feat(config): add per-auth disable_cooling override support

- Introduced `disable_cooling` metadata field for fine-grained control over cooldown scheduling.
- Updated `Auth` object to include `Metadata` with conditional logic for handling empty states.
- Added YAML configuration support for `disable_cooling` in API key definitions across providers.
- Enhanced unit tests to validate `disable_cooling` behavior in various scenarios.

config.example.yaml

@@ -157,6 +157,7 @@ nonstream-keepalive-interval: 0
 # gemini-api-key:
 #   - api-key: "AIzaSy...01"
 #     prefix: "test" # optional: require calls like "test/gemini-3-pro-preview" to target this credential
+#     disable-cooling: false # optional: per-auth override for auth/model cooldown scheduling
 #     base-url: "https://generativelanguage.googleapis.com"
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -176,6 +177,7 @@ nonstream-keepalive-interval: 0
 # codex-api-key:
 #   - api-key: "sk-atSM..."
 #     prefix: "test" # optional: require calls like "test/gpt-5-codex" to target this credential
+#     disable-cooling: false # optional: per-auth override for auth/model cooldown scheduling
 #     base-url: "https://www.example.com" # use the custom codex API endpoint
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -195,6 +197,7 @@ nonstream-keepalive-interval: 0
 #   - api-key: "sk-atSM..." # use the official claude API key, no need to set the base url
 #   - api-key: "sk-atSM..."
 #     prefix: "test" # optional: require calls like "test/claude-sonnet-latest" to target this credential
+#     disable-cooling: false # optional: per-auth override for auth/model cooldown scheduling
 #     base-url: "https://www.example.com" # use the custom claude API endpoint
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -250,6 +253,7 @@ nonstream-keepalive-interval: 0
 #     disabled: false # optional: set to true to disable this provider without removing it
 #     prefix: "test" # optional: require calls like "test/kimi-k2" to target this provider's credentials
 #     base-url: "https://openrouter.ai/api/v1" # The base URL of the provider.
+#     disable-cooling: false # optional: per-provider override for auth/model cooldown scheduling
 #     headers:
 #       X-Custom-Header: "custom-value"
 #     api-key-entries:

internal/config/config.go

@@ -226,12 +226,6 @@ type RoutingConfig struct {
 	// Supported values: "round-robin" (default), "fill-first".
 	Strategy string `yaml:"strategy,omitempty" json:"strategy,omitempty"`
 
-	// ClaudeCodeSessionAffinity enables session-sticky routing for Claude Code clients.
-	// When enabled, requests with the same session ID (extracted from metadata.user_id)
-	// are routed to the same auth credential when available.
-	// Deprecated: Use SessionAffinity instead for universal session support.
-	ClaudeCodeSessionAffinity bool `yaml:"claude-code-session-affinity,omitempty" json:"claude-code-session-affinity,omitempty"`
-
 	// SessionAffinity enables universal session-sticky routing for all clients.
 	// Session IDs are extracted from multiple sources:
 	// metadata.user_id (Claude Code session format), X-Session-ID, Session_id (Codex),
@@ -403,6 +397,9 @@ type ClaudeKey struct {
 	// ExcludedModels lists model IDs that should be excluded for this provider.
 	ExcludedModels []string `yaml:"excluded-models,omitempty" json:"excluded-models,omitempty"`
 
+	// DisableCooling disables auth/model cooldown scheduling for this credential when true.
+	DisableCooling bool `yaml:"disable-cooling,omitempty" json:"disable-cooling,omitempty"`
+
 	// Cloak configures request cloaking for non-Claude-Code clients.
 	Cloak *CloakConfig `yaml:"cloak,omitempty" json:"cloak,omitempty"`
 
@@ -458,6 +455,9 @@ type CodexKey struct {
 
 	// ExcludedModels lists model IDs that should be excluded for this provider.
 	ExcludedModels []string `yaml:"excluded-models,omitempty" json:"excluded-models,omitempty"`
+
+	// DisableCooling disables auth/model cooldown scheduling for this credential when true.
+	DisableCooling bool `yaml:"disable-cooling,omitempty" json:"disable-cooling,omitempty"`
 }
 
 func (k CodexKey) GetAPIKey() string  { return k.APIKey }
@@ -502,6 +502,9 @@ type GeminiKey struct {
 
