Improve config, diagnosis, and README demo

README.md

@@ -15,7 +15,7 @@ The live indicator and CPU activity area chart are animated with [termkit-go](ht
 
 Prometheus and load-test commands are optional integrations when you need application metrics or want to correlate a workload with system health.
 
-When Docker is available, the dashboard also shows live container CPU, memory, and network data. An optional `observe.yaml` can set `refresh_interval` (milliseconds), `thresholds` (`cpu`, `memory`, `swap`, `disk`, and `io_wait`), and visible `panels` while the defaults remain zero-config.
+When Docker is available, the dashboard also shows live container CPU, memory, and network data. An optional `observe.yaml` can set `refresh_interval` (milliseconds), `theme` (`default`, `dark`, `light`, or `mono`), `thresholds` (`cpu`, `memory`, `swap`, `disk`, and `io_wait`), visible `panels`, and custom `prometheus_presets` while the defaults remain zero-config. See [observe.yaml.example](observe.yaml.example).
 
 ## Install
 

cmd/observe/root.go

@@ -3,10 +3,12 @@ package observe
 import (
 	"fmt"
 	"os"
+	"time"
 
 	"github.com/pol-cova/observe/internal/assistant"
 	"github.com/pol-cova/observe/internal/config"
 	"github.com/pol-cova/observe/internal/detect"
+	"github.com/pol-cova/observe/internal/metrics/local"
 	"github.com/pol-cova/observe/internal/prometheus"
 	"github.com/pol-cova/observe/internal/snapshot"
 	"github.com/pol-cova/observe/internal/tui"
@@ -53,11 +55,15 @@ var initCmd = &cobra.Command{
 var askCmd = &cobra.Command{
 	Use: "ask <question>", Short: "Explain current system health in plain English", Args: cobra.MinimumNArgs(1),
 	RunE: func(cmd *cobra.Command, args []string) error {
-		snapshot, err := tui.Snapshot()
+		cfg, err := config.Load(configPath)
+		if err != nil {
+			return err
+		}
+		history, err := local.SampleHistory(3, 500*time.Millisecond)
 		if err != nil {
 			return err
 		}
-		fmt.Print(assistant.Answer(args, snapshot))
+		fmt.Print(assistant.Answer(args, history, cfg.Thresholds.Health()))
 		return nil
 	},
 }

docs/demo.tape

@@ -1,19 +1,47 @@
 Output docs/observe.gif
 
-Set FontSize 18
-Set Width 1440
-Set Height 820
-Set Margin 24
-Set MarginFill "#0b1118"
-Set Theme "Dracula"
-Set TypingSpeed 14ms
-Set PlaybackSpeed 1.15
+Set Shell bash
+Set FontSize 20
+Set Width 1280
+Set Height 760
+Set Margin 20
+Set MarginFill "#0a0e14"
+Set Theme "Catppuccin Mocha"
+Set TypingSpeed 12ms
+Set PlaybackSpeed 1
 
 Hide
-Type "clear"
-Enter
-Sleep 300ms
-Type "cat docs/demo-output.txt"
+Type "go build -o /tmp/observe-demo ."
 Enter
+Sleep 4s
 Show
-Sleep 5s
+
+Type "OBSERVE_NO_ALT_SCREEN=1 /tmp/observe-demo"
+Enter
+Sleep 4s
+Type "2"
+Sleep 2.5s
+Type "3"
+Sleep 2.5s
+Type "5"
+Sleep 2.5s
+Type "1"
+Sleep 3s
+Type "j"
+Sleep 400ms
+Type "j"
+Sleep 400ms
+Type "Enter"
+Sleep 2.5s
+Type "Esc"
+Sleep 800ms
+Type "?"
+Sleep 2s
+Type "?"
+Sleep 800ms
+Type "q"
+Enter
+Sleep 500ms
+Type "/tmp/observe-demo ask is my server cpu bound"
+Enter
+Sleep 2.5s

internal/assistant/assistant.go

@@ -2,32 +2,80 @@ package assistant
 
 import (
 	"fmt"
-	"github.com/pol-cova/observe/internal/metrics/local"
+	"sort"
 	"strings"
+
+	"github.com/pol-cova/observe/internal/health"
+	"github.com/pol-cova/observe/internal/metrics/local"
 )
 
