentencepiece python3-pip \
     && rm -rf /var/lib/apt/lists/*
 
 # Create non-root user when RUN_UID != 0
@@ -179,6 +180,9 @@ COPY --from=llama-build /install/bin/llama-cli /usr/local/bin/
 # Copy ik-llama-server (CUDA only; empty copy for vulkan)
 COPY --from=ik-llama-build /install/bin/ /usr/local/bin/
 
+# Install uv
+RUN pip install uv --break-system-packages
+
 # Copy llama-swap binary
 COPY --from=llama-swap-download /install/bin/llama-swap /usr/local/bin/
 COPY --from=llama-swap-download /install/llama-swap-version /tmp/

go.mod

@@ -4,7 +4,6 @@ go 1.26.1
 
 require (
 	github.com/billziss-gh/golib v0.2.0
-	github.com/fsnotify/fsnotify v1.9.0
 	github.com/gin-gonic/gin v1.10.0
 	github.com/klauspost/compress v1.18.5
 	github.com/stretchr/testify v1.9.0

go.sum

@@ -11,8 +11,6 @@ github.com/cloudwego/iasm v0.2.0/go.mod h1:8rXZaNYT2n95jn+zTI1sDr+IgcD2GVs0nlbbQ
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-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/gabriel-vasile/mimetype v1.4.3 h1:in2uUcidCuFcDKtdcBxlR0rJ1+fsokWf+uqxgUFjbI0=
 github.com/gabriel-vasile/mimetype v1.4.3/go.mod h1:d8uq/6HKRL6CGdk+aubisF/M5GcPfT7nKyLpA0lbSSk=
 github.com/gin-contrib/sse v0.1.0 h1:Y/yl/+YNO8GZSjAhjMsSuLt29uWRFHdHYUb5lYOV9qE=

llama-swap.go

@@ -9,14 +9,15 @@ import (
 	"os"
 	"os/signal"
 	"path/filepath"
+	"runtime"
 	"syscall"
 	"time"
 
-	"github.com/fsnotify/fsnotify"
 	"github.com/gin-gonic/gin"
 	"github.com/mostlygeek/llama-swap/event"
 	"github.com/mostlygeek/llama-swap/proxy"
 	"github.com/mostlygeek/llama-swap/proxy/config"
+	"github.com/mostlygeek/llama-swap/proxy/configwatcher"
 )
 
 var (
@@ -79,6 +80,17 @@ func main() {
 	sigChan := make(chan os.Signal, 1)
 	signal.Notify(sigChan, syscall.SIGINT, syscall.SIGTERM)
 
+	// Reload signals (SIGHUP on POSIX, none on Windows — Windows does not
+	// deliver SIGHUP). Always wired up so `kill -HUP` works regardless of
+	// --watch-config.
+	reloadChan := make(chan os.Signal, 1)
+	if runtime.GOOS != "windows" {
+		signal.Notify(reloadChan, syscall.SIGHUP)
+	}
+
+	// Context that bounds the lifetime of background watcher goroutines.
+	watcherCtx, watcherCancel := context.WithCancel(context.Background())
+
 	// Create server with initial handler
 	srv := &http.Server{
 		Addr: *listenStr,
@@ -121,52 +133,45 @@ func main() {
 	// load the initial proxy manager
 	reloadProxyManager()
 	debouncedReload := debounce(time.Second, reloadProxyManager)
-	if *watchConfig {
-		defer event.On(func(e proxy.ConfigFileChangedEvent) {
-			if e.ReloadingState == proxy.ReloadingStateStart {
-				debouncedReload()
-			}
-		})()
 
-		fmt.Println("Watching Configuration for changes")
+	// Listen for ConfigFileChangedEvent unconditionally so SIGHUP and the
+	// poll-based watcher both feed the same debounced reload pipeline. The
+	// UI also listens for the matching ReloadingStateEnd emitted from
+	// reloadProxyManager.
+	defer event.On(func(e proxy.ConfigFileChangedEvent) {
+		if e.ReloadingState == proxy.ReloadingStateStart {
+			debouncedReload()
+		}
+	})()
+
+	// SIGHUP (or platform-equivalent) → reload. Back-to-back signals collapse
+	// to one reload via the debounce window, which is the desired behavior.
+	go func() {
+		for range reloadChan {
+			fmt.Println("Received reload signal, reloading configuration")
+			event.Emit(proxy.ConfigFileChangedEvent{
+				ReloadingState: proxy.ReloadingStateStart,
+			})
+		}
+	}()
+
+	if *watchConfig {
 		go func() {
 			absConfigPath, err := filepath.Abs(*configPath)
 			if err != nil {
 				fmt.Printf("Error getting absolute path for watching config file: %v\n", err)
 				return
 			}
-			watcher, err := fsnotify.NewWatcher()
-			if err != nil {
-				fmt.Printf("Error creating file watcher: %v. File watching disabled.\n", err)
-				return
-			}
-
-			configDir := filepath.Dir(absConfigPath)
-			err = watcher.Add(configDir)
-			if err != nil {
-				fmt.Printf("Error adding config path directory (%s) to watcher: %v. File watching disabled.", configDir, err)
-				return
-			}
-
-			defer watcher.Close()
-			for {
-				select {
-				case changeEvent := <-watcher.Events:
-					if changeEvent.Name == absConfigPath && (changeEvent.Has(fsnotify.Write) || changeEvent.Has(fsnotify.Create) || changeEvent.Has(fsnotify.Remove)) {
-						event.Emit(proxy.ConfigFileChangedEvent{
-							ReloadingState: proxy.ReloadingStateStart,
-						})
-					} else if changeEvent.Name == filepath.Join(configDir, "..data") && changeEvent.Has(fsnotify.Create) {
-						// the change for k8s configmap
-						event.Emit(proxy.ConfigFileChangedEvent{
-							ReloadingState: proxy.ReloadingStateStart,
-						})
-					}
-
-				case err := <-watcher.Errors:
-					log.Printf("File watcher error: %v", err)
-				}
-			}
+			fmt.Println("Watching configuration for changes (poll-based, 2s interval)")
+			(&configwatcher.Watcher{
+				Path:     absConfigPath,
+				Interval: configwatcher.DefaultInterval,
+				OnChange: func() {
+					event.Emit(proxy.ConfigFileChangedEvent{
+						ReloadingState: proxy.ReloadingStateStart,
+					})
+				},
+			}).Run(watcherCtx)
 		}()
 	}
 