 	// ExcludedModels lists model IDs that should be excluded for this provider.
 	ExcludedModels []string `yaml:"excluded-models,omitempty" json:"excluded-models,omitempty"`
+
+	// DisableCooling disables auth/model cooldown scheduling for this credential when true.
+	DisableCooling bool `yaml:"disable-cooling,omitempty" json:"disable-cooling,omitempty"`
 }
 
 func (k GeminiKey) GetAPIKey() string  { return k.APIKey }
@@ -546,6 +549,9 @@ type OpenAICompatibility struct {
 
 	// Headers optionally adds extra HTTP headers for requests sent to this provider.
 	Headers map[string]string `yaml:"headers,omitempty" json:"headers,omitempty"`
+
+	// DisableCooling disables auth/model cooldown scheduling for this provider when true.
+	DisableCooling bool `yaml:"disable-cooling,omitempty" json:"disable-cooling,omitempty"`
 }
 
 // OpenAICompatibilityAPIKey represents an API key configuration with optional proxy setting.

internal/watcher/synthesizer/config.go

@@ -60,6 +60,10 @@ func (s *ConfigSynthesizer) synthesizeGeminiKeys(ctx *SynthesisContext) []*corea
 			"source":  fmt.Sprintf("config:gemini[%s]", token),
 			"api_key": key,
 		}
+		metadata := map[string]any{}
+		if entry.DisableCooling {
+			metadata["disable_cooling"] = true
+		}
 		if entry.Priority != 0 {
 			attrs["priority"] = strconv.Itoa(entry.Priority)
 		}
@@ -78,10 +82,14 @@ func (s *ConfigSynthesizer) synthesizeGeminiKeys(ctx *SynthesisContext) []*corea
 			Status:     coreauth.StatusActive,
 			ProxyURL:   proxyURL,
 			Attributes: attrs,
+			Metadata:   metadata,
 			CreatedAt:  now,
 			UpdatedAt:  now,
 		}
 		ApplyAuthExcludedModelsMeta(a, cfg, entry.ExcludedModels, "apikey")
+		if len(a.Metadata) == 0 {
+			a.Metadata = nil
+		}
 		out = append(out, a)
 	}
 	return out
@@ -107,6 +115,10 @@ func (s *ConfigSynthesizer) synthesizeClaudeKeys(ctx *SynthesisContext) []*corea
 			"source":  fmt.Sprintf("config:claude[%s]", token),
 			"api_key": key,
 		}
+		metadata := map[string]any{}
+		if ck.DisableCooling {
+			metadata["disable_cooling"] = true
+		}
 		if ck.Priority != 0 {
 			attrs["priority"] = strconv.Itoa(ck.Priority)
 		}
@@ -126,10 +138,14 @@ func (s *ConfigSynthesizer) synthesizeClaudeKeys(ctx *SynthesisContext) []*corea
 			Status:     coreauth.StatusActive,
 			ProxyURL:   proxyURL,
 			Attributes: attrs,
+			Metadata:   metadata,
 			CreatedAt:  now,
 			UpdatedAt:  now,
 		}
 		ApplyAuthExcludedModelsMeta(a, cfg, ck.ExcludedModels, "apikey")
+		if len(a.Metadata) == 0 {
+			a.Metadata = nil
+		}
 		out = append(out, a)
 	}
 	return out
@@ -154,6 +170,10 @@ func (s *ConfigSynthesizer) synthesizeCodexKeys(ctx *SynthesisContext) []*coreau
 			"source":  fmt.Sprintf("config:codex[%s]", token),
 			"api_key": key,
 		}
+		metadata := map[string]any{}
+		if ck.DisableCooling {
+			metadata["disable_cooling"] = true
+		}
 		if ck.Priority != 0 {
 			attrs["priority"] = strconv.Itoa(ck.Priority)
 		}
@@ -176,10 +196,14 @@ func (s *ConfigSynthesizer) synthesizeCodexKeys(ctx *SynthesisContext) []*coreau
 			Status:     coreauth.StatusActive,
 			ProxyURL:   proxyURL,
 			Attributes: attrs,
+			Metadata:   metadata,
 			CreatedAt:  now,
 			UpdatedAt:  now,
 		}
 		ApplyAuthExcludedModelsMeta(a, cfg, ck.ExcludedModels, "apikey")
+		if len(a.Metadata) == 0 {
+			a.Metadata = nil
+		}
 		out = append(out, a)
 	}
 	return out
@@ -203,6 +227,7 @@ func (s *ConfigSynthesizer) synthesizeOpenAICompat(ctx *SynthesisContext) []*cor
 			providerName = "openai-compatibility"
 		}
 		base := strings.TrimSpace(compat.BaseURL)
+		disableCooling := compat.DisableCooling
 