-func Answer(parts []string, s local.Snapshot) string {
+func Answer(parts []string, history []local.Snapshot, thresholds health.Thresholds) string {
+	if len(history) == 0 {
+		return "No samples collected yet.\n"
+	}
+	latest := history[len(history)-1]
 	q := strings.ToLower(strings.Join(parts, " "))
+
 	var b strings.Builder
+	fmt.Fprintf(&b, "%s\n\n", titleFor(q))
+	fmt.Fprintf(&b, "CPU %.1f%%, memory %.1f%%, disk %.1f%%.\n", latest.CPU, latest.Memory, latest.Disk)
+	if len(history) > 1 {
+		first := history[0]
+		fmt.Fprintf(&b, "Trend (%d samples): CPU %+.1f, memory %+.1f.\n",
+			len(history), latest.CPU-first.CPU, latest.Memory-first.Memory)
+	}
+	for _, process := range topProcesses(latest.Processes, 3) {
+		fmt.Fprintf(&b, "  • %s (pid %d) — %.1f%% CPU\n", process.Name, process.PID, process.CPU)
+	}
+	for _, hint := range health.HintsWithThresholds(history, thresholds) {
+		fmt.Fprintf(&b, "  • %s\n", hint)
+	}
+	fmt.Fprintf(&b, "\n%s\n", adviceFor(q, latest))
+	return b.String()
+}
+
+func titleFor(q string) string {
 	switch {
-	case strings.Contains(q, "prometheus") || strings.Contains(q, "5xx") || strings.Contains(q, "query"):
-		b.WriteString("For 5xx errors, use:\n\n  sum(rate(http_requests_total{status=~\"5..\"}[1m]))\n\n")
+	case strings.Contains(q, "prometheus") || strings.Contains(q, "5xx"):
+		return "Prometheus diagnosis"
 	case strings.Contains(q, "latency"):
-		b.WriteString("Latency diagnosis\n\n")
+		return "Latency diagnosis"
+	case strings.Contains(q, "memory") || strings.Contains(q, "ram"):
+		return "Memory diagnosis"
+	case strings.Contains(q, "disk"):
+		return "Disk diagnosis"
+	case strings.Contains(q, "network"):
+		return "Network diagnosis"
 	case strings.Contains(q, "cpu"):
-		b.WriteString("CPU diagnosis\n\n")
+		return "CPU diagnosis"
 	default:
-		b.WriteString("Current system health\n\n")
-	}
-	fmt.Fprintf(&b, "CPU is %.1f%%, memory is %.1f%%, and disk is %.1f%%.\n", s.CPU, s.Memory, s.Disk)
-	if s.CPU >= 90 {
-		b.WriteString("\nThis machine appears CPU-bound. Profile the top process or reduce test concurrency.\n")
-	} else if s.Memory >= 85 {
-		b.WriteString("\nMemory pressure is the strongest signal. Check process growth and swapping.\n")
-	} else if s.Disk >= 90 {
-		b.WriteString("\nDisk capacity is critically high; free space before trusting results.\n")
-	} else {
-		b.WriteString("\nNo clear local resource bottleneck is present in this single sample. For latency, compare p95 with CPU and error rate over time.\n")
+		return "Current system health"
 	}
-	return b.String()
+}
+
+func adviceFor(q string, s local.Snapshot) string {
+	switch {
+	case strings.Contains(q, "prometheus") || strings.Contains(q, "5xx"):
+		return "For 5xx errors, use:\n\n  sum(rate(http_requests_total{status=~\"5..\"}[1m]))"
+	case strings.Contains(q, "cpu") && s.CPU >= 90:
+		return "This machine appears CPU-bound. Profile the top process or reduce concurrency."
+	case strings.Contains(q, "memory") && (s.Memory >= 85 || s.Swap >= 25):
+		return "Memory pressure is the strongest signal. Check process growth and swapping."
+	case strings.Contains(q, "disk") && s.Disk >= 90:
+		return "Disk capacity is critically high; free space before trusting results."
+	default:
+		return "No clear local bottleneck in these samples. Compare metrics over time or run with --prometheus."
+	}
+}
+
+func topProcesses(processes []local.Process, limit int) []local.Process {
+	if limit <= 0 || len(processes) == 0 {
+		return nil
+	}
+	ranked := append([]local.Process(nil), processes...)
+	sort.Slice(ranked, func(i, j int) bool { return ranked[i].CPU > ranked[j].CPU })
+	if len(ranked) > limit {
+		return ranked[:limit]
+	}
+	return ranked
 }