@@ -174,6 +179,7 @@ func main() {
 	go func() {
 		sig := <-sigChan
 		fmt.Printf("Received signal %v, shutting down...\n", sig)
+		watcherCancel()
 		ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
 		defer cancel()
 

proxy/configwatcher/watcher.go

@@ -0,0 +1,85 @@
+// Package configwatcher provides a simple cross-platform file watcher based
+// on os.Stat polling. It works correctly inside Docker containers where the
+// config file is bind-mounted as an individual file, and for k8s ConfigMap
+// projections (which present the file as a symlink to an atomically swapped
+// target) — both cases where inotify-based watchers are unreliable.
+package configwatcher
+
+import (
+	"context"
+	"errors"
+	"io/fs"
+	"log"
+	"os"
+	"time"
+)
+
+const DefaultInterval = 2 * time.Second
+
+type Watcher struct {
+	Path     string
+	Interval time.Duration
+	OnChange func()
+}
+
+type snapshot struct {
+	exists  bool
+	modTime time.Time
+	size    int64
+}
+
+// Run blocks until ctx is canceled. It polls Path on Interval and invokes
+// OnChange whenever the file's modification time or size changes, or when
+// the file reappears after being missing. The baseline poll establishes
+// initial state and does not fire OnChange.
+func (w *Watcher) Run(ctx context.Context) {
+	interval := w.Interval
+	if interval <= 0 {
+		interval = DefaultInterval
+	}
+
+	prev := stat(w.Path)
+
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+			cur := stat(w.Path)
+			if changed(prev, cur) && w.OnChange != nil {
+				w.OnChange()
+			}
+			prev = cur
+		}
+	}
+}
+
+func stat(path string) snapshot {
+	fi, err := os.Stat(path)
+	if err != nil {
+		if !errors.Is(err, fs.ErrNotExist) {
+			log.Printf("configwatcher: stat %s: %v", path, err)
+		}
+		return snapshot{}
+	}
+	return snapshot{
+		exists:  true,
+		modTime: fi.ModTime(),
+		size:    fi.Size(),
+	}
+}
+
+func changed(prev, cur snapshot) bool {
+	// Present → missing: stay quiet (likely a transient rename-style write).
+	// Missing → present: fire so we reload as soon as the file comes back.
+	if !cur.exists {
+		return false
+	}
+	if !prev.exists {
+		return true
+	}
+	return !prev.modTime.Equal(cur.modTime) || prev.size != cur.size
+}

