feat(data): 废弃 usage 表 HTTP/结算列，迁移至 settlement_snapshots 与 http_audits

- 新增迁移 20260413030000：标记 billing_status、finalized_at、request_headers 等列为 DEPRECATED
- 更新 baseline_v2.sql 同步废弃注释，BASELINE_V2_CUTOFF_VERSION 升至 20260413030000
- usage/sql.rs：inline body 阈值归零，强制所有 body 走 blob 存储；upsert 时清空 legacy header/output_price 列
- 查询层优先读 usage_settlement_snapshots 的 billing_status、finalized_at、output_price_per_1m
- runtime.rs：stale usage 处理同步写入 usage_settlement_snapshots；SELECT FOR UPDATE 改为 FOR UPDATE OF usage
- 前端：PerformanceAnalysis 页面重构为实时面板，新增 prometheus 工具函数与 monitoring API

README.md

@@ -49,7 +49,7 @@ cp .env.example .env
 docker compose pull && docker compose up -d
 
 # 4. 如果后续版本包含 schema 变更，再显式执行数据库迁移
-docker compose run --rm app aether-gateway --migrate
+docker compose run --rm app --migrate
 
 # 5. 升级前备份 (可选)
 docker compose exec postgres pg_dump -U postgres aether | gzip > backup_$(date +%Y%m%d_%H%M%S).sql.gz

apps/aether-gateway/src/maintenance/runtime.rs

@@ -184,16 +184,18 @@ WHERE ledgers.billing_date = $1
 "#;
 const SELECT_STALE_PENDING_USAGE_BATCH_SQL: &str = r#"
 SELECT
-  id,
-  request_id,
-  status,
-  billing_status
+  usage.id,
+  usage.request_id,
+  usage.status,
+  COALESCE(usage_settlement_snapshots.billing_status, usage.billing_status) AS billing_status
 FROM usage
-WHERE status = ANY($1)
-  AND created_at < $2
-ORDER BY created_at ASC, id ASC
+LEFT JOIN usage_settlement_snapshots
+  ON usage_settlement_snapshots.request_id = usage.request_id
+WHERE usage.status = ANY($1)
+  AND usage.created_at < $2
+ORDER BY usage.created_at ASC, usage.id ASC
 LIMIT $3
-FOR UPDATE SKIP LOCKED
+FOR UPDATE OF usage SKIP LOCKED
 "#;
 const SELECT_COMPLETED_PENDING_REQUEST_IDS_SQL: &str = r#"
 SELECT DISTINCT request_id
@@ -222,17 +224,35 @@ SET status = 'failed',
 WHERE id = $1
 "#;
 const UPDATE_FAILED_VOID_STALE_USAGE_SQL: &str = r#"
-UPDATE usage
-SET status = 'failed',
-    status_code = 504,
-    error_message = $2,
-    billing_status = 'void',
-    finalized_at = $3,
-    total_cost_usd = 0,
-    request_cost_usd = 0,
-    actual_total_cost_usd = 0,
-    actual_request_cost_usd = 0
-WHERE id = $1
+WITH updated_usage AS (
+    UPDATE usage
+    SET status = 'failed',
+        status_code = 504,
+        error_message = $2,
+        billing_status = 'void',
+        finalized_at = $3,
+        total_cost_usd = 0,
+        request_cost_usd = 0,
+        actual_total_cost_usd = 0,
+        actual_request_cost_usd = 0
+    WHERE id = $1
+    RETURNING request_id
+)
+INSERT INTO usage_settlement_snapshots (
+    request_id,
+    billing_status,
+    finalized_at
+)
+SELECT request_id, 'void', $3
+FROM updated_usage
+ON CONFLICT (request_id)
+DO UPDATE SET
+    billing_status = EXCLUDED.billing_status,
+    finalized_at = COALESCE(
+        usage_settlement_snapshots.finalized_at,
+        EXCLUDED.finalized_at
+    ),
+    updated_at = NOW()
 "#;
 const UPDATE_RECOVERED_STREAMING_CANDIDATES_SQL: &str = r#"
 UPDATE request_candidates

apps/aether-gateway/src/maintenance/runtime/tests.rs

@@ -29,7 +29,8 @@ use super::{
     usage_cleanup_window, wallet_daily_usage_aggregation_target, AppState, DbMaintenanceRunSummary,
     FailedPendingUsageRow, GatewayDataState, ProxyUpgradeRolloutProbeConfig, StalePendingUsageRow,
     UsageCleanupSettings, DELETE_STALE_WALLET_DAILY_USAGE_LEDGERS_SQL,
-    SELECT_WALLET_DAILY_USAGE_AGGREGATION_ROWS_SQL, USAGE_CLEANUP_HOUR, USAGE_CLEANUP_MINUTE,
+    SELECT_STALE_PENDING_USAGE_BATCH_SQL, SELECT_WALLET_DAILY_USAGE_AGGREGATION_ROWS_SQL,
+    UPDATE_FAILED_VOID_STALE_USAGE_SQL, USAGE_CLEANUP_HOUR, USAGE_CLEANUP_MINUTE,
     WALLET_DAILY_USAGE_AGGREGATION_HOUR, WALLET_DAILY_USAGE_AGGREGATION_MINUTE,
 };
 
@@ -90,6 +91,15 @@ fn wallet_daily_usage_queries_use_settlement_snapshots_for_wallet_identity() {
         .contains("usage_settlement_snapshots.wallet_id = ledgers.wallet_id"));
 }
 
+#[test]
+fn pending_cleanup_queries_use_settlement_snapshots_for_billing_authority() {
+    assert!(SELECT_STALE_PENDING_USAGE_BATCH_SQL.contains("LEFT JOIN usage_settlement_snapshots"));
+    assert!(SELECT_STALE_PENDING_USAGE_BATCH_SQL
+        .contains("COALESCE(usage_settlement_snapshots.billing_status, usage.billing_status)"));
+    assert!(UPDATE_FAILED_VOID_STALE_USAGE_SQL.contains("INSERT INTO usage_settlement_snapshots"));
+    assert!(UPDATE_FAILED_VOID_STALE_USAGE_SQL.contains("billing_status = EXCLUDED.billing_status"));
+}
+
 fn sample_connected_proxy_node(
     node_id: &str,
     heartbeat_interval: i32,

crates/aether-data/bootstrap/20260413020000_baseline_v2.sql

@@ -4738,3 +4738,31 @@ COMMENT ON COLUMN public.usage.username IS
   'DEPRECATED: display cache only. Prefer join-time lookup from user/auth records. Legacy compatibility only.';
 COMMENT ON COLUMN public.usage.api_key_name IS
   'DEPRECATED: display cache only. Prefer join-time lookup from API key records. Legacy compatibility only.';
+COMMENT ON COLUMN public.usage.billing_status IS
+  'DEPRECATED: authoritative owner moved to public.usage_settlement_snapshots.billing_status. Compatibility/index mirror only; do not write new values.';
+COMMENT ON COLUMN public.usage.finalized_at IS
+  'DEPRECATED: authoritative owner moved to public.usage_settlement_snapshots.finalized_at. Compatibility/index mirror only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.request_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.provider_request_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.response_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.client_response_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.provider_request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.provider_request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.client_response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.client_response_body_ref. Legacy compatibility only; do not write new values.';

crates/aether-data/migrations/20260413030000_deprecate_usage_legacy_http_columns.sql

