fix: 修正时间戳精度、TTL 定价匹配逻辑及账单快照展示

- 将 created_at_unix_ms 从 current_unix_secs 改为 current_unix_ms，修正候选尝试和跳过记录的时间戳精度
- formula_engine: TTL 定价从「<=上限」模糊匹配改为精确匹配，null 值改为回退到基础价格而非透传
- 新增 ttl_pricing_requires_exact_match 和 ttl_pricing_null_value_falls_back_to_base_tier_value 测试用例
- service: 新增 5min/1h cache TTL 的端到端计费验证测试
- RequestDetailDrawer: 优先从 billing_snapshot 读取已解析的价格和费用，正确展示当前 TTL 对应的缓存创建/读取价格，修正输入/输出/缓存费用列的数据来源

apps/aether-gateway/src/ai_pipeline/planner/passthrough/provider/family/candidates.rs

@@ -9,7 +9,7 @@ use crate::ai_pipeline::{
     resolve_local_decision_execution_runtime_auth_context, ConversionMode, ExecutionStrategy,
     GatewayControlDecision, PlannerAppState,
 };
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::{append_execution_contract_fields_to_value, AppState, GatewayError};
 
 use super::{
@@ -105,7 +105,7 @@ pub(crate) async fn materialize_local_same_format_provider_candidate_attempts(
     )
     .await;
 
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
     for (candidate_index, candidate) in candidates.into_iter().enumerate() {

apps/aether-gateway/src/ai_pipeline/planner/passthrough/provider/family/payload.rs

@@ -21,7 +21,7 @@ use crate::ai_pipeline::transport::{
 use crate::ai_pipeline::{
     collect_control_headers, ConversionMode, ExecutionStrategy, PlannerAppState,
 };
-use crate::clock::current_unix_secs;
+use crate::clock::current_unix_ms;
 use crate::{
     append_execution_contract_fields_to_value, AppState, GatewayControlSyncDecisionResponse,
     EXECUTION_RUNTIME_STREAM_DECISION_ACTION, EXECUTION_RUNTIME_SYNC_DECISION_ACTION,
@@ -365,7 +365,7 @@ pub(super) async fn mark_skipped_local_same_format_provider_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local same-format decision failed to persist skipped candidate",
         )
         .await;

apps/aether-gateway/src/ai_pipeline/planner/specialized/files/support.rs

@@ -11,7 +11,7 @@ use crate::ai_pipeline::{
     resolve_local_decision_execution_runtime_auth_context, GatewayControlDecision,
 };
 use crate::ai_pipeline::{GatewayAuthApiKeySnapshot, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::{AppState, GatewayError};
 
 pub(super) const GEMINI_FILES_CANDIDATE_API_FORMAT: &str = "gemini:chat";
@@ -102,7 +102,7 @@ pub(super) async fn materialize_local_gemini_files_candidate_attempts(
     )
     .await;
 
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
     for (candidate_index, candidate) in candidates.into_iter().enumerate() {
@@ -174,7 +174,7 @@ pub(super) async fn mark_skipped_local_gemini_files_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local gemini files failed to persist skipped candidate",
         )
         .await;

apps/aether-gateway/src/ai_pipeline/planner/specialized/video/support.rs

@@ -12,7 +12,7 @@ use crate::ai_pipeline::{
     resolve_local_decision_execution_runtime_auth_context, GatewayControlDecision,
 };
 use crate::ai_pipeline::{GatewayAuthApiKeySnapshot, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::AppState;
 
 #[derive(Debug, Clone)]
@@ -148,7 +148,7 @@ async fn materialize_local_video_create_candidate_attempts(
         input.required_capabilities.as_ref(),
     )
     .await;
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
 
@@ -220,7 +220,7 @@ pub(super) async fn mark_skipped_local_video_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local video decision failed to persist skipped candidate",
         )
         .await;

apps/aether-gateway/src/ai_pipeline/planner/standard/family/candidates.rs

@@ -17,7 +17,7 @@ use crate::ai_pipeline::{
     GatewayControlDecision,
 };
 use crate::ai_pipeline::{GatewayAuthApiKeySnapshot, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::{append_execution_contract_fields_to_value, AppState, GatewayError};
 
 use super::{
@@ -145,7 +145,7 @@ pub(super) async fn materialize_local_standard_candidate_attempts(
     )
     .await;
 
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
     for (candidate_index, candidate) in candidates.into_iter().enumerate() {

apps/aether-gateway/src/ai_pipeline/planner/standard/family/payload.rs

@@ -14,7 +14,7 @@ use crate::ai_pipeline::transport::{
 };
 use crate::ai_pipeline::{collect_control_headers, ConversionMode, ExecutionStrategy};
 use crate::ai_pipeline::{LocalResolvedOAuthRequestAuth, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::current_unix_ms;
 use crate::{
     append_execution_contract_fields_to_value, AppState, GatewayControlSyncDecisionResponse,
     EXECUTION_RUNTIME_STREAM_DECISION_ACTION, EXECUTION_RUNTIME_SYNC_DECISION_ACTION,
@@ -390,7 +390,7 @@ pub(super) async fn mark_skipped_local_standard_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local standard decision failed to persist skipped candidate",
         )
         .await;

apps/aether-gateway/src/ai_pipeline/planner/standard/openai/chat/decision/support.rs