 		// Handle new APIKeyEntries format (preferred)
 		createdEntries := 0
@@ -218,6 +243,10 @@ func (s *ConfigSynthesizer) synthesizeOpenAICompat(ctx *SynthesisContext) []*cor
 				"compat_name":  compat.Name,
 				"provider_key": providerName,
 			}
+			metadata := map[string]any{}
+			if disableCooling {
+				metadata["disable_cooling"] = true
+			}
 			if compat.Priority != 0 {
 				attrs["priority"] = strconv.Itoa(compat.Priority)
 			}
@@ -236,9 +265,13 @@ func (s *ConfigSynthesizer) synthesizeOpenAICompat(ctx *SynthesisContext) []*cor
 				Status:     coreauth.StatusActive,
 				ProxyURL:   proxyURL,
 				Attributes: attrs,
+				Metadata:   metadata,
 				CreatedAt:  now,
 				UpdatedAt:  now,
 			}
+			if len(a.Metadata) == 0 {
+				a.Metadata = nil
+			}
 			out = append(out, a)
 			createdEntries++
 		}
@@ -252,6 +285,10 @@ func (s *ConfigSynthesizer) synthesizeOpenAICompat(ctx *SynthesisContext) []*cor
 				"compat_name":  compat.Name,
 				"provider_key": providerName,
 			}
+			metadata := map[string]any{}
+			if disableCooling {
+				metadata["disable_cooling"] = true
+			}
 			if compat.Priority != 0 {
 				attrs["priority"] = strconv.Itoa(compat.Priority)
 			}
@@ -266,9 +303,13 @@ func (s *ConfigSynthesizer) synthesizeOpenAICompat(ctx *SynthesisContext) []*cor
 				Prefix:     prefix,
 				Status:     coreauth.StatusActive,
 				Attributes: attrs,
+				Metadata:   metadata,
 				CreatedAt:  now,
 				UpdatedAt:  now,
 			}
+			if len(a.Metadata) == 0 {
+				a.Metadata = nil
+			}
 			out = append(out, a)
 		}
 	}

internal/watcher/synthesizer/config_test.go

@@ -68,11 +68,26 @@ func TestConfigSynthesizer_GeminiKeys(t *testing.T) {
 				if auths[0].Attributes["api_key"] != "test-key-123" {
 					t.Errorf("expected api_key test-key-123, got %s", auths[0].Attributes["api_key"])
 				}
+				if auths[0].Metadata != nil {
+					t.Errorf("expected metadata to be nil when disable_cooling not set, got %v", auths[0].Metadata)
+				}
 				if auths[0].Status != coreauth.StatusActive {
 					t.Errorf("expected status active, got %s", auths[0].Status)
 				}
 			},
 		},
+		{
+			name: "gemini key disable cooling",
+			geminiKeys: []config.GeminiKey{
+				{APIKey: "test-key-123", Prefix: "team-a", DisableCooling: true},
+			},
+			wantLen: 1,
+			validate: func(t *testing.T, auths []*coreauth.Auth) {
+				if v, ok := auths[0].Metadata["disable_cooling"].(bool); !ok || !v {
+					t.Errorf("expected disable_cooling=true, got %v", auths[0].Metadata["disable_cooling"])
+				}
+			},
+		},
 		{
 			name: "gemini key with base url and proxy",
 			geminiKeys: []config.GeminiKey{
@@ -160,9 +175,10 @@ func TestConfigSynthesizer_ClaudeKeys(t *testing.T) {
 		Config: &config.Config{
 			ClaudeKey: []config.ClaudeKey{
 				{
-					APIKey:  "sk-ant-api-xxx",
-					Prefix:  "main",
-					BaseURL: "https://api.anthropic.com",
+					APIKey:         "sk-ant-api-xxx",
+					Prefix:         "main",
+					BaseURL:        "https://api.anthropic.com",
+					DisableCooling: true,
 					Models: []config.ClaudeModel{
 						{Name: "claude-3-opus"},
 						{Name: "claude-3-sonnet"},
@@ -197,6 +213,9 @@ func TestConfigSynthesizer_ClaudeKeys(t *testing.T) {
 	if _, ok := auths[0].Attributes["models_hash"]; !ok {
 		t.Error("expected models_hash in attributes")
 	}
+	if v, ok := auths[0].Metadata["disable_cooling"].(bool); !ok || !v {
+		t.Errorf("expected disable_cooling=true, got %v", auths[0].Metadata["disable_cooling"])
+	}
 }
 