@@ -0,0 +1,28 @@
+COMMENT ON COLUMN public.usage.billing_status IS
+  'DEPRECATED: authoritative owner moved to public.usage_settlement_snapshots.billing_status. Compatibility/index mirror only; do not write new values.';
+COMMENT ON COLUMN public.usage.finalized_at IS
+  'DEPRECATED: authoritative owner moved to public.usage_settlement_snapshots.finalized_at. Compatibility/index mirror only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.request_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.provider_request_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.response_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_headers IS
+  'DEPRECATED: HTTP audit owner moved to public.usage_http_audits.client_response_headers. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.request_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.provider_request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.provider_request_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.provider_request_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.response_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_body IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.client_response_body_ref. Legacy compatibility only; do not write new values.';
+COMMENT ON COLUMN public.usage.client_response_body_compressed IS
+  'DEPRECATED: HTTP body owner moved to public.usage_body_blobs plus public.usage_http_audits.client_response_body_ref. Legacy compatibility only; do not write new values.';

crates/aether-data/src/migrate.rs

@@ -8,7 +8,7 @@ use tracing::{error, info, warn};
 
 static MIGRATOR: Migrator = sqlx::migrate!("./migrations");
 static BASELINE_V2_SQL: &str = include_str!("../bootstrap/20260413020000_baseline_v2.sql");
-const BASELINE_V2_CUTOFF_VERSION: i64 = 20260413020000;
+const BASELINE_V2_CUTOFF_VERSION: i64 = 20260413030000;
 const MIGRATIONS_TABLE_EXISTS_SQL: &str =
     "SELECT to_regclass('public._sqlx_migrations') IS NOT NULL";
 const EMPTY_DATABASE_USER_TABLE_COUNT_SQL: &str = r#"
@@ -404,6 +404,7 @@ mod tests {
                 20260406000000,
                 20260410000000,
                 20260413020000,
+                20260413030000,
             ]
         );
     }
@@ -425,27 +426,47 @@ mod tests {
 
     #[test]
     fn deprecation_migration_and_baseline_mark_legacy_usage_columns() {
-        let migration = MIGRATOR
+        let settlement_migration = MIGRATOR
             .iter()
             .find(|migration| migration.version == 20260413020000)
             .expect("deprecation migration should be embedded");
+        let http_migration = MIGRATOR
+            .iter()
+            .find(|migration| migration.version == 20260413030000)
+            .expect("http/body deprecation migration should be embedded");
 
-        assert!(migration
+        assert!(settlement_migration
             .sql
             .contains("COMMENT ON COLUMN public.usage.output_price_per_1m"));
-        assert!(migration
+        assert!(settlement_migration
             .sql
             .contains("COMMENT ON COLUMN public.usage.wallet_id"));
-        assert!(migration
+        assert!(settlement_migration
             .sql
             .contains("COMMENT ON COLUMN public.usage.username"));
-        assert!(migration
+        assert!(settlement_migration
             .sql
             .contains("COMMENT ON COLUMN public.usage.api_key_name"));
+        assert!(http_migration
+            .sql
+            .contains("COMMENT ON COLUMN public.usage.request_headers"));
+        assert!(http_migration
+            .sql
+            .contains("COMMENT ON COLUMN public.usage.request_body"));
+        assert!(http_migration
+            .sql
+            .contains("COMMENT ON COLUMN public.usage.billing_status"));
+        assert!(http_migration
+            .sql
+            .contains("COMMENT ON COLUMN public.usage.finalized_at"));
         assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.output_price_per_1m"));
         assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.wallet_id"));
         assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.username"));
         assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.api_key_name"));
+        assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.request_headers"));
+        assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.request_body"));
+        assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.billing_status"));
+        assert!(BASELINE_V2_SQL.contains("COMMENT ON COLUMN public.usage.finalized_at"));
     }
 
     #[test]
@@ -472,7 +493,10 @@ mod tests {
             .map(|migration| migration.version)
             .collect::<Vec<_>>();
 
-        assert_eq!(pending_versions, vec![20260410000000, 20260413020000]);
+        assert_eq!(
+            pending_versions,
+            vec![20260410000000, 20260413020000, 20260413030000]
+        );
     }
 
     #[test]

crates/aether-data/src/repository/settlement/sql.rs

@@ -10,7 +10,7 @@ const FIND_USAGE_FOR_SETTLEMENT_SQL: &str = r#"
 SELECT
   usage_record.request_id,
   COALESCE(usage_settlement_snapshots.wallet_id, usage_record.wallet_id) AS wallet_id,
-  usage_record.billing_status,
+  COALESCE(usage_settlement_snapshots.billing_status, usage_record.billing_status) AS billing_status,
   COALESCE(
     CAST(usage_settlement_snapshots.wallet_balance_before AS DOUBLE PRECISION),
     CAST(usage_record.wallet_balance_before AS DOUBLE PRECISION)
@@ -369,6 +369,7 @@ RETURNING CAST(monthly_used_usd AS DOUBLE PRECISION) AS monthly_used_usd
                         }
                     }
 
+                    sync_usage_settlement_snapshot(&mut **tx, &settlement).await?;
                     sqlx::query(FINALIZE_USAGE_BILLING_SQL)
                         .bind(&input.request_id)
                         .bind(final_billing_status)
@@ -376,7 +377,6 @@ RETURNING CAST(monthly_used_usd AS DOUBLE PRECISION) AS monthly_used_usd
                         .execute(&mut **tx)
                         .await
                         .map_postgres_err()?;
-                    sync_usage_settlement_snapshot(&mut **tx, &settlement).await?;
 
                     Ok(Some(settlement))
                 })
@@ -397,7 +397,9 @@ mod tests {
         assert!(
             super::FIND_USAGE_FOR_SETTLEMENT_SQL.contains("LEFT JOIN usage_settlement_snapshots")
         );
-        assert!(super::FIND_USAGE_FOR_SETTLEMENT_SQL.contains("COALESCE("));
+        assert!(super::FIND_USAGE_FOR_SETTLEMENT_SQL.contains(
+            "COALESCE(usage_settlement_snapshots.billing_status, usage_record.billing_status)"
+        ));
         assert!(super::FIND_USAGE_FOR_SETTLEMENT_SQL.contains("FOR UPDATE OF usage_record"));
     }
 

crates/aether-data/src/repository/usage/sql.rs

@@ -18,7 +18,9 @@ use super::{
 use crate::postgres::PostgresTransactionRunner;
 use crate::{error::SqlxResultExt, DataLayerError};
 