@@ -12,7 +12,7 @@ use crate::ai_pipeline::planner::candidate_affinity::{
 };
 use crate::ai_pipeline::GatewayAuthApiKeySnapshot;
 use crate::ai_pipeline::{ConversionMode, ExecutionStrategy, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::{append_execution_contract_fields_to_value, AppState};
 
 #[derive(Debug, Clone)]
@@ -57,7 +57,7 @@ pub(crate) async fn mark_skipped_local_openai_chat_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local openai chat decision failed to persist skipped candidate",
         )
         .await;
@@ -77,7 +77,7 @@ pub(crate) async fn materialize_local_openai_chat_candidate_attempts(
         input.required_capabilities.as_ref(),
     )
     .await;
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
 

apps/aether-gateway/src/ai_pipeline/planner/standard/openai/cli/decision/support.rs

@@ -18,7 +18,7 @@ use crate::ai_pipeline::{
     GatewayControlDecision,
 };
 use crate::ai_pipeline::{GatewayAuthApiKeySnapshot, PlannerAppState};
-use crate::clock::current_unix_secs;
+use crate::clock::{current_unix_ms, current_unix_secs};
 use crate::{append_execution_contract_fields_to_value, AppState, GatewayError};
 
 use super::LocalOpenAiCliSpec;
@@ -152,7 +152,7 @@ pub(crate) async fn materialize_local_openai_cli_candidate_attempts(
     )
     .await;
 
-    let created_at_unix_ms = current_unix_secs();
+    let created_at_unix_ms = current_unix_ms();
     let mut attempts = Vec::with_capacity(candidates.len());
     let mut affinity_remembered = false;
     for (candidate_index, candidate) in candidates.into_iter().enumerate() {
@@ -314,7 +314,7 @@ pub(crate) async fn mark_skipped_local_openai_cli_candidate(
             candidate_id,
             input.required_capabilities.as_ref(),
             skip_reason,
-            current_unix_secs(),
+            current_unix_ms(),
             "gateway local openai cli decision failed to persist skipped candidate",
         )
         .await;

crates/aether-billing/src/formula_engine.rs

@@ -380,28 +380,21 @@ fn resolve_tiered(
 
     if let (Some(cache_ttl_minutes), Some(ttl_value_key)) = (cache_ttl_minutes, ttl_value_key) {
         if let Some(ttl_pricing) = tier.get("cache_ttl_pricing").and_then(|v| v.as_array()) {
-            for ttl_entry in ttl_pricing {
+            if let Some(ttl_entry) = ttl_pricing.iter().find(|entry| {
-                let ttl_limit = ttl_entry
+                entry
                     .get("ttl_minutes")
                     .and_then(as_f64)
-                    .unwrap_or_default();
+                    .map(|value| value == cache_ttl_minutes)
-                if cache_ttl_minutes <= ttl_limit {
+                    .unwrap_or(false)
-                    if let Some(value) = ttl_entry.get(ttl_value_key) {
+            }) {
-                        return Ok((
+                if let Some(value) = ttl_entry.get(ttl_value_key).filter(|value| !value.is_null()) {
-                            value.clone(),
+                    return Ok((
-                            false,
+                        value.clone(),
-                            Some((matched_index.unwrap_or(0), tier)),
+                        false,
-                        ));
+                        Some((matched_index.unwrap_or(0), tier)),
-                    }
+                    ));
                 }
             }
-            if let Some(value) = ttl_pricing.last().and_then(|v| v.get(ttl_value_key)) {
-                return Ok((
-                    value.clone(),
-                    false,
-                    Some((matched_index.unwrap_or(0), tier)),
-                ));
-            }
         }
     }
 
