${t('recFilter')}
@@ -690,15 +696,16 @@ async function renderStatus() {
await paintStats();
statsTimerId = setInterval(paintStats, 3000);
}
-function renderKeySelect(keys) {
+function renderKeySelect(keys, keyNames) {
const sel = $('#rec-key');
if (!sel) return;
const cur = sel.value;
+ const nm = (k) => (keyNames && keyNames[k] ? keyNames[k] + ' · ' + k : k);
sel.innerHTML = `
` +
- keys.map(k => `
`).join('');
+ keys.map(k => `
`).join('');
if (cur) sel.value = cur;
}
-function renderKeyF(v) { statsKeyF = v; paintStats(); }
+function renderKeyF(v) { statsKeyF = (statsKeyF === v) ? '' : v; paintStats(); }
function openExportModal() {
const now = new Date();
const ago = d => { const x = new Date(now); x.setDate(x.getDate() - d); return x.toISOString().slice(0, 10); };
@@ -731,14 +738,14 @@ function downloadKeysCsv() {
}
async function paintStats() {
try {
- const q = '/api/stats?limit=500' + (statsKeyF ? '&key=' + encodeURIComponent(statsKeyF) : '');
+ const q = '/api/stats?limit=2000' + (statsKeyF ? '&key=' + encodeURIComponent(statsKeyF) : '');
const st = await api(q);
const tot = st.total || {};
const okr = tot.reqs ? Math.round(tot.ok * 100 / tot.reqs) : 0;
const avg = fmtLat(tot.latency_sum_ms, tot.reqs);
const kpi = $('#kpi-row');
if (kpi) kpi.innerHTML = `
-
${t('kpiActive')}
${st.active || 0}
${statsKeyF ? esc(statsKeyF) : ''}
+
${t('kpiActive')}
${st.active || 0}
${statsKeyF ? esc(st.key_names ? (st.key_names[statsKeyF] ? st.key_names[statsKeyF] + ' · ' : '') + statsKeyF : statsKeyF) : ''}
${t('kpiReqs')}
${fmtN(tot.reqs)}
${t('kpiOk')} ${okr}%
${t('kpiTokens')}
${fmtTok(tot.tokens)}
${t('thPrompt')} ${fmtTok(tot.prompt_tokens)} · ${t('thCompl')} ${fmtTok(tot.completion_tokens)}
@@ -748,7 +755,9 @@ async function paintStats() {
paintStatusTable(st.by_status || []);
paintKeyTable(st.by_key || [], st.key_names || {});
paintRecords(st.records || [], st.key_names || {});
- renderKeySelect((st.by_key || []).map(k => k.name));
+ const rex = $('#rec-exit');
+ if (rex) rex.innerHTML = statsKeyF ? `
` : '';
+ renderKeySelect((st.by_key || []).map(k => k.name), st.key_names || {});
} catch (e) { console.error('[stats]', e); }
}
function paintModelTable(rows) {
diff --git a/internal/scheduler/scheduler.go b/internal/scheduler/scheduler.go
index 4dc4853..789095f 100644
--- a/internal/scheduler/scheduler.go
+++ b/internal/scheduler/scheduler.go
@@ -92,7 +92,7 @@ func (tn *TierNode) NextStart() int64 {
// requests (rotation state resets when the chain is rebuilt, e.g. after
// editing the rules — acceptable, the swap also resets cooldowns).
type Chain struct {
- Tiers []*TierNode // descending tier order
+ Tiers []*TierNode // ascending tier order (tier 1 = highest priority, tried first)
}
// TierErrors is the per-tier failure summary carried by ChainErr. Errors
@@ -133,10 +133,10 @@ func (e *ChainErr) Error() string {
return b.String()
}
-// BuildChain groups rules into descending tiers and resolves each slot's
-// provider via prov. Rules whose provider resolves to nil are dropped (the
-// source no longer serves the model). Slot order within a tier follows the
-// configured rule order.