-const MAX_INLINE_USAGE_BODY_BYTES: usize = 16 * 1024;
+// Legacy inline body columns on public.usage are deprecated. Keep the threshold at zero so
+// newly captured bodies always spill to usage_body_blobs and resolve through usage_http_audits.
+const MAX_INLINE_USAGE_BODY_BYTES: usize = 0;
 const FIND_USAGE_BODY_BLOB_BY_REF_SQL: &str =
     r#"SELECT payload_gzip FROM usage_body_blobs WHERE body_ref = $1 LIMIT 1"#;
 const UPSERT_USAGE_BODY_BLOB_SQL: &str = r#"
@@ -258,7 +260,10 @@ SELECT
   COALESCE("usage".cache_read_input_tokens, 0) AS cache_read_input_tokens,
   COALESCE(CAST("usage".cache_creation_cost_usd AS DOUBLE PRECISION), 0) AS cache_creation_cost_usd,
   COALESCE(CAST("usage".cache_read_cost_usd AS DOUBLE PRECISION), 0) AS cache_read_cost_usd,
-  CAST("usage".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m,
+  COALESCE(
+    CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION),
+    CAST("usage".output_price_per_1m AS DOUBLE PRECISION)
+  ) AS output_price_per_1m,
   COALESCE(CAST("usage".total_cost_usd AS DOUBLE PRECISION), 0) AS total_cost_usd,
   COALESCE(CAST("usage".actual_total_cost_usd AS DOUBLE PRECISION), 0) AS actual_total_cost_usd,
   "usage".status_code,
@@ -310,9 +315,19 @@ SELECT
   CAST(usage_settlement_snapshots.price_per_request AS DOUBLE PRECISION) AS settlement_price_per_request,
   CAST(EXTRACT(EPOCH FROM "usage".created_at) AS BIGINT) AS created_at_unix_ms,
   CAST(
-    EXTRACT(EPOCH FROM COALESCE("usage".finalized_at, "usage".created_at)) AS BIGINT
+    EXTRACT(
+      EPOCH FROM COALESCE(
+        usage_settlement_snapshots.finalized_at,
+        "usage".finalized_at,
+        "usage".created_at
+      )
+    ) AS BIGINT
   ) AS updated_at_unix_secs,
-  CAST(EXTRACT(EPOCH FROM "usage".finalized_at) AS BIGINT) AS finalized_at_unix_secs
+  CAST(
+    EXTRACT(
+      EPOCH FROM COALESCE(usage_settlement_snapshots.finalized_at, "usage".finalized_at)
+    ) AS BIGINT
+  ) AS finalized_at_unix_secs
 FROM "usage"
 LEFT JOIN usage_http_audits
   ON usage_http_audits.request_id = "usage".request_id
@@ -356,7 +371,10 @@ SELECT
   COALESCE("usage".cache_read_input_tokens, 0) AS cache_read_input_tokens,
   COALESCE(CAST("usage".cache_creation_cost_usd AS DOUBLE PRECISION), 0) AS cache_creation_cost_usd,
   COALESCE(CAST("usage".cache_read_cost_usd AS DOUBLE PRECISION), 0) AS cache_read_cost_usd,
-  CAST("usage".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m,
+  COALESCE(
+    CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION),
+    CAST("usage".output_price_per_1m AS DOUBLE PRECISION)
+  ) AS output_price_per_1m,
   COALESCE(CAST("usage".total_cost_usd AS DOUBLE PRECISION), 0) AS total_cost_usd,
   COALESCE(CAST("usage".actual_total_cost_usd AS DOUBLE PRECISION), 0) AS actual_total_cost_usd,
   "usage".status_code,
@@ -408,9 +426,19 @@ SELECT
   CAST(usage_settlement_snapshots.price_per_request AS DOUBLE PRECISION) AS settlement_price_per_request,
   CAST(EXTRACT(EPOCH FROM "usage".created_at) AS BIGINT) AS created_at_unix_ms,
   CAST(
-    EXTRACT(EPOCH FROM COALESCE("usage".finalized_at, "usage".created_at)) AS BIGINT
+    EXTRACT(
+      EPOCH FROM COALESCE(
+        usage_settlement_snapshots.finalized_at,
+        "usage".finalized_at,
+        "usage".created_at
+      )
+    ) AS BIGINT
   ) AS updated_at_unix_secs,
-  CAST(EXTRACT(EPOCH FROM "usage".finalized_at) AS BIGINT) AS finalized_at_unix_secs
+  CAST(
+    EXTRACT(
+      EPOCH FROM COALESCE(usage_settlement_snapshots.finalized_at, "usage".finalized_at)
+    ) AS BIGINT
+  ) AS finalized_at_unix_secs
 FROM "usage"
 LEFT JOIN usage_http_audits
   ON usage_http_audits.request_id = "usage".request_id
@@ -505,7 +533,10 @@ SELECT
   COALESCE("usage".cache_read_input_tokens, 0) AS cache_read_input_tokens,
   COALESCE(CAST("usage".cache_creation_cost_usd AS DOUBLE PRECISION), 0) AS cache_creation_cost_usd,
   COALESCE(CAST("usage".cache_read_cost_usd AS DOUBLE PRECISION), 0) AS cache_read_cost_usd,
-  CAST("usage".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m,
+  COALESCE(
+    CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION),
+    CAST("usage".output_price_per_1m AS DOUBLE PRECISION)
+  ) AS output_price_per_1m,
   COALESCE(CAST("usage".total_cost_usd AS DOUBLE PRECISION), 0) AS total_cost_usd,
   COALESCE(CAST("usage".actual_total_cost_usd AS DOUBLE PRECISION), 0) AS actual_total_cost_usd,
   "usage".status_code,
@@ -551,9 +582,19 @@ SELECT
   CAST(usage_settlement_snapshots.price_per_request AS DOUBLE PRECISION) AS settlement_price_per_request,
   CAST(EXTRACT(EPOCH FROM "usage".created_at) AS BIGINT) AS created_at_unix_ms,
   CAST(
-    EXTRACT(EPOCH FROM COALESCE("usage".finalized_at, "usage".created_at)) AS BIGINT
+    EXTRACT(
+      EPOCH FROM COALESCE(
+        usage_settlement_snapshots.finalized_at,
+        "usage".finalized_at,
+        "usage".created_at
+      )
+    ) AS BIGINT
   ) AS updated_at_unix_secs,
-  CAST(EXTRACT(EPOCH FROM "usage".finalized_at) AS BIGINT) AS finalized_at_unix_secs
+  CAST(
+    EXTRACT(
+      EPOCH FROM COALESCE(usage_settlement_snapshots.finalized_at, "usage".finalized_at)
+    ) AS BIGINT
+  ) AS finalized_at_unix_secs
 FROM "usage"
 LEFT JOIN usage_settlement_snapshots
   ON usage_settlement_snapshots.request_id = "usage".request_id
@@ -591,7 +632,10 @@ SELECT
   COALESCE("usage".cache_read_input_tokens, 0) AS cache_read_input_tokens,
   COALESCE(CAST("usage".cache_creation_cost_usd AS DOUBLE PRECISION), 0) AS cache_creation_cost_usd,
   COALESCE(CAST("usage".cache_read_cost_usd AS DOUBLE PRECISION), 0) AS cache_read_cost_usd,
-  CAST("usage".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m,
+  COALESCE(
+    CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION),