@@ -857,4 +850,82 @@ mod tests {
         assert_eq!(result.status, FormulaEvaluationStatus::Incomplete);
         assert_eq!(result.missing_required, vec!["input_tokens".to_string()]);
     }
+
+    #[test]
+    fn ttl_pricing_requires_exact_match() {
+        let engine = FormulaEngine::new();
+        let dimensions = BTreeMap::from([
+            ("total_input_context".to_string(), serde_json::json!(22_562)),
+            ("cache_ttl_minutes".to_string(), serde_json::json!(5)),
+        ]);
+        let mappings = BTreeMap::from([(
+            "cache_creation_price_per_1m".to_string(),
+            serde_json::json!({
+                "source": "tiered",
+                "tier_key": "total_input_context",
+                "ttl_key": "cache_ttl_minutes",
+                "ttl_value_key": "cache_creation_price_per_1m",
+                "tiers": [{
+                    "up_to": null,
+                    "value": 3.125,
+                    "cache_ttl_pricing": [{
+                        "ttl_minutes": 60,
+                        "cache_creation_price_per_1m": 5.0
+                    }]
+                }],
+                "default": 0.0
+            }),
+        )]);
+
+        let result = engine
+            .evaluate(
+                "cache_creation_price_per_1m",
+                None,
+                Some(&dimensions),
+                Some(&mappings),
+                false,
+            )
+            .expect("tiered mapping should evaluate");
+
+        assert_eq!(result.cost, 3.125);
+    }
+
+    #[test]
+    fn ttl_pricing_null_value_falls_back_to_base_tier_value() {
+        let engine = FormulaEngine::new();
+        let dimensions = BTreeMap::from([
+            ("total_input_context".to_string(), serde_json::json!(22_562)),
+            ("cache_ttl_minutes".to_string(), serde_json::json!(60)),
+        ]);
+        let mappings = BTreeMap::from([(
+            "cache_read_price_per_1m".to_string(),
+            serde_json::json!({
+                "source": "tiered",
+                "tier_key": "total_input_context",
+                "ttl_key": "cache_ttl_minutes",
+                "ttl_value_key": "cache_read_price_per_1m",
+                "tiers": [{
+                    "up_to": null,
+                    "value": 0.25,
+                    "cache_ttl_pricing": [{
+                        "ttl_minutes": 60,
+                        "cache_read_price_per_1m": null
+                    }]
+                }],
+                "default": 0.0
+            }),
+        )]);
+
+        let result = engine
+            .evaluate(
+                "cache_read_price_per_1m",
+                None,
+                Some(&dimensions),
+                Some(&mappings),
+                false,
+            )
+            .expect("tiered mapping should evaluate");
+
+        assert_eq!(result.cost, 0.25);
+    }
 }

crates/aether-billing/src/service.rs

@@ -264,4 +264,142 @@ mod tests {
         assert!(result.actual_total_cost > 0.0);
         assert_eq!(result.rate_multiplier, 0.5);
     }
+
+    #[test]
+    fn five_minute_cache_ttl_uses_base_cache_prices() {
+        let pricing = BillingModelPricingSnapshot {
+            provider_id: "provider-1".to_string(),
+            provider_billing_type: Some("pay_as_you_go".to_string()),
+            provider_api_key_id: Some("key-1".to_string()),
+            provider_api_key_rate_multipliers: None,
+            provider_api_key_cache_ttl_minutes: Some(5),
+            global_model_id: "global-model-1".to_string(),
+            global_model_name: "gpt-5.4".to_string(),
+            global_model_config: None,
+            default_price_per_request: None,
+            default_tiered_pricing: Some(json!({
+                "tiers": [{
+                    "up_to": null,
+                    "input_price_per_1m": 2.5,
+                    "output_price_per_1m": 15.0,
+                    "cache_creation_price_per_1m": 3.125,
+                    "cache_read_price_per_1m": 0.25,
+                    "cache_ttl_pricing": [{
+                        "ttl_minutes": 60,
+                        "cache_creation_price_per_1m": 5.0,
+                        "cache_read_price_per_1m": null
+                    }]
+                }]
+            })),
+            model_id: None,
+            model_provider_model_name: None,
+            model_config: None,
+            model_price_per_request: None,
+            model_tiered_pricing: None,
+        };
+
+        let result = BillingService::new()
+            .calculate(
+                &pricing,
+                &BillingUsageInput {
+                    task_type: "chat".to_string(),
+                    api_format: None,
+                    request_count: 1,
+                    input_tokens: 1_000,
+                    output_tokens: 10,
+                    cache_creation_tokens: 0,
+                    cache_creation_ephemeral_5m_tokens: 0,
+                    cache_creation_ephemeral_1h_tokens: 0,
+                    cache_read_tokens: 100,
+                    cache_ttl_minutes: Some(5),
+                },
+            )
+            .expect("billing should calculate");
+
+        assert_eq!(
+            result
+                .cost_result
+                .snapshot
+                .resolved_variables
+                .get("cache_creation_price_per_1m"),
+            Some(&json!(3.125))
+        );
+        assert_eq!(
+            result
+                .cost_result
+                .snapshot
+                .resolved_variables
+                .get("cache_read_price_per_1m"),
+            Some(&json!(0.25))
+        );
+    }
+
+    #[test]
+    fn one_hour_cache_ttl_keeps_base_cache_read_when_ttl_entry_omits_it() {
+        let pricing = BillingModelPricingSnapshot {
+            provider_id: "provider-1".to_string(),
+            provider_billing_type: Some("pay_as_you_go".to_string()),
+            provider_api_key_id: Some("key-1".to_string()),
+            provider_api_key_rate_multipliers: None,
+            provider_api_key_cache_ttl_minutes: Some(60),
+            global_model_id: "global-model-1".to_string(),
+            global_model_name: "gpt-5.4".to_string(),
+            global_model_config: None,
+            default_price_per_request: None,
+            default_tiered_pricing: Some(json!({
+                "tiers": [{
+                    "up_to": null,
+                    "input_price_per_1m": 2.5,
+                    "output_price_per_1m": 15.0,
+                    "cache_creation_price_per_1m": 3.125,
+                    "cache_read_price_per_1m": 0.25,
+                    "cache_ttl_pricing": [{
+                        "ttl_minutes": 60,
+                        "cache_creation_price_per_1m": 5.0,
+                        "cache_read_price_per_1m": null
+                    }]
+                }]
+            })),
+            model_id: None,
+            model_provider_model_name: None,
+            model_config: None,
+            model_price_per_request: None,
+            model_tiered_pricing: None,
+        };
+
+        let result = BillingService::new()
+            .calculate(
+                &pricing,
+                &BillingUsageInput {
+                    task_type: "chat".to_string(),
+                    api_format: None,
+                    request_count: 1,
+                    input_tokens: 1_000,
+                    output_tokens: 10,
+                    cache_creation_tokens: 0,
+                    cache_creation_ephemeral_5m_tokens: 0,
+                    cache_creation_ephemeral_1h_tokens: 0,
+                    cache_read_tokens: 100,
+                    cache_ttl_minutes: Some(60),
+                },
+            )
+            .expect("billing should calculate");
+
+        assert_eq!(
+            result
+                .cost_result
+                .snapshot
+                .resolved_variables
+                .get("cache_creation_price_per_1m"),
+            Some(&json!(5.0))
+        );
+        assert_eq!(
+            result
+                .cost_result
+                .snapshot
+                .resolved_variables
+                .get("cache_read_price_per_1m"),
+            Some(&json!(0.25))
+        );
+    }
 }