proxy/configwatcher/watcher_test.go

@@ -0,0 +1,191 @@
+package configwatcher
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"runtime"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+const testInterval = 25 * time.Millisecond
+
+// startWatcher launches w.Run in a goroutine and returns a function that
+// cancels the context and waits for Run to return.
+func startWatcher(t *testing.T, w *Watcher) func() {
+	t.Helper()
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+	go func() {
+		w.Run(ctx)
+		close(done)
+	}()
+	return func() {
+		cancel()
+		select {
+		case <-done:
+		case <-time.After(2 * time.Second):
+			t.Fatal("watcher did not stop within 2s of cancel")
+		}
+	}
+}
+
+// waitForCount blocks until counter reaches want or timeout elapses.
+func waitForCount(t *testing.T, counter *int64, want int64, timeout time.Duration) bool {
+	t.Helper()
+	deadline := time.Now().Add(timeout)
+	for time.Now().Before(deadline) {
+		if atomic.LoadInt64(counter) >= want {
+			return true
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+	return false
+}
+
+func TestWatcher_NoFireOnBaseline(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "config.yaml")
+	require.NoError(t, os.WriteFile(path, []byte("a"), 0o644))
+
+	var n int64
+	stop := startWatcher(t, &Watcher{
+		Path:     path,
+		Interval: testInterval,
+		OnChange: func() { atomic.AddInt64(&n, 1) },
+	})
+	defer stop()
+
+	time.Sleep(testInterval * 5)
+	require.Equal(t, int64(0), atomic.LoadInt64(&n), "baseline poll must not fire")
+}
+
+func TestWatcher_DetectsModTimeChange(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "config.yaml")
+	require.NoError(t, os.WriteFile(path, []byte("a"), 0o644))
+
+	// Force a known baseline mtime.
+	base := time.Now().Add(-1 * time.Hour).Truncate(time.Second)
+	require.NoError(t, os.Chtimes(path, base, base))
+
+	var n int64
+	stop := startWatcher(t, &Watcher{
+		Path:     path,
+		Interval: testInterval,
+		OnChange: func() { atomic.AddInt64(&n, 1) },
+	})
+	defer stop()
+
+	// Let the baseline settle.
+	time.Sleep(testInterval * 2)
+
+	// Bump mtime well above the baseline so low-resolution filesystems still notice.
+	require.NoError(t, os.Chtimes(path, base.Add(10*time.Second), base.Add(10*time.Second)))
+
+	require.True(t, waitForCount(t, &n, 1, time.Second), "callback should fire after mtime change")
+}
+
+func TestWatcher_DetectsSizeChangeWithSameModTime(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "config.yaml")
+	require.NoError(t, os.WriteFile(path, []byte("a"), 0o644))
+
+	fi, err := os.Stat(path)
+	require.NoError(t, err)
+	originalMtime := fi.ModTime()
+
+	var n int64
+	stop := startWatcher(t, &Watcher{
+		Path:     path,
+		Interval: testInterval,
+		OnChange: func() { atomic.AddInt64(&n, 1) },
+	})
+	defer stop()
+	time.Sleep(testInterval * 2)
+
+	require.NoError(t, os.WriteFile(path, []byte("aaaaa"), 0o644))
+	// Reset mtime back to the original so size is the only signal.
+	require.NoError(t, os.Chtimes(path, originalMtime, originalMtime))
+
+	require.True(t, waitForCount(t, &n, 1, time.Second), "callback should fire on size change")
+}
+
+func TestWatcher_SymlinkTargetSwap(t *testing.T) {
+	dir := t.TempDir()
+	targetA := filepath.Join(dir, "targetA")
+	targetB := filepath.Join(dir, "targetB")
+	link := filepath.Join(dir, "config.yaml")
+
+	require.NoError(t, os.WriteFile(targetA, []byte("AAAA"), 0o644))
+	require.NoError(t, os.WriteFile(targetB, []byte("BBBBBBBB"), 0o644))
+
+	if err := os.Symlink(targetA, link); err != nil {
+		if runtime.GOOS == "windows" {
+			t.Skipf("symlink creation requires privilege on Windows: %v", err)
+		}
+		t.Fatalf("os.Symlink: %v", err)
+	}
+
+	var n int64
+	stop := startWatcher(t, &Watcher{
+		Path:     link,
+		Interval: testInterval,
+		OnChange: func() { atomic.AddInt64(&n, 1) },
+	})
+	defer stop()
+	time.Sleep(testInterval * 2)
+
+	// Atomic symlink swap (k8s ConfigMap pattern): create new symlink at a
+	// temp name, then rename over the existing one.
+	tmpLink := filepath.Join(dir, "config.yaml.tmp")
+	require.NoError(t, os.Symlink(targetB, tmpLink))
+	require.NoError(t, os.Rename(tmpLink, link))
+
+	require.True(t, waitForCount(t, &n, 1, time.Second), "callback should fire after symlink target swap")
+}
+
+func TestWatcher_FileMissingThenReturns(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "config.yaml")
+	require.NoError(t, os.WriteFile(path, []byte("a"), 0o644))
+
+	var n int64
+	stop := startWatcher(t, &Watcher{
+		Path:     path,
+		Interval: testInterval,
+		OnChange: func() { atomic.AddInt64(&n, 1) },
+	})
+	defer stop()
+	time.Sleep(testInterval * 2)
+
+	require.NoError(t, os.Remove(path))
+	time.Sleep(testInterval * 3)
+	require.Equal(t, int64(0), atomic.LoadInt64(&n), "removal alone must not fire")
+
+	require.NoError(t, os.WriteFile(path, []byte("b"), 0o644))
+	require.True(t, waitForCount(t, &n, 1, time.Second), "callback should fire when file returns")
+}
+
+func TestWatcher_ContextCancelStopsRun(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "config.yaml")
+	require.NoError(t, os.WriteFile(path, []byte("a"), 0o644))
+
+	w := &Watcher{Path: path, Interval: testInterval}
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+	go func() { w.Run(ctx); close(done) }()
+
+	time.Sleep(testInterval * 2)
+	cancel()
+	select {
+	case <-done:
+	case <-time.After(2 * time.Second):
+		t.Fatal("Run did not return within 2s of cancel")
+	}
+}

proxy/matrix.go

@@ -297,7 +297,7 @@ func (m *Matrix) Shutdown() {
 	wg.Wait()
 }
 
-// RunningModels returns model names currently in StateReady.
+// RunningModels returns model names currently in an active (non-stopped) state.
 func (m *Matrix) RunningModels() []string {
 	m.Lock()
 	defer m.Unlock()
@@ -308,7 +308,7 @@ func (m *Matrix) RunningModels() []string {
 func (m *Matrix) runningModels() []string {
 	var running []string
 	for id, process := range m.processes {
-		if process.CurrentState() == StateReady {
+		if process.CurrentState() != StateStopped && process.CurrentState() != StateShutdown {
 			running = append(running, id)
 		}
 	}

proxy/process.go

@@ -432,7 +432,10 @@ func (p *Process) start() error {
 
 // Stop will wait for inflight requests to complete before stopping the process.
 func (p *Process) Stop() {
+
+	// guard to prevent multiple goroutines from stopping
 	if !isValidTransition(p.CurrentState(), StateStopping) {
+		p.proxyLogger.Debugf("<%s> Stop() suppressing invalid transition from %s to StateStopping", p.ID, p.CurrentState())
 		return
 	}
 