+    CAST("usage".output_price_per_1m AS DOUBLE PRECISION)
+  ) AS output_price_per_1m,
   COALESCE(CAST("usage".total_cost_usd AS DOUBLE PRECISION), 0) AS total_cost_usd,
   COALESCE(CAST("usage".actual_total_cost_usd AS DOUBLE PRECISION), 0) AS actual_total_cost_usd,
   "usage".status_code,
@@ -637,9 +681,19 @@ SELECT
   CAST(usage_settlement_snapshots.price_per_request AS DOUBLE PRECISION) AS settlement_price_per_request,
   CAST(EXTRACT(EPOCH FROM "usage".created_at) AS BIGINT) AS created_at_unix_ms,
   CAST(
-    EXTRACT(EPOCH FROM COALESCE("usage".finalized_at, "usage".created_at)) AS BIGINT
+    EXTRACT(
+      EPOCH FROM COALESCE(
+        usage_settlement_snapshots.finalized_at,
+        "usage".finalized_at,
+        "usage".created_at
+      )
+    ) AS BIGINT
   ) AS updated_at_unix_secs,
-  CAST(EXTRACT(EPOCH FROM "usage".finalized_at) AS BIGINT) AS finalized_at_unix_secs
+  CAST(
+    EXTRACT(
+      EPOCH FROM COALESCE(usage_settlement_snapshots.finalized_at, "usage".finalized_at)
+    ) AS BIGINT
+  ) AS finalized_at_unix_secs
 FROM "usage"
 LEFT JOIN usage_settlement_snapshots
   ON usage_settlement_snapshots.request_id = "usage".request_id
@@ -792,7 +846,7 @@ DO UPDATE SET
   cache_read_input_tokens = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.cache_read_input_tokens, "usage".cache_read_input_tokens) ELSE "usage".cache_read_input_tokens END,
   cache_creation_cost_usd = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.cache_creation_cost_usd, "usage".cache_creation_cost_usd) ELSE "usage".cache_creation_cost_usd END,
   cache_read_cost_usd = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.cache_read_cost_usd, "usage".cache_read_cost_usd) ELSE "usage".cache_read_cost_usd END,
-  output_price_per_1m = "usage".output_price_per_1m,
+  output_price_per_1m = NULL,
   total_cost_usd = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.total_cost_usd, "usage".total_cost_usd) ELSE "usage".total_cost_usd END,
   actual_total_cost_usd = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.actual_total_cost_usd, "usage".actual_total_cost_usd) ELSE "usage".actual_total_cost_usd END,
   status_code = CASE WHEN "usage".billing_status = 'pending' THEN CASE
@@ -811,7 +865,7 @@ DO UPDATE SET
   first_byte_time_ms = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.first_byte_time_ms, "usage".first_byte_time_ms) ELSE "usage".first_byte_time_ms END,
   status = CASE WHEN "usage".billing_status = 'pending' THEN EXCLUDED.status ELSE "usage".status END,
   billing_status = CASE WHEN "usage".billing_status = 'pending' THEN EXCLUDED.billing_status ELSE "usage".billing_status END,
-  request_headers = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.request_headers, "usage".request_headers) ELSE "usage".request_headers END,
+  request_headers = NULL,
   request_body = CASE WHEN "usage".billing_status = 'pending' THEN CASE
     WHEN EXCLUDED.request_body_compressed IS NOT NULL OR $56 THEN NULL
     ELSE COALESCE(EXCLUDED.request_body, "usage".request_body)
@@ -820,7 +874,7 @@ DO UPDATE SET
     WHEN EXCLUDED.request_body IS NOT NULL OR $56 THEN NULL
     ELSE COALESCE(EXCLUDED.request_body_compressed, "usage".request_body_compressed)
   END ELSE "usage".request_body_compressed END,
-  provider_request_headers = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.provider_request_headers, "usage".provider_request_headers) ELSE "usage".provider_request_headers END,
+  provider_request_headers = NULL,
   provider_request_body = CASE WHEN "usage".billing_status = 'pending' THEN CASE
     WHEN EXCLUDED.provider_request_body_compressed IS NOT NULL OR $57 THEN NULL
     ELSE COALESCE(EXCLUDED.provider_request_body, "usage".provider_request_body)
@@ -829,7 +883,7 @@ DO UPDATE SET
     WHEN EXCLUDED.provider_request_body IS NOT NULL OR $57 THEN NULL
     ELSE COALESCE(EXCLUDED.provider_request_body_compressed, "usage".provider_request_body_compressed)
   END ELSE "usage".provider_request_body_compressed END,
-  response_headers = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.response_headers, "usage".response_headers) ELSE "usage".response_headers END,
+  response_headers = NULL,
   response_body = CASE WHEN "usage".billing_status = 'pending' THEN CASE
     WHEN EXCLUDED.response_body_compressed IS NOT NULL OR $58 THEN NULL
     ELSE COALESCE(EXCLUDED.response_body, "usage".response_body)
@@ -838,7 +892,7 @@ DO UPDATE SET
     WHEN EXCLUDED.response_body IS NOT NULL OR $58 THEN NULL
     ELSE COALESCE(EXCLUDED.response_body_compressed, "usage".response_body_compressed)
   END ELSE "usage".response_body_compressed END,
-  client_response_headers = CASE WHEN "usage".billing_status = 'pending' THEN COALESCE(EXCLUDED.client_response_headers, "usage".client_response_headers) ELSE "usage".client_response_headers END,
+  client_response_headers = NULL,
   client_response_body = CASE WHEN "usage".billing_status = 'pending' THEN CASE
     WHEN EXCLUDED.client_response_body_compressed IS NOT NULL OR $59 THEN NULL
     ELSE COALESCE(EXCLUDED.client_response_body, "usage".client_response_body)
@@ -1550,6 +1604,7 @@ impl SqlxUsageReadRepository {
                     stored.execution_path = routing_snapshot.execution_path.clone();
                     stored.local_execution_runtime_miss_reason =
                         routing_snapshot.local_execution_runtime_miss_reason.clone();
+                    stored.output_price_per_1m = settlement_pricing_snapshot.output_price_per_1m;
                     stored.request_metadata = request_metadata_value;
                     Ok(stored)
                 }) as BoxFuture<'_, Result<StoredRequestUsageAudit, DataLayerError>>
@@ -1910,7 +1965,8 @@ struct UsageSettlementPricingSnapshot {
 