frontend/src/features/usage/components/RequestDetailDrawer.vue

@@ -237,19 +237,12 @@
                         <div class="text-muted-foreground flex items-center gap-2 flex-wrap">
                           <span>输入 ${{ formatPrice(tier.input_price_per_1m) }}/M</span>
                           <span>输出 ${{ formatPrice(tier.output_price_per_1m) }}/M</span>
-                          <template v-if="hasTierCacheCreationSplitPricing(tier)">
+                          <span v-if="getTierActiveCacheCreationDisplay(tier)">
-                            <span v-if="getTierCachePriceForTTL(tier, 5, 'cache_creation_price_per_1m') !== null">
+                            {{ getTierActiveCacheCreationDisplay(tier)?.label }}
-                              缓存创建(5min) ${{ formatPrice(getTierCachePriceForTTL(tier, 5, 'cache_creation_price_per_1m') || 0) }}/M
+                            ${{ formatPrice(getTierActiveCacheCreationDisplay(tier)?.price || 0) }}/M
-                            </span>
-                            <span v-if="getTierCachePriceForTTL(tier, 60, 'cache_creation_price_per_1m') !== null">
-                              缓存创建(1h) ${{ formatPrice(getTierCachePriceForTTL(tier, 60, 'cache_creation_price_per_1m') || 0) }}/M
-                            </span>
-                          </template>
-                          <span v-else-if="tier.cache_creation_price_per_1m">
-                            缓存创建 ${{ formatPrice(tier.cache_creation_price_per_1m) }}/M
                           </span>
-                          <span v-if="tier.cache_read_price_per_1m">
+                          <span v-if="shouldShowCacheReadPrice(tier)">
-                            缓存读取 ${{ formatPrice(tier.cache_read_price_per_1m) }}/M
+                            缓存读取 ${{ formatPrice(getTierActiveCacheReadPrice(tier) ?? 0) }}/M
                           </span>
                         </div>
                       </div>
@@ -261,7 +254,7 @@
                           <div class="flex items-center flex-1">
                             <span class="text-xs text-muted-foreground w-[56px]">输入</span>
                             <span class="text-sm font-semibold font-mono flex-1 text-center">{{ displayInputTokens }}</span>
-                            <span class="text-xs font-mono">${{ (detail.cost?.input || detail.input_cost || 0).toFixed(6) }}</span>
+                            <span class="text-xs font-mono">${{ effectiveInputCost.toFixed(6) }}</span>
                           </div>
                           <Separator
                             orientation="vertical"
@@ -270,15 +263,15 @@
                           <div class="flex items-center flex-1">
                             <span class="text-xs text-muted-foreground w-[56px]">输出</span>
                             <span class="text-sm font-semibold font-mono flex-1 text-center">{{ detail.tokens?.output || detail.output_tokens || 0 }}</span>
-                            <span class="text-xs font-mono">${{ (detail.cost?.output || detail.output_cost || 0).toFixed(6) }}</span>
+                            <span class="text-xs font-mono">${{ effectiveOutputCost.toFixed(6) }}</span>
                           </div>
                         </div>
                         <!-- 缓存创建 缓存读取 -->
                         <div class="flex items-center">
                           <div class="flex items-center flex-1">
                             <span class="text-xs text-muted-foreground w-[56px]">{{ cacheCreationSplitRows.length > 0 ? '创建合计' : '缓存创建' }}</span>
-                            <span class="text-sm font-semibold font-mono flex-1 text-center">{{ detail.cache_creation_input_tokens || 0 }}</span>
+                            <span class="text-sm font-semibold font-mono flex-1 text-center">{{ totalCacheCreationTokens }}</span>
-                            <span class="text-xs font-mono">${{ (detail.cache_creation_cost || 0).toFixed(6) }}</span>
+                            <span class="text-xs font-mono">${{ effectiveCacheCreationCost.toFixed(6) }}</span>
                           </div>
                           <Separator
                             orientation="vertical"
@@ -287,7 +280,7 @@
                           <div class="flex items-center flex-1">
                             <span class="text-xs text-muted-foreground w-[56px]">缓存读取</span>
                             <span class="text-sm font-semibold font-mono flex-1 text-center">{{ detail.cache_read_input_tokens || 0 }}</span>
-                            <span class="text-xs font-mono">${{ (detail.cache_read_cost || 0).toFixed(6) }}</span>
+                            <span class="text-xs font-mono">${{ effectiveCacheReadCost.toFixed(6) }}</span>
                           </div>
                         </div>
                         <!-- 缓存创建 5m/1h 细分 -->
@@ -323,10 +316,10 @@
                     </div>
                     <div class="rounded-lg p-3 bg-primary/5 border border-primary/30 space-y-2">
                       <div
-                        v-if="detail.price_per_request"
+                        v-if="effectivePricePerRequest > 0"
                         class="flex items-center justify-end text-xs"
                       >
-                        <span class="text-muted-foreground">${{ detail.price_per_request.toFixed(6) }}/次</span>
+                        <span class="text-muted-foreground">${{ effectivePricePerRequest.toFixed(6) }}/次</span>
                       </div>
                       <div class="flex items-center">
                         <div class="flex items-center flex-1">
@@ -753,10 +746,55 @@ type CacheTTLPriceEntry = {
 }
 