@@ -445,13 +448,17 @@ func (p *Process) Stop() {
 // StopImmediately will transition the process to the stopping state and stop the process with a SIGTERM.
 // If the process does not stop within the specified timeout, it will be forcefully stopped with a SIGKILL.
 func (p *Process) StopImmediately() {
-	if !isValidTransition(p.CurrentState(), StateStopping) {
+
+	// guard to prevent multiple goroutines from stopping the process
+	enterState := p.CurrentState()
+	if !isValidTransition(enterState, StateStopping) {
+		p.proxyLogger.Debugf("<%s> StopImmediate() suppressing invalid transition from %s to StateStopping", p.ID, p.CurrentState())
 		return
 	}
 
-	p.proxyLogger.Debugf("<%s> Stopping process, current state: %s", p.ID, p.CurrentState())
-	if curState, err := p.swapState(StateReady, StateStopping); err != nil {
-		p.proxyLogger.Infof("<%s> Stop() Ready -> StateStopping err: %v, current state: %v", p.ID, err, curState)
+	p.proxyLogger.Debugf("<%s> Stopping process, enter state: %s", p.ID, enterState)
+	if curState, err := p.swapState(enterState, StateStopping); err != nil {
+		p.proxyLogger.Infof("<%s> Stop() %s -> StateStopping err: %v, current state: %v", p.ID, enterState, err, curState)
 		return
 	}
 

ui-svelte/src/components/ActivityStats.svelte

@@ -0,0 +1,72 @@
+<script lang="ts">
+  import { inFlightRequests, metrics } from "../stores/api";
+  import { persistentStore } from "../stores/persistent";
+  import { calculateHistogramData } from "../lib/histogram";
+  import TokenHistogram from "./TokenHistogram.svelte";
+
+  const nf = new Intl.NumberFormat();
+  const histogramCollapsed = persistentStore<boolean>("activity-histogram-collapsed", false);
+
+  let stats = $derived.by(() => {
+    const totalRequests = $metrics.length;
+    const totalInputTokens = $metrics.reduce((sum, m) => sum + m.input_tokens, 0);
+    const totalOutputTokens = $metrics.reduce((sum, m) => sum + m.output_tokens, 0);
+
+    const tokensPerSecond = $metrics
+      .filter((m) => m.tokens_per_second > 0)
+      .map((m) => m.tokens_per_second);
+
+    const histogramData = tokensPerSecond.length > 0
+      ? calculateHistogramData(tokensPerSecond, { minBins: 20, maxBins: 80, binScaling: 3 })
+      : null;
+
+    return {
+      totalRequests,
+      totalInputTokens,
+      totalOutputTokens,
+      inFlightRequests: $inFlightRequests,
+      histogramData,
+    };
+  });
+</script>
+
+<div class="card">
+  <button
+    class="flex items-center gap-1 px-4 pt-3 text-xs font-medium text-gray-500 dark:text-gray-400 hover:text-gray-700 dark:hover:text-gray-200 transition-colors"
+    onclick={() => $histogramCollapsed = !$histogramCollapsed}
+  >
+    <svg
+      class="w-3 h-3 transition-transform"
+      style="transform: rotate({$histogramCollapsed ? -90 : 0}deg)"
+      viewBox="0 0 16 16"
+      fill="currentColor"
+    >
+      <path d="M4.5 6l3.5 4 3.5-4H4.5z" />
+    </svg>
+    Tokens/sec Distribution
+  </button>
+  {#if !$histogramCollapsed}
+    {#if stats.histogramData}
+      <TokenHistogram data={stats.histogramData} />
+    {:else}
+      <div class="px-4 py-6 text-center text-sm text-gray-500 dark:text-gray-400">
+        No token speed data yet
+      </div>
+    {/if}
+  {/if}
+  <div class="grid grid-cols-3 gap-x-6 gap-y-1 px-4 pb-3 text-sm">
+    <div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Requests</div>
+    <div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Processed</div>
+    <div class="text-xs uppercase tracking-wider text-gray-500 dark:text-gray-400">Generated</div>
+    <div class="text-sm text-gray-700 dark:text-gray-300">
+      <span class="font-semibold">{nf.format(stats.totalRequests)}</span> completed,
+      <span class="font-semibold">{nf.format(stats.inFlightRequests)}</span> waiting
+    </div>
+    <div class="text-sm text-gray-700 dark:text-gray-300">
+      <span class="font-semibold">{nf.format(stats.totalInputTokens)}</span> tokens
+    </div>
+    <div class="text-sm text-gray-700 dark:text-gray-300">
+      <span class="font-semibold">{nf.format(stats.totalOutputTokens)}</span> tokens
+    </div>
+  </div>
+</div>