 func TestConfigSynthesizer_ClaudeKeys_SkipsEmptyAndHeaders(t *testing.T) {
@@ -231,11 +250,12 @@ func TestConfigSynthesizer_CodexKeys(t *testing.T) {
 		Config: &config.Config{
 			CodexKey: []config.CodexKey{
 				{
-					APIKey:     "codex-key-123",
-					Prefix:     "dev",
-					BaseURL:    "https://api.openai.com",
-					ProxyURL:   "http://proxy.local",
-					Websockets: true,
+					APIKey:         "codex-key-123",
+					Prefix:         "dev",
+					BaseURL:        "https://api.openai.com",
+					ProxyURL:       "http://proxy.local",
+					Websockets:     true,
+					DisableCooling: true,
 				},
 			},
 		},
@@ -263,6 +283,9 @@ func TestConfigSynthesizer_CodexKeys(t *testing.T) {
 	if auths[0].Attributes["websockets"] != "true" {
 		t.Errorf("expected websockets=true, got %s", auths[0].Attributes["websockets"])
 	}
+	if v, ok := auths[0].Metadata["disable_cooling"].(bool); !ok || !v {
+		t.Errorf("expected disable_cooling=true, got %v", auths[0].Metadata["disable_cooling"])
+	}
 }
 
 func TestConfigSynthesizer_CodexKeys_SkipsEmptyAndHeaders(t *testing.T) {
@@ -301,8 +324,9 @@ func TestConfigSynthesizer_OpenAICompat(t *testing.T) {
 			name: "with APIKeyEntries",
 			compat: []config.OpenAICompatibility{
 				{
-					Name:    "CustomProvider",
-					BaseURL: "https://custom.api.com",
+					Name:           "CustomProvider",
+					BaseURL:        "https://custom.api.com",
+					DisableCooling: true,
 					APIKeyEntries: []config.OpenAICompatibilityAPIKey{
 						{APIKey: "key-1"},
 						{APIKey: "key-2"},
@@ -365,6 +389,13 @@ func TestConfigSynthesizer_OpenAICompat(t *testing.T) {
 			if len(auths) != tt.wantLen {
 				t.Fatalf("expected %d auths, got %d", tt.wantLen, len(auths))
 			}
+			if tt.name == "with APIKeyEntries" {
+				for i := range auths {
+					if v, ok := auths[i].Metadata["disable_cooling"].(bool); !ok || !v {
+						t.Fatalf("expected auth[%d].disable_cooling=true, got %v", i, auths[i].Metadata["disable_cooling"])
+					}
+				}
+			}
 		})
 	}
 }

sdk/cliproxy/auth/types.go

@@ -355,19 +355,28 @@ func (a *Auth) ProxyInfo() string {
 	return "via proxy"
 }
 
-// DisableCoolingOverride returns the auth-file scoped disable_cooling override when present.
+// DisableCoolingOverride returns the auth scoped disable_cooling override when present.
 // The value is read from metadata key "disable_cooling" (or legacy "disable-cooling").
+//
+// NOTE: This override is intentionally "true-only". When the metadata value is false, it is treated
+// as "not set" so the global disable-cooling flag can still take effect.
 func (a *Auth) DisableCoolingOverride() (bool, bool) {
 	if a == nil || a.Metadata == nil {
 		return false, false
 	}
 	if val, ok := a.Metadata["disable_cooling"]; ok {
 		if parsed, okParse := parseBoolAny(val); okParse {
+			if !parsed {
+				return false, false
+			}
 			return parsed, true
 		}
 	}
 	if val, ok := a.Metadata["disable-cooling"]; ok {
 		if parsed, okParse := parseBoolAny(val); okParse {
+			if !parsed {
+				return false, false
+			}
 			return parsed, true
 		}
 	}