 type PricingTierLike = {
+  up_to?: number | null
+  input_price_per_1m?: number | null
+  output_price_per_1m?: number | null
   cache_creation_price_per_1m?: number | null
   cache_read_price_per_1m?: number | null
   cache_ttl_pricing?: CacheTTLPriceEntry[] | null
 }
+
+type JsonRecord = Record<string, unknown>
+
+function asRecord(value: unknown): JsonRecord | null {
+  if (!value || typeof value !== 'object' || Array.isArray(value)) return null
+  return value as JsonRecord
+}
+
+function toNumber(value: unknown): number | null {
+  if (typeof value === 'number' && Number.isFinite(value)) return value
+  if (typeof value === 'string') {
+    const parsed = Number(value)
+    return Number.isFinite(parsed) ? parsed : null
+  }
+  return null
+}
+
+function getNestedValue(record: JsonRecord | null, ...path: string[]): unknown {
+  let current: unknown = record
+  for (const key of path) {
+    const object = asRecord(current)
+    if (!object) return null
+    current = object[key]
+  }
+  return current
+}
+
+function getNestedNumber(record: JsonRecord | null, ...path: string[]): number | null {
+  return toNumber(getNestedValue(record, ...path))
+}
+
+function normalizeCacheTtlPricing(value: unknown): CacheTTLPriceEntry[] {
+  if (!Array.isArray(value)) return []
+  return value
+    .map((entry) => asRecord(entry))
+    .filter((entry): entry is JsonRecord => entry !== null)
+    .map((entry) => ({
+      ttl_minutes: toNumber(entry.ttl_minutes),
+      cache_creation_price_per_1m: toNumber(entry.cache_creation_price_per_1m),
+      cache_read_price_per_1m: toNumber(entry.cache_read_price_per_1m),
+    }))
+}
 const autoRefreshTimer = ref<ReturnType<typeof setInterval> | null>(null)
 const autoRefreshing = ref(false)
 const isPageVisible = ref(typeof document === 'undefined' ? true : !document.hidden)
@@ -813,6 +851,22 @@ const traceRequestMetadata = computed<Record<string, unknown> | null>(() => {
   return meta as Record<string, unknown>
 })
 
+const billingSnapshot = computed<JsonRecord | null>(() =>
+  asRecord(traceRequestMetadata.value?.billing_snapshot),
+)
+
+const billingResolvedVariables = computed<JsonRecord | null>(() =>
+  asRecord(billingSnapshot.value?.resolved_variables),
+)
+
+const billingCostBreakdown = computed<JsonRecord | null>(() =>
+  asRecord(billingSnapshot.value?.cost_breakdown),
+)
+
+const billingTierInfo = computed<JsonRecord | null>(() =>
+  asRecord(billingSnapshot.value?.tier_info),
+)
+
 function hasBodyContent(flag: boolean | undefined, data: unknown): boolean {
   return Boolean(flag) || hasContent(data)
 }
@@ -1022,6 +1076,136 @@ const priceSourceLabel = computed(() => {
   return '历史定价'
 })
 
