quota: z.ai credit-bucket back-pressure + resource gate + usage accounting (Redmine 490+491)

The loop now consults quota and host state before every dispatch and
DEFERS gated work with a logged reason instead of letting turns die at
the provider (the 2026-08-28 19:00 quota-wall failure mode, replayed as
a test). Adds internal/quota: 5h/weekly credit buckets (provider poll
when z.ai ships an endpoint - fake-server tested - else locally
estimated from the documented credit formula), TZ-aware peak window
(default 01:00-05:00 America/Chicago weekdays, matching the documented
z.ai peak Mon-Fri 14:00-18:00 Singapore), block/defer thresholds, a
read-only load/mem/disk/IO-PSI monitor, an optional redis shared-state
hop (stdlib RESP2 mini-client) so all instances of an account
coordinate, per-class token+credit accounting in loop.jsonl, and
`harness quota status|probe|gate`. Config: [quota] + [resources]
sections; README runbook covers the redis container and deploy-time
cgroup enforcement.

💘 Generated with Crush

Assisted-by: Crush:glm-5.2

internal/config/quota_test.go

@@ -0,0 +1,131 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestQuotaAndResourcesLoad(t *testing.T) {
+	src := `
+vertical = "t"
+[loop]
+[redmine]
+url = "https://rm"
+key_ref = "env:K"
+scope_query = "p=1"
+[litellm]
+base_url = "http://l"
+key_ref = "env:K"
+[models]
+a = "m"
+default_tier = "a"
+[models.classes]
+x = "a"
+[quota]
+enabled = true
+account = "zai-max-1"
+plan_5h_credits = 28000
+plan_weekly_credits = 140000
+usage_url = "https://api.z.ai/api/coding/paas/v4/usage"
+key_ref = "env:HARNESS_ZAI_KEY"
+poll_interval_secs = 120
+defer_at_pct = 80
+block_at_pct = 92.5
+peak_start = "22:30"
+peak_end = "04:15"
+timezone = "America/Chicago"
+peak_weekdays_only = false
+peak_classes = ["study", "read", "summarize"]
+redis_url = "redis://192.168.3.78:6390/2"
+[resources]
+enabled = true
+max_load_avg = 5.5
+min_mem_available_mb = 1024
+min_disk_free_mb = 2048
+max_io_delay_pct = 75
+`
+	path := filepath.Join(t.TempDir(), "harness.toml")
+	if err := os.WriteFile(path, []byte(src), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	cfg, err := Load(path)
+	if err != nil {
+		t.Fatalf("Load: %v", err)
+	}
+	q := cfg.Quota
+	if !q.Enabled || q.Account != "zai-max-1" || q.Plan5hCredits != 28000 || q.PlanWeeklyCredits != 140000 {
+		t.Errorf("quota basics: %+v", q)
+	}
+	if q.UsageURL == "" || q.KeyRef != "env:HARNESS_ZAI_KEY" || q.PollIntervalSecs != 120 {
+		t.Errorf("poller config: %+v", q)
+	}
+	if q.DeferAtPct != 80 || q.BlockAtPct != 92.5 {
+		t.Errorf("thresholds: %+v", q)
+	}
+	if q.PeakStart != "22:30" || q.PeakEnd != "04:15" || q.Timezone != "America/Chicago" || q.PeakWeekdaysOnly {
+		t.Errorf("peak window: %+v", q)
+	}
+	if len(q.PeakClasses) != 3 || q.PeakClasses[0] != "study" {
+		t.Errorf("peak classes: %v", q.PeakClasses)
+	}
+	if q.RedisURL != "redis://192.168.3.78:6390/2" {
+		t.Errorf("redis url: %q", q.RedisURL)
+	}
+	r := cfg.Resources
+	if !r.Enabled || r.MaxLoadAvg != 5.5 || r.MinMemAvailableMB != 1024 || r.MinDiskFreeMB != 2048 || r.MaxIODelayPct != 75 {
+		t.Errorf("resources: %+v", r)
+	}
+}
+
+func TestQuotaValidateErrors(t *testing.T) {
+	base := `
+vertical = "t"
+[litellm]
+base_url = "http://l"
+key_ref = "env:K"
+[models]
+a = "m"
+default_tier = "a"
+`
+	cases := []struct {
+		name, extra, want string
+	}{
+		{"no account", "[quota]\nenabled = true\n", "account is required"},
+		{"bad bucket", "[quota]\nenabled = true\naccount = \"a\"\nplan_5h_credits = 0\n", "plan_5h_credits"},
+		{"defer > block", "[quota]\nenabled = true\naccount = \"a\"\ndefer_at_pct = 99\nblock_at_pct = 95\n", "defer_at_pct"},
+		{"bad window", "[quota]\nenabled = true\naccount = \"a\"\npeak_start = \"99:00\"\n", "peak_start"},
+		{"bad tz", "[quota]\nenabled = true\naccount = \"a\"\ntimezone = \"Mars/Olympus\"\n", "timezone"},
+		{"usage url without key", "[quota]\nenabled = true\naccount = \"a\"\nusage_url = \"https://x\"\n", "key_ref is required"},
+		{"bad threshold", "[resources]\nenabled = true\nmax_load_avg = -1\n", "thresholds must be > 0"},
+	}
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			path := filepath.Join(t.TempDir(), "harness.toml")
+			if err := os.WriteFile(path, []byte(base+c.extra), 0o600); err != nil {
+				t.Fatal(err)
+			}
+			_, err := Load(path)
+			if err == nil || !strings.Contains(err.Error(), c.want) {
+				t.Fatalf("err = %v, want containing %q", err, c.want)
+			}
+		})
+	}
+}
+
+func TestFloatTomLValues(t *testing.T) {
+	doc, err := ParseTOML("a = 6\nb = 6.25\nc = 1_000.5\n")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if v, ok := doc.Float("a"); !ok || v != 6 {
+		t.Errorf("int as float = %v %v", v, ok)
+	}
+	if v, ok := doc.Float("b"); !ok || v != 6.25 {
+		t.Errorf("float = %v %v", v, ok)
+	}
+	if v, ok := doc.Float("c"); !ok || v != 1000.5 {
+		t.Errorf("underscore float = %v %v", v, ok)
+	}
+}