+// BuildChain groups rules into ascending tiers (tier 1 = highest priority,
+// tried first) and resolves each slot's provider via prov. Rules whose
+// provider resolves to nil are dropped (the source no longer serves the
+// model). Slot order within a tier follows the configured rule order.
func BuildChain(rules []Rule, prov func(model, source string) Provider) *Chain {
byTier := map[int][]*Slot{}
var tiers []int
@@ -157,7 +157,7 @@ func BuildChain(rules []Rule, prov func(model, source string) Provider) *Chain {
Prov: p,
})
}
- sort.Slice(tiers, func(i, j int) bool { return tiers[i] > tiers[j] })
+ sort.Slice(tiers, func(i, j int) bool { return tiers[i] < tiers[j] })
ch := &Chain{}
for _, t := range tiers {
tn := &TierNode{Tier: t, Slots: byTier[t]}
@@ -221,12 +221,12 @@ func runTier(ctx context.Context, tn *TierNode, cands []*Slot, base int64, req *
return tierResult{hard: hard}
}
-// chainDrive runs a request down the chain (plan 2.3): tiers descending,
-// per-tier round-robin starting at the tier cursor, same-tier runs ordered by
-// preference (negative prefs sink but stay reachable). Quota-exhausted and
-// cooling slots are filtered up front; a fully busy tier is polled for a
-// bounded time before falling through. Failures are summarized in *ChainErr
-// for the caller to map to HTTP 503.
+// chainDrive runs a request down the chain (plan 2.3): tiers ascending (tier
+// 1, the highest priority, first), per-tier round-robin starting at the tier
+// cursor, same-tier runs ordered by preference (negative prefs sink but stay
+// reachable). Quota-exhausted and cooling slots are filtered up front; a
+// fully busy tier is polled for a bounded time before falling through.
+// Failures are summarized in *ChainErr for the caller to map to HTTP 503.
func (s *Scheduler) chainDrive(ctx context.Context, chain *Chain, req *types.ChatRequest, exhausted func(*Slot) bool, stream bool) (*types.UnifiedResponse, <-chan types.UnifiedChunk, string, string, error) {
if chain == nil || len(chain.Tiers) == 0 {
return nil, nil, "", "", fmt.Errorf("no auto slot configured")
diff --git a/internal/scheduler/scheduler_test.go b/internal/scheduler/scheduler_test.go
index a53ae74..3fab878 100644
--- a/internal/scheduler/scheduler_test.go
+++ b/internal/scheduler/scheduler_test.go
@@ -97,12 +97,12 @@ func TestBuildChainTierOrdering(t *testing.T) {
if len(ch.Tiers) != 3 {
t.Fatalf("tiers = %d, want 3", len(ch.Tiers))
}
- if ch.Tiers[0].Tier != 3 || ch.Tiers[1].Tier != 2 || ch.Tiers[2].Tier != 1 {
- t.Fatalf("tier order = %d,%d,%d, want 3,2,1", ch.Tiers[0].Tier, ch.Tiers[1].Tier, ch.Tiers[2].Tier)
+ if ch.Tiers[0].Tier != 1 || ch.Tiers[1].Tier != 2 || ch.Tiers[2].Tier != 3 {
+ t.Fatalf("tier order = %d,%d,%d, want 1,2,3", ch.Tiers[0].Tier, ch.Tiers[1].Tier, ch.Tiers[2].Tier)
}
// same-tier slots keep configured order, unresolvable rule is dropped
- if len(ch.Tiers[2].Slots) != 2 || ch.Tiers[2].Slots[0].Model != "a" || ch.Tiers[2].Slots[1].Model != "d" {
- t.Fatalf("tier 1 slots = %+v", ch.Tiers[2].Slots)
+ if len(ch.Tiers[0].Slots) != 2 || ch.Tiers[0].Slots[0].Model != "a" || ch.Tiers[0].Slots[1].Model != "d" {
+ t.Fatalf("tier 1 slots = %+v", ch.Tiers[0].Slots)
}
}
@@ -182,9 +182,9 @@ func TestChainAllBusyBoundedWaitThenNextTier(t *testing.T) {
a.busy.Store(true)
b.busy.Store(true)
ch := BuildChain([]Rule{
- {Tier: 5, Model: "a", Source: "s1"},
- {Tier: 5, Model: "b", Source: "s2"},
- {Tier: 4, Model: "c", Source: "s3"},
+ {Tier: 1, Model: "a", Source: "s1"},
+ {Tier: 1, Model: "b", Source: "s2"},
+ {Tier: 2, Model: "c", Source: "s3"},
}, bySource(a, b, c))
s := New(0)
t0 := time.Now()