ui-svelte/src/components/StatsPanel.svelte

@@ -1,167 +0,0 @@
-<script lang="ts">
-  import { inFlightRequests, metrics } from "../stores/api";
-  import TokenHistogram from "./TokenHistogram.svelte";
-
-  interface HistogramData {
-    bins: number[];
-    min: number;
-    max: number;
-    binSize: number;
-    p99: number;
-    p95: number;
-    p50: number;
-  }
-
-  let stats = $derived.by(() => {
-    const totalRequests = $metrics.length;
-    if (totalRequests === 0) {
-      return {
-        totalRequests: 0,
-        totalInputTokens: 0,
-        totalOutputTokens: 0,
-        inFlightRequests: $inFlightRequests,
-        tokenStats: { p99: "0", p95: "0", p50: "0" },
-        histogramData: null,
-      };
-    }
-
-    const totalInputTokens = $metrics.reduce((sum, m) => sum + m.input_tokens, 0);
-    const totalOutputTokens = $metrics.reduce((sum, m) => sum + m.output_tokens, 0);
-
-    // Calculate token statistics using output_tokens and duration_ms
-    const validMetrics = $metrics.filter((m) => m.duration_ms > 0 && m.output_tokens > 0);
-    if (validMetrics.length === 0) {
-      return {
-        totalRequests,
-        totalInputTokens,
-        totalOutputTokens,
-        inFlightRequests: $inFlightRequests,
-        tokenStats: { p99: "0", p95: "0", p50: "0" },
-        histogramData: null,
-      };
-    }
-
-    // Calculate tokens/second for each valid metric
-    const tokensPerSecond = validMetrics.map((m) => m.output_tokens / (m.duration_ms / 1000));
-
-    // Sort for percentile calculation
-    const sortedTokensPerSecond = [...tokensPerSecond].sort((a, b) => a - b);
-
-    const p99 = sortedTokensPerSecond[Math.floor(sortedTokensPerSecond.length * 0.99)];
-    const p95 = sortedTokensPerSecond[Math.floor(sortedTokensPerSecond.length * 0.95)];
-    const p50 = sortedTokensPerSecond[Math.floor(sortedTokensPerSecond.length * 0.5)];
-
-    // Create histogram data
-    const min = Math.min(...tokensPerSecond);
-    const max = Math.max(...tokensPerSecond);
-    const binCount = Math.min(30, Math.max(10, Math.floor(tokensPerSecond.length / 5)));
-    const binSize = (max - min) / binCount;
-
-    const bins = Array(binCount).fill(0);
-    tokensPerSecond.forEach((value) => {
-      const binIndex = Math.min(Math.floor((value - min) / binSize), binCount - 1);
-      bins[binIndex]++;
-    });
-
-    const histogramData: HistogramData = {
-      bins,
-      min,
-      max,
-      binSize,
-      p99,
-      p95,
-      p50,
-    };
-
-    return {
-      totalRequests,
-      totalInputTokens,
-      totalOutputTokens,
-      inFlightRequests: $inFlightRequests,
-      tokenStats: {
-        p99: p99.toFixed(2),
-        p95: p95.toFixed(2),
-        p50: p50.toFixed(2),
-      },
-      histogramData,
-    };
-  });
-
-  const nf = new Intl.NumberFormat();
-</script>
-
-<div class="card">
-  <div class="rounded-lg overflow-hidden border border-card-border-inner">
-    <table class="min-w-full divide-y divide-card-border-inner">
-      <thead class="bg-secondary">
-        <tr>
-          <th class="px-4 py-3 text-left text-xs font-semibold uppercase tracking-wider text-txtmain">Requests</th>
-          <th class="px-4 py-3 text-left text-xs font-semibold uppercase tracking-wider text-txtmain border-l border-card-border-inner">
-            Processed
-          </th>
-          <th class="px-4 py-3 text-left text-xs font-semibold uppercase tracking-wider text-txtmain border-l border-card-border-inner">
-            Generated
-          </th>
-          <th class="px-4 py-3 text-left text-xs font-semibold uppercase tracking-wider text-txtmain border-l border-card-border-inner">
-            Token Stats (tokens/sec)
-          </th>
-        </tr>
-      </thead>
-
-      <tbody class="bg-surface divide-y divide-card-border-inner">
-        <tr class="hover:bg-secondary">
-          <td class="px-4 py-4 text-sm font-semibold text-gray-900 dark:text-white">
-            <div class="flex flex-col gap-1">
-              <span class="text-xs font-medium text-gray-500 dark:text-gray-400">Completed: {nf.format(stats.totalRequests)}</span>
-              <span class="text-xs font-medium text-gray-500 dark:text-gray-400">Waiting: {nf.format(stats.inFlightRequests)}</span>
-            </div>
-          </td>
-
-          <td class="px-4 py-4 text-sm text-gray-700 dark:text-gray-300 border-l border-gray-200 dark:border-white/10">
-            <div class="flex items-center gap-2">
-              <span class="text-sm font-medium">{nf.format(stats.totalInputTokens)}</span>
-              <span class="text-xs text-gray-500 dark:text-gray-400">tokens</span>
-            </div>
-          </td>
-
-          <td class="px-4 py-4 text-sm text-gray-700 dark:text-gray-300 border-l border-gray-200 dark:border-white/10">
-            <div class="flex items-center gap-2">
-              <span class="text-sm font-medium">{nf.format(stats.totalOutputTokens)}</span>
-              <span class="text-xs text-gray-500 dark:text-gray-400">tokens</span>
-            </div>
-          </td>
-
-          <td class="px-4 py-4 border-l border-gray-200 dark:border-white/10">
-            <div class="space-y-3">
-              <div class="grid grid-cols-3 gap-2 items-center">
-                <div class="text-center">
-                  <div class="text-xs text-gray-500 dark:text-gray-400">P50</div>
-                  <div class="mt-1 inline-block rounded-full bg-gray-100 dark:bg-white/5 px-3 py-1 text-sm font-semibold text-gray-800 dark:text-white">
-                    {stats.tokenStats.p50}
-                  </div>
-                </div>
-
-                <div class="text-center">
-                  <div class="text-xs text-gray-500 dark:text-gray-400">P95</div>
-                  <div class="mt-1 inline-block rounded-full bg-gray-100 dark:bg-white/5 px-3 py-1 text-sm font-semibold text-gray-800 dark:text-white">
-                    {stats.tokenStats.p95}
-                  </div>
-                </div>
-
-                <div class="text-center">
-                  <div class="text-xs text-gray-500 dark:text-gray-400">P99</div>
-                  <div class="mt-1 inline-block rounded-full bg-gray-100 dark:bg-white/5 px-3 py-1 text-sm font-semibold text-gray-800 dark:text-white">
-                    {stats.tokenStats.p99}
-                  </div>
-                </div>
-              </div>
-              {#if stats.histogramData}
-                <TokenHistogram data={stats.histogramData} />
-              {/if}
-            </div>
-          </td>
-        </tr>
-      </tbody>
-    </table>
-  </div>
-</div>