 impl UsageSettlementPricingSnapshot {
     fn any_present(&self) -> bool {
-        self.billing_snapshot_schema_version.is_some()
+        self.billing_status.is_some()
+            || self.billing_snapshot_schema_version.is_some()
             || self.billing_snapshot_status.is_some()
             || self.rate_multiplier.is_some()
             || self.is_free_tier.is_some()
@@ -2907,11 +2963,10 @@ mod tests {
                 )
             );
             assert!(sql.contains(
-                "CAST(\"usage\".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m"
+                "CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION)"
             ));
             assert!(sql.contains("EXTRACT(EPOCH FROM \"usage\".created_at)"));
-            assert!(sql.contains("COALESCE(\"usage\".finalized_at, \"usage\".created_at)"));
-            assert!(sql.contains("EXTRACT(EPOCH FROM \"usage\".finalized_at)"));
+            assert!(sql.contains("usage_settlement_snapshots.finalized_at"));
             assert!(!sql.contains("CAST(EXTRACT(EPOCH FROM created_at) AS BIGINT)"));
             assert!(!sql.contains("CAST(output_price_per_1m AS DOUBLE PRECISION)"));
         }
@@ -2955,17 +3010,11 @@ mod tests {
     }
 
     #[test]
-    fn usage_sql_keeps_list_output_price_reads_legacy_only() {
-        assert!(super::LIST_USAGE_AUDITS_PREFIX.contains(
-            "CAST(\"usage\".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m"
-        ));
-        assert!(super::LIST_RECENT_USAGE_AUDITS_PREFIX.contains(
-            "CAST(\"usage\".output_price_per_1m AS DOUBLE PRECISION) AS output_price_per_1m"
-        ));
-        assert!(!super::LIST_USAGE_AUDITS_PREFIX
-            .contains("COALESCE(\n      usage_settlement_snapshots.output_price_per_1m,"));
-        assert!(!super::LIST_RECENT_USAGE_AUDITS_PREFIX
-            .contains("COALESCE(\n      usage_settlement_snapshots.output_price_per_1m,"));
+    fn usage_sql_reads_list_output_price_from_settlement_snapshots_before_legacy_usage_column() {
+        assert!(super::LIST_USAGE_AUDITS_PREFIX
+            .contains("CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION)"));
+        assert!(super::LIST_RECENT_USAGE_AUDITS_PREFIX
+            .contains("CAST(usage_settlement_snapshots.output_price_per_1m AS DOUBLE PRECISION)"));
     }
 