internal/assistant/assistant_test.go

@@ -0,0 +1,36 @@
+package assistant
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/pol-cova/observe/internal/health"
+	"github.com/pol-cova/observe/internal/metrics/local"
+)
+
+func TestAnswerIncludesTrendAndHints(t *testing.T) {
+	history := []local.Snapshot{
+		{CPU: 40, Memory: 50, Disk: 60, Processes: []local.Process{{PID: 1, Name: "worker", CPU: 12}}},
+		{CPU: 55, Memory: 62, Disk: 61, Processes: []local.Process{{PID: 1, Name: "worker", CPU: 18}}},
+	}
+	answer := Answer([]string{"cpu"}, history, health.Thresholds{})
+	for _, phrase := range []string{"Trend (2 samples)", "worker", "No immediate resource bottleneck detected"} {
+		if !strings.Contains(answer, phrase) {
+			t.Fatalf("answer missing %q:\n%s", phrase, answer)
+		}
+	}
+}
+
+func TestAnswerCPUAdviceWhenSaturated(t *testing.T) {
+	answer := Answer([]string{"cpu"}, []local.Snapshot{{CPU: 95, Memory: 40, Disk: 50}}, health.Thresholds{})
+	if !strings.Contains(answer, "CPU-bound") {
+		t.Fatalf("expected CPU-bound advice: %s", answer)
+	}
+}
+
+func TestAnswerUsesThresholdHints(t *testing.T) {
+	answer := Answer([]string{"health"}, []local.Snapshot{{CPU: 96, Memory: 40, Disk: 50}}, health.Thresholds{CPU: 95})
+	if !strings.Contains(answer, "CPU saturated") {
+		t.Fatalf("expected threshold hint: %s", answer)
+	}
+}

internal/config/config.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"os"
 
+	"github.com/pol-cova/observe/internal/health"
 	"gopkg.in/yaml.v3"
 )
 
@@ -23,6 +24,10 @@ type Thresholds struct {
 	IOWait float64 `yaml:"io_wait"`
 }
 
+func (t Thresholds) Health() health.Thresholds {
+	return health.Thresholds{CPU: t.CPU, Memory: t.Memory, Swap: t.Swap, Disk: t.Disk, IOWait: t.IOWait}
+}
+
 func Load(path string) (Config, error) {
 	data, err := os.ReadFile(path)
 	if errors.Is(err, os.ErrNotExist) {