ui-svelte/src/components/TokenHistogram.svelte

@@ -1,23 +1,11 @@
 <script lang="ts">
-  interface HistogramData {
-    bins: number[];
-    min: number;
-    max: number;
-    binSize: number;
-    p99: number;
-    p95: number;
-    p50: number;
-  }
+  import type { HistogramData } from "../lib/types";
 
-  interface Props {
-    data: HistogramData;
-  }
+  let { data }: { data: HistogramData } = $props();
 
-  let { data }: Props = $props();
-
-  const height = 120;
-  const padding = { top: 10, right: 15, bottom: 25, left: 45 };
-  const viewBoxWidth = 600;
+  const height = 55;
+  const padding = { top: 5, right: 45, bottom: 15, left: 45 };
+  const viewBoxWidth = 1200;
   const chartWidth = viewBoxWidth - padding.left - padding.right;
   const chartHeight = height - padding.top - padding.bottom;
 
@@ -121,9 +109,5 @@
       {data.max.toFixed(1)}
     </text>
 
-    <!-- X-axis label -->
-    <text x={padding.left + chartWidth / 2} y={height - 2} font-size="10" fill="currentColor" opacity="0.6" text-anchor="middle">
-      Tokens/Second Distribution
-    </text>
   </svg>
 </div>