     #[test]
@@ -3007,11 +3056,19 @@ mod tests {
     }
 
     #[test]
-    fn usage_sql_preserves_legacy_output_price_column_on_upsert() {
-        assert!(super::UPSERT_SQL.contains("output_price_per_1m = \"usage\".output_price_per_1m"));
+    fn usage_sql_clears_legacy_output_price_column_on_upsert() {
+        assert!(super::UPSERT_SQL.contains("output_price_per_1m = NULL"));
         assert!(include_str!("sql.rs").contains(".bind(None::<f64>)"));
     }
 
+    #[test]
+    fn usage_sql_clears_legacy_header_columns_on_upsert() {
+        assert!(super::UPSERT_SQL.contains("request_headers = NULL"));
+        assert!(super::UPSERT_SQL.contains("provider_request_headers = NULL"));
+        assert!(super::UPSERT_SQL.contains("response_headers = NULL"));
+        assert!(super::UPSERT_SQL.contains("client_response_headers = NULL"));
+    }
+
     #[test]
     fn usage_sql_detached_body_flags_clear_inline_and_compressed_columns() {
         assert!(super::UPSERT_SQL
@@ -3040,14 +3097,18 @@ mod tests {
     }
 
     #[test]
-    fn prepare_usage_body_storage_keeps_small_payloads_inline() {
+    fn prepare_usage_body_storage_detaches_small_payloads_into_blob_storage() {
         let payload = json!({"message": "hello"});
         let storage = prepare_usage_body_storage(Some(&payload)).expect("storage should serialize");
 
-        assert!(storage.detached_blob_bytes.is_none());
+        assert!(storage.inline_json.is_none());
+        let compressed = storage
+            .detached_blob_bytes
+            .as_deref()
+            .expect("small payload should now be ref-backed");
         assert_eq!(
-            storage.inline_json.as_deref(),
-            Some(payload.to_string().as_str())
+            inflate_usage_json_value(compressed).expect("payload should inflate"),
+            payload
         );
     }
 
@@ -3748,6 +3809,16 @@ mod tests {
         );
     }
 
+    #[test]
+    fn usage_settlement_pricing_snapshot_with_billing_status_only_is_still_persisted() {
+        let snapshot = UsageSettlementPricingSnapshot {
+            billing_status: Some("pending".to_string()),
+            ..UsageSettlementPricingSnapshot::default()
+        };
+
+        assert!(snapshot.any_present());
+    }
+
     #[test]
     fn attach_usage_settlement_pricing_snapshot_metadata_adds_missing_values_without_overwriting() {
         let metadata = attach_usage_settlement_pricing_snapshot_metadata(

frontend/src/api/dashboard.ts

@@ -183,6 +183,18 @@ export interface RequestDetail {
   has_response_body?: boolean
   has_client_response_body?: boolean
   metadata?: Record<string, unknown>
+  settlement?: {
+    billing_snapshot?: Record<string, unknown>
+    billing_snapshot_schema_version?: string
+    billing_snapshot_status?: string
+    rate_multiplier?: number
+    is_free_tier?: boolean
+    input_price_per_1m?: number
+    output_price_per_1m?: number
+    cache_creation_price_per_1m?: number
+    cache_read_price_per_1m?: number
+    price_per_request?: number
+  } | null
   // 阶梯计费信息
   tiered_pricing?: {
     total_input_context: number  // 总输入上下文 (input + cache_read)

frontend/src/api/monitoring.ts

@@ -0,0 +1,208 @@
+import apiClient from './client'
+import {
+  findMetricValueNumber,
+  parsePrometheusSamples,
+  sumMetricValues,
+} from '@/utils/prometheus'
+
+export interface AdminMonitoringSystemStatus {
+  timestamp: string
+  users: {
+    total: number
+    active: number
+  }
+  providers: {
+    total: number
+    active: number
+  }
+  api_keys: {
+    total: number
+    active: number
+  }
+  today_stats: {
+    requests: number
+    tokens: number
+    cost_usd: string
+  }
+  tunnel: {
+    proxy_connections: number
+    nodes: number
+    active_streams: number
+  }
+  internal_gateway: {
+    status: string
+    path_prefixes: string[]
+  }
+  recent_errors: number
+}
+
+export interface AdminMonitoringCircuitBreakerSummary {
+  state: string
+  provider_id?: string
+  provider_name?: string | null
+  key_name?: string | null
+  health_score?: number
+  consecutive_failures?: number
+  last_failure_at?: string | null
+  open_formats?: string[]
+}
+
+export interface AdminMonitoringErrorStatistics {
+  total_errors: number
+  active_keys: number
+  degraded_keys: number
+  unhealthy_keys: number
+  open_circuit_breakers: number
+  circuit_breakers: Record<string, AdminMonitoringCircuitBreakerSummary>
+}
+
+export interface AdminMonitoringRecentError {
+  error_id: string
+  error_type: string
+  operation: string
+  timestamp: string | null
+  context: {
+    request_id?: string | null
+    provider_id?: string | null
+    provider_name?: string | null
+    model?: string | null
+    api_format?: string | null
+    status_code?: number | null
+    error_message?: string | null
+  }
+}
+
+export interface AdminMonitoringResilienceStatus {
+  timestamp: string
+  health_score: number
+  status: 'healthy' | 'degraded' | 'critical' | string
+  error_statistics: AdminMonitoringErrorStatistics
+  recent_errors: AdminMonitoringRecentError[]
+  recommendations: string[]
+}
+
+export interface AdminMonitoringCircuitHistoryItem {
+  event: string
+  key_id: string
+  provider_id: string
+  provider_name?: string | null
+  key_name?: string | null
+  api_format?: string | null
+  reason?: string | null
+  recovery_seconds?: number | null
+  timestamp?: string | null
+}
+
+export interface AdminMonitoringCircuitHistoryResponse {
+  items: AdminMonitoringCircuitHistoryItem[]
+  count: number
+}
+
+export interface GatewayGateMetrics {
+  inFlight: number | null
+  availablePermits: number | null
+  highWatermark: number | null
+  rejectedTotal: number | null
+  unavailable: boolean
+}
+
+export interface GatewayFallbackMetricSummary {
+  name: string
+  label: string
+  total: number
+}
+
+export interface GatewayMetricsSummary {
+  serviceUp: number | null
+  local: GatewayGateMetrics
+  distributed: GatewayGateMetrics
+  tunnel: {
+    proxyConnections: number | null
+    nodes: number | null
+    activeStreams: number | null
+  }
+  fallbackTotal: number
+  fallbacks: GatewayFallbackMetricSummary[]
+}
+
+const FALLBACK_METRICS: Array<{ name: string; label: string }> = [
+  { name: 'decision_remote_total', label: '远端决策回退' },
+  { name: 'plan_fallback_total', label: 'Plan 回退' },
+  { name: 'control_execute_fallback_total', label: '控制执行回退' },
+  { name: 'remote_execute_emergency_total', label: '紧急远端执行' },
+  { name: 'local_execution_runtime_miss_total', label: '本地运行时缺失' },
+]
+
+function buildGateMetrics(
+  samples: ReturnType<typeof parsePrometheusSamples>,
+  gate: string
+): GatewayGateMetrics {
+  return {
+    inFlight: findMetricValueNumber(samples, 'concurrency_in_flight', { gate }),
+    availablePermits: findMetricValueNumber(samples, 'concurrency_available_permits', { gate }),
+    highWatermark: findMetricValueNumber(samples, 'concurrency_high_watermark', { gate }),
+    rejectedTotal: findMetricValueNumber(samples, 'concurrency_rejected_total', { gate }),
+    unavailable: findMetricValueNumber(samples, 'concurrency_unavailable', { gate }) === 1,
+  }
+}
+
+export function buildGatewayMetricsSummary(text: string): GatewayMetricsSummary {
+  const samples = parsePrometheusSamples(text)
+  const fallbacks = FALLBACK_METRICS.map(item => ({
+    ...item,
+    total: sumMetricValues(samples, item.name),
+  }))
+
+  return {
+    serviceUp: findMetricValueNumber(samples, 'service_up', { service: 'aether-gateway' }),
+    local: buildGateMetrics(samples, 'gateway_requests'),
+    distributed: buildGateMetrics(samples, 'gateway_requests_distributed'),
+    tunnel: {
+      proxyConnections: findMetricValueNumber(samples, 'tunnel_proxy_connections'),
+      nodes: findMetricValueNumber(samples, 'tunnel_nodes'),
+      activeStreams: findMetricValueNumber(samples, 'tunnel_active_streams'),
+    },
+    fallbackTotal: fallbacks.reduce((total, item) => total + item.total, 0),
+    fallbacks,
+  }
+}
+
+async function fetchGatewayMetricsText(): Promise<string> {
+  const response = await apiClient.get<string>('/_gateway/metrics', {
+    responseType: 'text',
+    transformResponse: [(data: string) => data],
+  })
+  return typeof response.data === 'string' ? response.data : String(response.data ?? '')
+}
+
+export const monitoringApi = {
+  async getSystemStatus(): Promise<AdminMonitoringSystemStatus> {
+    const response = await apiClient.get<AdminMonitoringSystemStatus>(
+      '/api/admin/monitoring/system-status'
+    )
+    return response.data
+  },
+
+  async getResilienceStatus(): Promise<AdminMonitoringResilienceStatus> {
+    const response = await apiClient.get<AdminMonitoringResilienceStatus>(
+      '/api/admin/monitoring/resilience-status'
+    )
+    return response.data
+  },
+
+  async getCircuitHistory(limit = 10): Promise<AdminMonitoringCircuitHistoryResponse> {
+    const response = await apiClient.get<AdminMonitoringCircuitHistoryResponse>(
+      '/api/admin/monitoring/resilience/circuit-history',
+      { params: { limit } }
+    )
+    return response.data
+  },
+
+  async getGatewayMetricsText(): Promise<string> {
+    return fetchGatewayMetricsText()
+  },
+
+  async getGatewayMetricsSummary(): Promise<GatewayMetricsSummary> {
+    return buildGatewayMetricsSummary(await fetchGatewayMetricsText())
+  },
+}