internal/config/config_test.go

@@ -8,16 +8,19 @@ import (
 
 func TestLoadConfig(t *testing.T) {
 	path := filepath.Join(t.TempDir(), "observe.yaml")
-	if err := os.WriteFile(path, []byte("refresh_interval: 750\nthresholds:\n  cpu: 80\npanels: [overview, docker]\n"), 0600); err != nil {
+	if err := os.WriteFile(path, []byte("refresh_interval: 750\ntheme: light\nthresholds:\n  cpu: 80\npanels: [overview, docker]\nprometheus_presets:\n  - name: Queue depth\n    query: queue_depth\n"), 0600); err != nil {
 		t.Fatal(err)
 	}
 	cfg, err := Load(path)
 	if err != nil {
 		t.Fatal(err)
 	}
-	if cfg.RefreshInterval != 750 || cfg.Thresholds.CPU != 80 || len(cfg.Panels) != 2 {
+	if cfg.RefreshInterval != 750 || cfg.Thresholds.CPU != 80 || len(cfg.Panels) != 2 || cfg.Theme != "light" {
 		t.Fatalf("unexpected config: %#v", cfg)
 	}
+	if len(cfg.Prometheus) != 1 || cfg.Prometheus[0].Name != "Queue depth" {
+		t.Fatalf("unexpected prometheus presets: %#v", cfg.Prometheus)
+	}
 }
 
 func TestMissingConfigUsesDefaults(t *testing.T) {

internal/metrics/local/sample.go

@@ -0,0 +1,22 @@
+package local
+
+import "time"
+
+func SampleHistory(samples int, interval time.Duration) ([]Snapshot, error) {
+	if samples < 1 {
+		samples = 1
+	}
+	collector := New()
+	history := make([]Snapshot, 0, samples)
+	for i := 0; i < samples; i++ {
+		snapshot, err := collector.Collect()
+		if err != nil {
+			return history, err
+		}
+		history = append(history, snapshot)
+		if i < samples-1 {
+			time.Sleep(interval)
+		}
+	}
+	return history, nil
+}

internal/prometheus/presets.go

@@ -0,0 +1,29 @@
+package prometheus
+
+import "github.com/pol-cova/observe/internal/config"
+
+func MergePresets(custom []config.PrometheusPreset) []Preset {
+	if len(custom) == 0 {
+		return append([]Preset(nil), Presets...)
+	}
+	byName := make(map[string]Preset, len(Presets)+len(custom))
+	order := make([]string, 0, len(Presets)+len(custom))
+	for _, preset := range Presets {
+		byName[preset.Name] = preset
+		order = append(order, preset.Name)
+	}
+	for _, preset := range custom {
+		if preset.Name == "" || preset.Query == "" {
+			continue
+		}
+		if _, ok := byName[preset.Name]; !ok {
+			order = append(order, preset.Name)
+		}
+		byName[preset.Name] = Preset{Name: preset.Name, Description: preset.Description, Query: preset.Query}
+	}
+	merged := make([]Preset, len(order))
+	for i, name := range order {
+		merged[i] = byName[name]
+	}
+	return merged
+}

internal/prometheus/presets_test.go

@@ -0,0 +1,37 @@
+package prometheus
+
+import (
+	"testing"
+
+	"github.com/pol-cova/observe/internal/config"
+)
+
+func TestMergePresetsUsesBuiltinsWhenEmpty(t *testing.T) {
+	merged := MergePresets(nil)
+	if len(merged) != len(Presets) {
+		t.Fatalf("len(MergePresets(nil)) = %d, want %d", len(merged), len(Presets))
+	}
+}
+
+func TestMergePresetsOverridesAndAddsCustom(t *testing.T) {
+	custom := []config.PrometheusPreset{
+		{Name: "5xx errors", Description: "custom 5xx", Query: "sum(rate(errors_total[1m]))"},
+		{Name: "Queue depth", Description: "pending jobs", Query: "queue_depth"},
+	}
+	merged := MergePresets(custom)
+	foundOverride, foundCustom := false, false
+	for _, preset := range merged {
+		if preset.Name == "5xx errors" {
+			foundOverride = true
+			if preset.Query != "sum(rate(errors_total[1m]))" {
+				t.Fatalf("override query = %q", preset.Query)
+			}
+		}
+		if preset.Name == "Queue depth" {
+			foundCustom = true
+		}
+	}
+	if !foundOverride || !foundCustom {
+		t.Fatalf("merged presets missing override/custom: %#v", merged)
+	}
+}