ui-svelte/src/lib/histogram.test.ts

@@ -0,0 +1,163 @@
+import { describe, it, expect } from "vitest";
+import { calculateHistogramData } from "./histogram";
+
+describe("calculateHistogramData", () => {
+  describe("edge cases", () => {
+    it("returns null for empty input", () => {
+      expect(calculateHistogramData([])).toBeNull();
+    });
+
+    it("handles single value", () => {
+      const result = calculateHistogramData([42]);
+      expect(result).not.toBeNull();
+      expect(result!.bins).toEqual([1]);
+      expect(result!.min).toBe(42);
+      expect(result!.max).toBe(42);
+      expect(result!.binSize).toBe(0);
+      expect(result!.p50).toBe(42);
+      expect(result!.p95).toBe(42);
+      expect(result!.p99).toBe(42);
+    });
+
+    it("handles all identical values", () => {
+      const result = calculateHistogramData([10, 10, 10, 10, 10]);
+      expect(result).not.toBeNull();
+      expect(result!.bins).toEqual([5]);
+      expect(result!.min).toBe(10);
+      expect(result!.max).toBe(10);
+      expect(result!.binSize).toBe(0);
+    });
+
+    it("handles two distinct values", () => {
+      const result = calculateHistogramData([10, 20]);
+      expect(result).not.toBeNull();
+      expect(result!.min).toBe(10);
+      expect(result!.max).toBe(20);
+      expect(result!.p50).toBe(15);
+      const binSum = result!.bins.reduce((s, b) => s + b, 0);
+      expect(binSum).toBe(2);
+    });
+  });
+
+  describe("bin distribution", () => {
+    it("bins sum to total number of values", () => {
+      const values = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      const binSum = result!.bins.reduce((s, b) => s + b, 0);
+      expect(binSum).toBe(values.length);
+    });
+
+    it("distributes uniform values across bins", () => {
+      const values = Array.from({ length: 100 }, (_, i) => i);
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.bins.length).toBe(20);
+      const binSum = result!.bins.reduce((s, b) => s + b, 0);
+      expect(binSum).toBe(100);
+    });
+
+    it("places values in correct bins", () => {
+      const values = [1, 1, 1, 5, 5, 9, 9, 9];
+      const result = calculateHistogramData(values, { minBins: 3, maxBins: 3, binScaling: 1 });
+      expect(result).not.toBeNull();
+      expect(result!.bins.length).toBe(3);
+      expect(result!.bins.reduce((s, b) => s + b, 0)).toBe(8);
+    });
+
+    it("handles skewed distribution", () => {
+      const values = [1, 1, 1, 1, 1, 100];
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      const binSum = result!.bins.reduce((s, b) => s + b, 0);
+      expect(binSum).toBe(6);
+    });
+  });
+
+  describe("percentiles", () => {
+    it("calculates correct p50 for even-length array", () => {
+      const values = [1, 2, 3, 4];
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.p50).toBe(2.5);
+    });
+
+    it("calculates correct p50 for odd-length array", () => {
+      const values = [1, 2, 3, 4, 5];
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.p50).toBe(3);
+    });
+
+    it("calculates p99 with interpolation", () => {
+      const values = Array.from({ length: 100 }, (_, i) => i + 1);
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.p99).toBeCloseTo(99.01);
+    });
+
+    it("calculates p95 with interpolation", () => {
+      const values = Array.from({ length: 100 }, (_, i) => i + 1);
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.p95).toBeCloseTo(95.05);
+    });
+
+    it("percentiles are monotonically increasing", () => {
+      const values = Array.from({ length: 200 }, () => Math.random() * 100);
+      const result = calculateHistogramData(values);
+      expect(result).not.toBeNull();
+      expect(result!.p50).toBeLessThanOrEqual(result!.p95);
+      expect(result!.p95).toBeLessThanOrEqual(result!.p99);
+    });
+  });
+
+  describe("bin count adaptation", () => {
+    it("uses minimum bins for small datasets", () => {
+      const values = Array.from({ length: 20 }, (_, i) => i);
+      const result = calculateHistogramData(values);
+      expect(result!.bins.length).toBe(10);
+    });
+
+    it("scales bins with dataset size", () => {
+      const values = Array.from({ length: 100 }, (_, i) => i);
+      const result = calculateHistogramData(values);
+      expect(result!.bins.length).toBe(20);
+    });
+
+    it("caps bins at maximum", () => {
+      const values = Array.from({ length: 200 }, (_, i) => i);
+      const result = calculateHistogramData(values);
+      expect(result!.bins.length).toBe(30);
+    });
+
+    it("respects custom options", () => {
+      const values = Array.from({ length: 100 }, (_, i) => i);
+      const result = calculateHistogramData(values, { minBins: 5, maxBins: 10, binScaling: 2 });
+      expect(result!.bins.length).toBe(10);
+    });
+  });
+
+  describe("min and max", () => {
+    it("correctly identifies min and max", () => {
+      const values = [5, 3, 8, 1, 9, 2];
+      const result = calculateHistogramData(values);
+      expect(result!.min).toBe(1);
+      expect(result!.max).toBe(9);
+    });
+
+    it("handles negative values", () => {
+      const values = [-10, -5, 0, 5, 10];
+      const result = calculateHistogramData(values);
+      expect(result!.min).toBe(-10);
+      expect(result!.max).toBe(10);
+    });
+
+    it("handles floating point values", () => {
+      const values = [1.5, 2.7, 3.14, 0.5, 4.99];
+      const result = calculateHistogramData(values);
+      expect(result!.min).toBe(0.5);
+      expect(result!.max).toBe(4.99);
+    });
+  });
+});

ui-svelte/src/lib/histogram.ts

@@ -0,0 +1,72 @@
+import type { HistogramData } from "./types";
+
+export interface HistogramOptions {
+  minBins?: number;
+  maxBins?: number;
+  binScaling?: number;
+}
+
+const DEFAULT_OPTIONS: HistogramOptions = {
+  minBins: 10,
+  maxBins: 30,
+  binScaling: 5,
+};
+
+function percentile(sorted: number[], p: number): number {
+  if (sorted.length === 0) return 0;
+  if (sorted.length === 1) return sorted[0];
+
+  const rank = (p / 100) * (sorted.length - 1);
+  const lower = Math.floor(rank);
+  const upper = Math.ceil(rank);
+  const fraction = rank - lower;
+
+  return sorted[lower] + fraction * (sorted[upper] - sorted[lower]);
+}
+
+export function calculateHistogramData(
+  values: number[],
+  options: HistogramOptions = DEFAULT_OPTIONS,
+): HistogramData | null {
+  if (values.length === 0) return null;
+
+  const sorted = [...values].sort((a, b) => a - b);
+  const min = sorted[0];
+  const max = sorted[sorted.length - 1];
+
+  const p50 = percentile(sorted, 50);
+  const p95 = percentile(sorted, 95);
+  const p99 = percentile(sorted, 99);
+
+  if (min === max) {
+    return {
+      bins: [values.length],
+      min,
+      max,
+      binSize: 0,
+      p50,
+      p95,
+      p99,
+    };
+  }
+
+  const { minBins = 10, maxBins = 30, binScaling = 5 } = options;
+  const binCount = Math.min(maxBins, Math.max(minBins, Math.floor(values.length / binScaling)));
+  const binSize = (max - min) / binCount;
+
+  const bins = new Array(binCount).fill(0);
+  for (const value of values) {
+    const binIndex = Math.min(Math.floor((value - min) / binSize), binCount - 1);
+    bins[binIndex]++;
+  }
+
+  return {
+    bins,
+    min,
+    max,
+    binSize,
+    p50,
+    p95,
+    p99,
+  };
+}