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

@@ -855,8 +855,13 @@ const traceRequestMetadata = computed<Record<string, unknown> | null>(() => {
   return meta as Record<string, unknown>
 })
 
+const settlementInfo = computed<JsonRecord | null>(() =>
+  asRecord(detail.value?.settlement ?? null),
+)
+
 const billingSnapshot = computed<JsonRecord | null>(() =>
-  asRecord(traceRequestMetadata.value?.billing_snapshot),
+  asRecord(settlementInfo.value?.billing_snapshot)
+  ?? asRecord(traceRequestMetadata.value?.billing_snapshot),
 )
 
 const billingResolvedVariables = computed<JsonRecord | null>(() =>

frontend/src/mocks/handler.ts

@@ -3016,6 +3016,194 @@ mockHandlers['GET /api/admin/stats/errors/distribution'] = async () => {
   })
 }
 
+mockHandlers['GET /api/admin/monitoring/system-status'] = async () => {
+  await delay()
+  requireAdmin()
+  return createMockResponse({
+    timestamp: new Date().toISOString(),
+    users: {
+      total: 124,
+      active: 111
+    },
+    providers: {
+      total: 18,
+      active: 15
+    },
+    api_keys: {
+      total: 263,
+      active: 241
+    },
+    today_stats: {
+      requests: 12483,
+      tokens: 48751234,
+      cost_usd: '$182.4631'
+    },
+    tunnel: {
+      proxy_connections: 28,
+      nodes: 6,
+      active_streams: 164
+    },
+    internal_gateway: {
+      status: 'rust_native_control_plane',
+      path_prefixes: ['/api/', '/v1/', '/v1beta/', '/_gateway/']
+    },
+    recent_errors: 9
+  })
+}
+
+mockHandlers['GET /api/admin/monitoring/resilience-status'] = async () => {
+  await delay()
+  requireAdmin()
+  return createMockResponse({
+    timestamp: new Date().toISOString(),
+    health_score: 86,
+    status: 'healthy',
+    error_statistics: {
+      total_errors: 14,
+      active_keys: 24,
+      degraded_keys: 3,
+      unhealthy_keys: 1,
+      open_circuit_breakers: 1,
+      circuit_breakers: {
+        'provider-key-1': {
+          state: 'open',
+          provider_id: 'provider-openai',
+          provider_name: 'OpenAI',
+          key_name: 'prod-key-a',
+          health_score: 0.42,
+          consecutive_failures: 4,
+          last_failure_at: new Date(Date.now() - 8 * 60 * 1000).toISOString(),
+          open_formats: ['openai:chat']
+        }
+      }
+    },
+    recent_errors: [
+      {
+        error_id: 'usage-request-1',
+        error_type: 'timeout',
+        operation: 'OpenAI:openai:chat',
+        timestamp: new Date(Date.now() - 3 * 60 * 1000).toISOString(),
+        context: {
+          request_id: 'req-live-001',
+          provider_id: 'provider-openai',
+          provider_name: 'OpenAI',
+          model: 'gpt-5',
+          api_format: 'openai:chat',
+          status_code: 504,
+          error_message: '上游响应超时，等待首字节超过阈值'
+        }
+      },
+      {
+        error_id: 'usage-request-2',
+        error_type: 'server_error',
+        operation: 'Anthropic:claude:chat',
+        timestamp: new Date(Date.now() - 11 * 60 * 1000).toISOString(),
+        context: {
+          request_id: 'req-live-002',
+          provider_id: 'provider-anthropic',
+          provider_name: 'Anthropic',
+          model: 'claude-sonnet-4-5',
+          api_format: 'claude:chat',
+          status_code: 502,
+          error_message: '上游返回 502 Bad Gateway'
+        }
+      }
+    ],
+    recommendations: [
+      '以下服务熔断器已打开：OpenAI/prod-key-a',
+      '建议检查最近的 timeout 与 5xx 错误峰值',
+      '当前整体健康度可接受，但需要关注单 Key 退化'
+    ]
+  })
+}
+
+mockHandlers['GET /api/admin/monitoring/resilience/circuit-history'] = async () => {
+  await delay()
+  requireAdmin()
+  return createMockResponse({
+    count: 2,
+    items: [
+      {
+        event: 'opened',
+        key_id: 'provider-key-1',
+        provider_id: 'provider-openai',
+        provider_name: 'OpenAI',
+        key_name: 'prod-key-a',
+        api_format: 'openai:chat',
+        reason: '错误率过高',
+        recovery_seconds: 300,
+        timestamp: new Date(Date.now() - 10 * 60 * 1000).toISOString()
+      },
+      {
+        event: 'half_open',
+        key_id: 'provider-key-7',
+        provider_id: 'provider-gemini',
+        provider_name: 'Gemini',
+        key_name: 'gemini-burst',
+        api_format: 'gemini:chat',
+        reason: '正在探测恢复',
+        recovery_seconds: 120,
+        timestamp: new Date(Date.now() - 22 * 60 * 1000).toISOString()
+      }
+    ]
+  })
+}
+
+mockHandlers['GET /_gateway/metrics'] = async () => {
+  await delay(60)
+  requireAdmin()
+  return createMockResponse(`# HELP aether_gateway_service_up Whether the service process is currently up.
+# TYPE aether_gateway_service_up gauge
+aether_gateway_service_up{service="aether-gateway"} 1
+# HELP aether_gateway_concurrency_in_flight Current number of in-flight operations guarded by the concurrency gate.
+# TYPE aether_gateway_concurrency_in_flight gauge
+aether_gateway_concurrency_in_flight{gate="gateway_requests"} 82
+# HELP aether_gateway_concurrency_available_permits Currently available permits for the concurrency gate.
+# TYPE aether_gateway_concurrency_available_permits gauge
+aether_gateway_concurrency_available_permits{gate="gateway_requests"} 174
+# HELP aether_gateway_concurrency_high_watermark Highest observed in-flight count for the concurrency gate.
+# TYPE aether_gateway_concurrency_high_watermark gauge
+aether_gateway_concurrency_high_watermark{gate="gateway_requests"} 121
+# HELP aether_gateway_concurrency_rejected_total Number of operations rejected by the concurrency gate.
+# TYPE aether_gateway_concurrency_rejected_total counter
+aether_gateway_concurrency_rejected_total{gate="gateway_requests"} 6
+# HELP aether_gateway_concurrency_in_flight Current number of in-flight operations guarded by the concurrency gate.
+# TYPE aether_gateway_concurrency_in_flight gauge
+aether_gateway_concurrency_in_flight{gate="gateway_requests_distributed"} 94
+# HELP aether_gateway_concurrency_available_permits Currently available permits for the concurrency gate.
+# TYPE aether_gateway_concurrency_available_permits gauge
+aether_gateway_concurrency_available_permits{gate="gateway_requests_distributed"} 418
+# HELP aether_gateway_concurrency_high_watermark Highest observed in-flight count for the concurrency gate.
+# TYPE aether_gateway_concurrency_high_watermark gauge
+aether_gateway_concurrency_high_watermark{gate="gateway_requests_distributed"} 137
+# HELP aether_gateway_concurrency_rejected_total Number of operations rejected by the concurrency gate.
+# TYPE aether_gateway_concurrency_rejected_total counter
+aether_gateway_concurrency_rejected_total{gate="gateway_requests_distributed"} 11
+# HELP aether_gateway_tunnel_proxy_connections Current number of connected proxy sockets.
+# TYPE aether_gateway_tunnel_proxy_connections gauge
+aether_gateway_tunnel_proxy_connections 28
+# HELP aether_gateway_tunnel_nodes Current number of connected logical nodes.
+# TYPE aether_gateway_tunnel_nodes gauge
+aether_gateway_tunnel_nodes 6
+# HELP aether_gateway_tunnel_active_streams Current number of active local relay streams.
+# TYPE aether_gateway_tunnel_active_streams gauge
+aether_gateway_tunnel_active_streams 164
+# HELP aether_gateway_decision_remote_total Number of requests that fell back to Python decision endpoints.
+# TYPE aether_gateway_decision_remote_total counter
+aether_gateway_decision_remote_total{route_kind="chat",reason="local_decision_miss"} 4
+aether_gateway_decision_remote_total{route_kind="responses",reason="remote_decision_miss"} 2
+# HELP aether_gateway_plan_fallback_total Number of requests that fell back to Python plan endpoints.
+# TYPE aether_gateway_plan_fallback_total counter
+aether_gateway_plan_fallback_total{route_kind="chat",reason="scheduler_decision_unsupported"} 3
+# HELP aether_gateway_control_execute_fallback_total Number of requests that fell back to Python control execution.
+# TYPE aether_gateway_control_execute_fallback_total counter
+aether_gateway_control_execute_fallback_total{route_kind="chat",reason="control_execute_emergency"} 1
+# HELP aether_gateway_remote_execute_emergency_total Number of requests that used remote emergency execution fallback.
+# TYPE aether_gateway_remote_execute_emergency_total counter
+aether_gateway_remote_execute_emergency_total{route_kind="chat",reason="control_execute_emergency"} 2
+`)
+}
+
 mockHandlers['GET /api/admin/stats/comparison'] = async () => {
   await delay()
   requireAdmin()

frontend/src/utils/__tests__/prometheus.spec.ts

@@ -0,0 +1,37 @@
+import { describe, expect, it } from 'vitest'
+import {
+  findMetricValueNumber,
+  parsePrometheusSamples,
+  sumMetricValues,
+} from '../prometheus'
+
+describe('parsePrometheusSamples', () => {
+  it('parses labeled samples and finds gate metrics by suffix name', () => {
+    const samples = parsePrometheusSamples(`
+# HELP aether_gateway_concurrency_in_flight Current number of in-flight operations.
+# TYPE aether_gateway_concurrency_in_flight gauge
+aether_gateway_concurrency_in_flight{gate="gateway_requests"} 7
+aether_gateway_concurrency_rejected_total{gate="gateway_requests"} 12
+`)
+
+    expect(
+      findMetricValueNumber(samples, 'concurrency_in_flight', {
+        gate: 'gateway_requests',
+      })
+    ).toBe(7)
+    expect(
+      findMetricValueNumber(samples, 'concurrency_rejected_total', {
+        gate: 'gateway_requests',
+      })
+    ).toBe(12)
+  })
+
+  it('sums fallback counters across labeled samples', () => {
+    const samples = parsePrometheusSamples(`
+decision_remote_total{route_kind="chat",reason="local_decision_miss"} 2
+decision_remote_total{route_kind="responses",reason="remote_decision_miss"} 3
+`)
+
+    expect(sumMetricValues(samples, 'decision_remote_total')).toBe(5)
+  })
+})