+const cacheCreationInputTokens5m = computed(() => {
+  if (!detail.value) return 0
+  return toNumber(detail.value.cache_creation_input_tokens_5m)
+    ?? getNestedNumber(detail.value as unknown as JsonRecord, 'cache_creation_ephemeral_5m_input_tokens')
+    ?? 0
+})
+
+const cacheCreationInputTokens1h = computed(() => {
+  if (!detail.value) return 0
+  return toNumber(detail.value.cache_creation_input_tokens_1h)
+    ?? getNestedNumber(detail.value as unknown as JsonRecord, 'cache_creation_ephemeral_1h_input_tokens')
+    ?? 0
+})
+
+const totalCacheCreationTokens = computed(() => {
+  if (!detail.value) return 0
+  const classified = cacheCreationInputTokens5m.value + cacheCreationInputTokens1h.value
+  const total = detail.value.cache_creation_input_tokens || 0
+  if (total === 0 && classified > 0) return classified
+  return total
+})
+
+const effectiveInputPricePer1M = computed(() =>
+  getNestedNumber(billingResolvedVariables.value, 'input_price_per_1m')
+  ?? toNumber(detail.value?.input_price_per_1m)
+)
+
+const effectiveOutputPricePer1M = computed(() =>
+  getNestedNumber(billingResolvedVariables.value, 'output_price_per_1m')
+  ?? toNumber(detail.value?.output_price_per_1m)
+)
+
+const effectiveCacheCreationPricePer1M = computed(() =>
+  getNestedNumber(billingResolvedVariables.value, 'cache_creation_price_per_1m')
+  ?? toNumber(detail.value?.cache_creation_price_per_1m)
+)
+
+const effectiveCacheReadPricePer1M = computed(() =>
+  getNestedNumber(billingResolvedVariables.value, 'cache_read_price_per_1m')
+  ?? toNumber(detail.value?.cache_read_price_per_1m)
+)
+
+const effectivePricePerRequest = computed(() =>
+  getNestedNumber(billingResolvedVariables.value, 'price_per_request')
+  ?? toNumber(detail.value?.price_per_request)
+  ?? 0,
+)
+
+const effectiveInputCost = computed(() =>
+  getNestedNumber(billingCostBreakdown.value, 'input_cost')
+  ?? toNumber(detail.value?.input_cost)
+  ?? 0,
+)
+
+const effectiveOutputCost = computed(() =>
+  getNestedNumber(billingCostBreakdown.value, 'output_cost')
+  ?? toNumber(detail.value?.output_cost)
+  ?? 0,
+)
+
+const effectiveCacheCreationCost = computed(() => {
+  const snapshotCost = [
+    getNestedNumber(billingCostBreakdown.value, 'cache_creation_uncategorized_cost'),
+    getNestedNumber(billingCostBreakdown.value, 'cache_creation_ephemeral_5m_cost'),
+    getNestedNumber(billingCostBreakdown.value, 'cache_creation_ephemeral_1h_cost'),
+  ].reduce((sum, value) => sum + (value ?? 0), 0)
+  if (snapshotCost > 0) return snapshotCost
+  return toNumber(detail.value?.cache_creation_cost) ?? 0
+})
+
+const effectiveCacheReadCost = computed(() =>
+  getNestedNumber(billingCostBreakdown.value, 'cache_read_cost')
+  ?? toNumber(detail.value?.cache_read_cost)
+  ?? 0,
+)
+
+const effectiveRequestCost = computed(() => {
+  const snapshotCost = getNestedNumber(billingCostBreakdown.value, 'request_cost')
+  if (snapshotCost !== null) return snapshotCost
+  if (effectivePricePerRequest.value > 0) {
+    return toNumber(detail.value?.request_cost) ?? 0
+  }
+  return 0
+})
+
+const fallbackCacheTtlPricing = computed<CacheTTLPriceEntry[]>(() => {
+  const tierPricing = normalizeCacheTtlPricing(billingTierInfo.value?.cache_ttl_pricing)
+  if (tierPricing.length > 0) return tierPricing
+
+  const rows: CacheTTLPriceEntry[] = []
+  const cache5mCreationPrice = getNestedNumber(
+    billingResolvedVariables.value,
+    'cache_creation_ephemeral_5m_price_per_1m',
+  )
+  const cache1hCreationPrice = getNestedNumber(
+    billingResolvedVariables.value,
+    'cache_creation_ephemeral_1h_price_per_1m',
+  )
+
+  if (cache5mCreationPrice !== null) {
+    rows.push({
+      ttl_minutes: 5,
+      cache_creation_price_per_1m: cache5mCreationPrice,
+      cache_read_price_per_1m: effectiveCacheReadPricePer1M.value,
+    })
+  }
+  if (cache1hCreationPrice !== null) {
+    rows.push({
+      ttl_minutes: 60,
+      cache_creation_price_per_1m: cache1hCreationPrice,
+      cache_read_price_per_1m: effectiveCacheReadPricePer1M.value,
+    })
+  }
+  return rows
+})
+
+const activeCacheTtlMinutes = computed(() => {
+  const snapshotTtl = getNestedNumber(billingSnapshot.value, 'resolved_dimensions', 'cache_ttl_minutes')
+  if (snapshotTtl !== null && snapshotTtl > 0) {
+    return Math.trunc(snapshotTtl)
+  }
+  if (cacheCreationInputTokens1h.value > 0 && cacheCreationInputTokens5m.value === 0) {
+    return 60
+  }
+  if (cacheCreationInputTokens5m.value > 0 && cacheCreationInputTokens1h.value === 0) {
+    return 5
+  }
+  return null
+})
+
 // 统一的阶梯显示数据
 // 如果有 tiered_pricing，使用它；否则用历史价格构建单阶梯
 const displayTiers = computed(() => {
@@ -1035,10 +1219,11 @@ const displayTiers = computed(() => {
   // 否则用历史价格构建单阶梯（无上限）
   return [{
     up_to: null,
-    input_price_per_1m: detail.value.input_price_per_1m || 0,
+    input_price_per_1m: effectiveInputPricePer1M.value ?? 0,
-    output_price_per_1m: detail.value.output_price_per_1m || 0,
+    output_price_per_1m: effectiveOutputPricePer1M.value ?? 0,
-    cache_creation_price_per_1m: detail.value.cache_creation_price_per_1m,
+    cache_creation_price_per_1m: effectiveCacheCreationPricePer1M.value,
-    cache_read_price_per_1m: detail.value.cache_read_price_per_1m
+    cache_read_price_per_1m: effectiveCacheReadPricePer1M.value,
+    cache_ttl_pricing: fallbackCacheTtlPricing.value,
   }]
 })
 
