feat: add runner failover rules and cache affinity

apps/api/internal/runner/cache_affinity_test.go

@@ -0,0 +1,42 @@
+package runner
+
+import "testing"
+
+func TestBuildCacheAffinityKeyUsesExplicitKeyFirst(t *testing.T) {
+	body := map[string]any{
+		"cacheAffinityKey": "session-a",
+		"sessionId":        "session-b",
+		"messages":         []any{map[string]any{"role": "user", "content": "hello"}},
+	}
+
+	got := buildCacheAffinityKey("chat.completions", "text_generate", body)
+	if got == "" || got[:9] != "explicit:" {
+		t.Fatalf("expected explicit cache affinity key, got %q", got)
+	}
+}
+
+func TestBuildCacheAffinityKeyUsesSessionContinuity(t *testing.T) {
+	got := buildCacheAffinityKey("chat.completions", "text_generate", map[string]any{
+		"session_id": "conversation-1",
+	})
+
+	if got == "" || got[:11] != "session_id:" {
+		t.Fatalf("expected session cache affinity key, got %q", got)
+	}
+}
+
+func TestBuildCacheAffinityKeyFallsBackToPromptHash(t *testing.T) {
+	body := map[string]any{
+		"messages": []any{
+			map[string]any{"role": "system", "content": "stable"},
+			map[string]any{"role": "user", "content": "hello"},
+		},
+		"tools": []any{map[string]any{"type": "function", "function": map[string]any{"name": "lookup"}}},
+	}
+
+	first := buildCacheAffinityKey("chat.completions", "text_generate", body)
+	second := buildCacheAffinityKey("chat.completions", "text_generate", body)
+	if first == "" || first != second || first[:7] != "prompt:" {
+		t.Fatalf("expected stable prompt hash cache affinity key, first=%q second=%q", first, second)
+	}
+}