ui-svelte/src/lib/types.ts

@@ -48,6 +48,16 @@ export interface APIEventEnvelope {
   data: string;
 }
 
+export interface HistogramData {
+  bins: number[];
+  min: number;
+  max: number;
+  binSize: number;
+  p99: number;
+  p95: number;
+  p50: number;
+}
+
 export interface VersionInfo {
   build_date: string;
   commit: string;

ui-svelte/src/routes/Activity.svelte

@@ -1,5 +1,6 @@
 <script lang="ts">
   import { metrics, getCapture } from "../stores/api";
+  import ActivityStats from "../components/ActivityStats.svelte";
   import Tooltip from "../components/Tooltip.svelte";
   import CaptureDialog from "../components/CaptureDialog.svelte";
   import type { ReqRespCapture } from "../lib/types";
@@ -63,33 +64,38 @@
 
 <div class="p-2">
   <h1 class="text-2xl font-bold">Activity</h1>
+  <div class="mt-4 mb-4">
+    <ActivityStats />
+  </div>
 
-  {#if $metrics.length === 0}
-    <div class="text-center py-8">
-      <p class="text-gray-600">No metrics data available</p>
-    </div>
-  {:else}
-    <div class="card overflow-auto">
-      <table class="min-w-full divide-y">
-        <thead class="border-gray-200 dark:border-white/10">
-          <tr class="text-left text-xs uppercase tracking-wider">
-            <th class="px-6 py-3">ID</th>
-            <th class="px-6 py-3">Time</th>
-            <th class="px-6 py-3">Model</th>
-            <th class="px-6 py-3">
-              Cached <Tooltip content="prompt tokens from cache" />
-            </th>
-            <th class="px-6 py-3">
-              Prompt <Tooltip content="new prompt tokens processed" />
-            </th>
-            <th class="px-6 py-3">Generated</th>
-            <th class="px-6 py-3">Prompt Processing</th>
-            <th class="px-6 py-3">Generation Speed</th>
-            <th class="px-6 py-3">Duration</th>
-            <th class="px-6 py-3">Capture</th>
+  <div class="card overflow-auto">
+    <table class="min-w-full divide-y">
+      <thead class="border-gray-200 dark:border-white/10">
+        <tr class="text-left text-xs uppercase tracking-wider">
+          <th class="px-6 py-3">ID</th>
+          <th class="px-6 py-3">Time</th>
+          <th class="px-6 py-3">Model</th>
+          <th class="px-6 py-3">
+            Cached <Tooltip content="prompt tokens from cache" />
+          </th>
+          <th class="px-6 py-3">
+            Prompt <Tooltip content="new prompt tokens processed" />
+          </th>
+          <th class="px-6 py-3">Generated</th>
+          <th class="px-6 py-3">Prompt Processing</th>
+          <th class="px-6 py-3">Generation Speed</th>
+          <th class="px-6 py-3">Duration</th>
+          <th class="px-6 py-3">Capture</th>
+        </tr>
+      </thead>
+      <tbody class="divide-y">
+        {#if sortedMetrics.length === 0}
+          <tr>
+            <td colspan="10" class="px-6 py-8 text-center text-sm text-gray-500 dark:text-gray-400">
+              No activity recorded
+            </td>
           </tr>
-        </thead>
-        <tbody class="divide-y">
+        {:else}
           {#each sortedMetrics as metric (metric.id)}
             <tr class="whitespace-nowrap text-sm border-gray-200 dark:border-white/10">
               <td class="px-4 py-4">{metric.id + 1}</td>
@@ -116,10 +122,10 @@
               </td>
             </tr>
           {/each}
-        </tbody>
-      </table>
-    </div>
-  {/if}
+        {/if}
+      </tbody>
+    </table>
+  </div>
 </div>
 
 <CaptureDialog capture={selectedCapture} open={dialogOpen} onclose={closeDialog} />

ui-svelte/src/routes/Models.svelte

@@ -2,7 +2,6 @@
   import { isNarrow } from "../stores/theme";
   import { upstreamLogs } from "../stores/api";
   import ModelsPanel from "../components/ModelsPanel.svelte";
-  import StatsPanel from "../components/StatsPanel.svelte";
   import LogPanel from "../components/LogPanel.svelte";
   import ResizablePanels from "../components/ResizablePanels.svelte";
 
@@ -14,13 +13,6 @@
     <ModelsPanel />
   {/snippet}
   {#snippet rightPanel()}
-    <div class="flex flex-col h-full space-y-4">
-      {#if direction === "horizontal"}
-        <StatsPanel />
-      {/if}
-      <div class="flex-1 min-h-0">
-        <LogPanel id="modelsupstream" title="Upstream Logs" logData={$upstreamLogs} />
-      </div>
-    </div>
+    <LogPanel id="modelsupstream" title="Upstream Logs" logData={$upstreamLogs} />
   {/snippet}
 </ResizablePanels>