@@ -1051,6 +1236,11 @@ const currentTierIndex = computed(() => {
     return detail.value.tiered_pricing.tier_index
   }
 
+  const snapshotTierIndex = getNestedNumber(billingSnapshot.value, 'tier_index')
+  if (snapshotTierIndex !== null) {
+    return Math.max(0, Math.trunc(snapshotTierIndex))
+  }
+
   // 单阶梯时默认是第0阶
   return 0
 })
@@ -1064,9 +1254,9 @@ const currentTier = computed<PricingTierLike | null>(() => {
 const cacheCreationSummaryText = computed(() => {
   if (!detail.value) return '0'
 
-  const total = detail.value.cache_creation_input_tokens || 0
+  const total = totalCacheCreationTokens.value
-  const cache5m = detail.value.cache_creation_input_tokens_5m || 0
+  const cache5m = cacheCreationInputTokens5m.value
-  const cache1h = detail.value.cache_creation_input_tokens_1h || 0
+  const cache1h = cacheCreationInputTokens1h.value
 
   if (cache5m <= 0 && cache1h <= 0) {
     return formatNumber(total)
@@ -1091,8 +1281,8 @@ const cacheCreationSplitRows = computed(() => {
     cost: number | null
   }> = []
 
-  const cache5m = detail.value.cache_creation_input_tokens_5m || 0
+  const cache5m = cacheCreationInputTokens5m.value
-  const cache1h = detail.value.cache_creation_input_tokens_1h || 0
+  const cache1h = cacheCreationInputTokens1h.value
 
   if (cache5m > 0) {
     const pricePer1M = getActiveCachePriceForTTL(5, 'cache_creation_price_per_1m')
@@ -1130,7 +1320,7 @@ const _totalInputContext = computed(() => {
 
   // 否则手动计算
   const input = detail.value.tokens?.input || detail.value.input_tokens || 0
-  const cacheCreation = detail.value.cache_creation_input_tokens || 0
+  const cacheCreation = totalCacheCreationTokens.value
   const cacheRead = detail.value.cache_read_input_tokens || 0
   return input + cacheCreation + cacheRead
 })
@@ -1138,11 +1328,10 @@ const _totalInputContext = computed(() => {
 // Token 费用总计
 const tokenCostTotal = computed(() => {
   if (!detail.value) return 0
-  const inputCost = detail.value.cost?.input || detail.value.input_cost || 0
+  return effectiveInputCost.value
-  const outputCost = detail.value.cost?.output || detail.value.output_cost || 0
+    + effectiveOutputCost.value
-  const cacheCreationCost = detail.value.cache_creation_cost || 0
+    + effectiveCacheCreationCost.value
-  const cacheReadCost = detail.value.cache_read_cost || 0
+    + effectiveCacheReadCost.value
-  return inputCost + outputCost + cacheCreationCost + cacheReadCost
 })
 
 // 按次计费费用（非视频任务时）
@@ -1150,7 +1339,7 @@ const perRequestCost = computed(() => {
   if (!detail.value) return 0
   // 视频任务的 request_cost 实际上是视频费用，不算按次
   if (detail.value.video_billing) return 0
-  return detail.value.request_cost || 0
+  return effectiveRequestCost.value
 })
 
 // 视频/图像/音频费用
@@ -1167,7 +1356,7 @@ const hasTokenCost = computed(() => {
   if (!detail.value) return false
   const inputTokens = detail.value.tokens?.input || detail.value.input_tokens || 0
   const outputTokens = detail.value.tokens?.output || detail.value.output_tokens || 0
-  const cacheCreation = detail.value.cache_creation_input_tokens || 0
+  const cacheCreation = totalCacheCreationTokens.value
   const cacheRead = detail.value.cache_read_input_tokens || 0
   return (inputTokens + outputTokens + cacheCreation + cacheRead) > 0 || tokenCostTotal.value > 0
 })
@@ -1208,12 +1397,25 @@ function getTierCachePriceForTTL(
 
   if (ttlPricing.length === 0) return fallback
 
-  const matched = ttlPricing.find((entry) => Number(entry.ttl_minutes || 0) >= ttlMinutes)
+  const matched = ttlPricing.find((entry) => Number(entry.ttl_minutes || 0) === ttlMinutes)
-    || ttlPricing[ttlPricing.length - 1]
   const price = toFiniteNumber(matched?.[priceKey])
   return price ?? fallback
 }
 
+function getTierMatchedCachePricingEntry(
+  tier: PricingTierLike | null | undefined,
+  ttlMinutes: number,
+): CacheTTLPriceEntry | null {
+  const ttlPricing = Array.isArray(tier?.cache_ttl_pricing)
+    ? tier.cache_ttl_pricing
+        .filter((entry): entry is CacheTTLPriceEntry => !!entry && typeof entry === 'object')
+        .sort((a, b) => Number(a.ttl_minutes || 0) - Number(b.ttl_minutes || 0))
+    : []
+
+  if (ttlPricing.length === 0) return null
+  return ttlPricing.find((entry) => Number(entry.ttl_minutes || 0) === ttlMinutes) ?? null
+}
+
 function hasTierCacheCreationSplitPricing(tier: PricingTierLike | null | undefined): boolean {
   const ttlPricing = Array.isArray(tier?.cache_ttl_pricing) ? tier.cache_ttl_pricing : []
   return ttlPricing.some((entry) =>
@@ -1222,6 +1424,45 @@ function hasTierCacheCreationSplitPricing(tier: PricingTierLike | null | undefin
   )
 }
 
+function formatCacheTtlLabel(ttlMinutes: number | null | undefined): string {
+  if (!ttlMinutes || ttlMinutes <= 0) return '缓存创建'
+  if (ttlMinutes >= 60) return '缓存创建(1h)'
+  if (ttlMinutes <= 5) return '缓存创建(5min)'
+  return `缓存创建(${ttlMinutes}min)`
+}
+
+function getTierActiveCacheCreationDisplay(
+  tier: PricingTierLike | null | undefined,
+): { label: string, price: number } | null {
+  if (hasTierCacheCreationSplitPricing(tier)) {
+    const activeTtl = activeCacheTtlMinutes.value
+    if (activeTtl !== null) {
+      const matchedEntry = getTierMatchedCachePricingEntry(tier, activeTtl)
+      const matchedPrice = getTierCachePriceForTTL(tier, activeTtl, 'cache_creation_price_per_1m')
+      if (matchedEntry && matchedPrice !== null) {
+        return {
+          label: formatCacheTtlLabel(matchedEntry.ttl_minutes),
+          price: matchedPrice,
+        }
+      }
+      const fallbackPrice = toFiniteNumber(tier?.cache_creation_price_per_1m)
+      if (fallbackPrice !== null) {
+        return {
+          label: formatCacheTtlLabel(activeTtl),
+          price: fallbackPrice,
+        }
+      }
+    }
+  }
+
+  const fallbackPrice = toFiniteNumber(tier?.cache_creation_price_per_1m)
+  if (fallbackPrice === null) return null
+  return {
+    label: '缓存创建',
+    price: fallbackPrice,
+  }
+}
+
 function getActiveCachePriceForTTL(
   ttlMinutes: number,
   priceKey: 'cache_creation_price_per_1m' | 'cache_read_price_per_1m',
@@ -1235,6 +1476,22 @@ function getActiveCachePriceForTTL(
   return toFiniteNumber(detail.value?.cache_read_price_per_1m)
 }
 
+function getTierActiveCacheReadPrice(tier: PricingTierLike | null | undefined): number | null {
+  const activeTtl = activeCacheTtlMinutes.value
+  if (activeTtl !== null) {
+    const matchedPrice = getTierCachePriceForTTL(tier, activeTtl, 'cache_read_price_per_1m')
+    if (matchedPrice !== null) return matchedPrice
+  }
+
+  return toFiniteNumber(tier?.cache_read_price_per_1m)
+    ?? effectiveCacheReadPricePer1M.value
+    ?? null
+}
+
+function shouldShowCacheReadPrice(tier: PricingTierLike | null | undefined): boolean {
+  return getTierActiveCacheReadPrice(tier) !== null
+}
+
 function getDefaultDataSourceForTab(tab: string): 'client' | 'provider' {
   if (!detail.value) {
     if (['request-headers', 'request-body'].includes(tab)) return 'provider'