frontend/src/utils/prometheus.ts

@@ -0,0 +1,106 @@
+export interface PrometheusSample {
+  name: string
+  labels: Record<string, string>
+  value: string
+}
+
+export function parsePrometheusSamples(text: string): PrometheusSample[] {
+  return text
+    .split(/\r?\n/)
+    .map(line => line.trim())
+    .filter(line => line.length > 0 && !line.startsWith('#'))
+    .map(parsePrometheusLine)
+    .filter((sample): sample is PrometheusSample => sample !== null)
+}
+
+export function findMetricValueNumber(
+  samples: PrometheusSample[],
+  metricName: string,
+  labels: Record<string, string> = {}
+): number | null {
+  const sample = samples.find(item =>
+    metricNameMatches(item.name, metricName) && labelsMatch(item.labels, labels)
+  )
+
+  if (!sample) {
+    return null
+  }
+
+  const value = Number(sample.value)
+  return Number.isFinite(value) ? value : null
+}
+
+export function sumMetricValues(
+  samples: PrometheusSample[],
+  metricName: string
+): number {
+  return samples.reduce((total, sample) => {
+    if (!metricNameMatches(sample.name, metricName)) {
+      return total
+    }
+
+    const value = Number(sample.value)
+    return Number.isFinite(value) ? total + value : total
+  }, 0)
+}
+
+function metricNameMatches(actual: string, expected: string): boolean {
+  return actual === expected || actual.split('_').pop() === expected || actual.endsWith(`_${expected}`)
+}
+
+function labelsMatch(
+  actual: Record<string, string>,
+  expected: Record<string, string>
+): boolean {
+  return Object.entries(expected).every(([key, value]) => actual[key] === value)
+}
+
+function parsePrometheusLine(line: string): PrometheusSample | null {
+  const separatorIndex = line.lastIndexOf(' ')
+  if (separatorIndex === -1) {
+    return null
+  }
+
+  const metric = line.slice(0, separatorIndex).trim()
+  const value = line.slice(separatorIndex + 1).trim()
+  if (!metric || !value) {
+    return null
+  }
+
+  const labelStart = metric.indexOf('{')
+  if (labelStart === -1 || !metric.endsWith('}')) {
+    return {
+      name: metric,
+      labels: {},
+      value,
+    }
+  }
+
+  return {
+    name: metric.slice(0, labelStart),
+    labels: parseLabels(metric.slice(labelStart + 1, -1)),
+    value,
+  }
+}
+
+function parseLabels(raw: string): Record<string, string> {
+  const labels: Record<string, string> = {}
+  const pattern = /([^=,\s]+)="((?:\\.|[^"])*)"/g
+
+  for (const match of raw.matchAll(pattern)) {
+    const [, key, value] = match
+    if (!key) {
+      continue
+    }
+    labels[key] = unescapePrometheusLabel(value ?? '')
+  }
+
+  return labels
+}
+
+function unescapePrometheusLabel(value: string): string {
+  return value
+    .replace(/\\"/g, '"')
+    .replace(/\\n/g, '\n')
+    .replace(/\\\\/g, '\\')
+}

frontend/src/views/admin/PerformanceAnalysis.vue

@@ -1,18 +1,478 @@
 <template>
   <div class="space-y-6 px-4 sm:px-6 lg:px-0">
-    <div class="flex flex-col sm:flex-row sm:items-center sm:justify-between gap-3">
+    <div class="flex flex-col gap-3 xl:flex-row xl:items-center xl:justify-between">
       <div>
         <h1 class="text-lg font-semibold">
           性能分析
         </h1>
         <p class="text-xs text-muted-foreground">
-          延迟分布与错误统计
+          实时性能监控与历史延迟趋势
         </p>
       </div>
-      <TimeRangePicker v-model="timeRange" />
+      <div class="flex flex-wrap items-center gap-2">
+        <Badge variant="outline">
+          实时 10s 刷新
+        </Badge>
+        <span class="text-xs text-muted-foreground">
+          上次更新 {{ liveLastUpdatedLabel }}
+        </span>
+        <RefreshButton
+          :loading="isRefreshing"
+          title="刷新实时与历史性能数据"
+          @click="handleManualRefresh"
+        />
+        <TimeRangePicker v-model="timeRange" />
+      </div>
     </div>
 
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
+    <Card class="overflow-hidden">
+      <div class="border-b border-border/70 bg-muted/20 px-4 py-3 sm:px-5">
+        <div class="flex flex-col gap-2 sm:flex-row sm:items-center sm:justify-between">
+          <div>
+            <h2 class="text-sm font-semibold">
+              实时性能面板
+            </h2>
+            <p class="text-xs text-muted-foreground">
+              聚合系统状态、并发门、Tunnel 与 fallback 指标
+            </p>
+          </div>
+          <div class="flex flex-wrap items-center gap-2">
+            <Badge :variant="healthStatusVariant">
+              {{ healthStatusText }}
+            </Badge>
+            <Badge variant="outline">
+              {{ metricsAvailabilityText }}
+            </Badge>
+          </div>
+        </div>
+      </div>
+
+      <div class="p-4 sm:p-5">
+        <div
+          v-if="liveLoading && !liveReady"
+          class="py-6"
+        >
+          <LoadingState message="加载实时性能数据中" />
+        </div>
+
+        <div
+          v-else-if="!liveReady"
+          class="rounded-xl border border-dashed border-border/70 bg-muted/15 px-4 py-6 text-sm text-muted-foreground"
+        >
+          实时性能数据暂不可用，请稍后重试。
+        </div>
+
+        <div
+          v-else
+          class="space-y-4"
+        >
+          <div
+            v-if="liveLoadError"
+            class="rounded-lg border border-yellow-300/70 bg-yellow-50/80 px-3 py-2 text-xs text-yellow-900 dark:border-yellow-900/60 dark:bg-yellow-950/30 dark:text-yellow-100"
+          >
+            {{ liveLoadError }}
+          </div>
+
+          <div class="grid grid-cols-1 gap-3 sm:grid-cols-2 xl:grid-cols-3 2xl:grid-cols-6">
+            <div
+              v-for="card in liveSummaryCards"
+              :key="card.title"
+              class="rounded-xl border border-border/70 bg-card/70 px-4 py-3"
+            >
+              <div class="flex items-center justify-between gap-3">
+                <span class="text-xs text-muted-foreground">{{ card.title }}</span>
+                <component
+                  :is="card.icon"
+                  class="h-4 w-4"
+                  :class="card.iconClass"
+                />
+              </div>
+              <div class="mt-3 text-2xl font-semibold tracking-tight">
+                {{ card.value }}
+              </div>
+              <div class="mt-2 text-xs text-muted-foreground">
+                {{ card.hint }}
+              </div>
+            </div>
+          </div>
+
+          <div class="grid grid-cols-1 gap-4 xl:grid-cols-3">
+            <div class="grid grid-cols-1 gap-4 lg:grid-cols-2 xl:col-span-2">
+              <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+                <div class="flex items-center justify-between gap-3">
+                  <h3 class="text-sm font-semibold">
+                    本地并发门
+                  </h3>
+                  <Badge variant="outline">
+                    gateway_requests
+                  </Badge>
+                </div>
+                <div class="mt-4 grid grid-cols-2 gap-3 text-sm">
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      In Flight
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.local.inFlight) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Available
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.local.availablePermits) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      High Watermark
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.local.highWatermark) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Rejected Total
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.local.rejectedTotal) }}
+                    </div>
+                  </div>
+                </div>
+              </section>
+
+              <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+                <div class="flex items-center justify-between gap-3">
+                  <h3 class="text-sm font-semibold">
+                    分布式并发门
+                  </h3>
+                  <Badge :variant="distributedGateVariant">
+                    {{ distributedGateText }}
+                  </Badge>
+                </div>
+                <div class="mt-4 grid grid-cols-2 gap-3 text-sm">
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      In Flight
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.distributed.inFlight) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Available
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.distributed.availablePermits) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      High Watermark
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.distributed.highWatermark) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Rejected Total
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(gatewayMetrics?.distributed.rejectedTotal) }}
+                    </div>
+                  </div>
+                </div>
+                <p
+                  v-if="gatewayMetrics?.distributed.unavailable"
+                  class="mt-3 text-xs text-yellow-700 dark:text-yellow-300"
+                >
+                  Redis 分布式并发快照当前不可用，需要检查 gate 后端连接。
+                </p>
+              </section>
+
+              <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+                <div class="flex items-center justify-between gap-3">
+                  <h3 class="text-sm font-semibold">
+                    Tunnel / 代理
+                  </h3>
+                  <Badge variant="outline">
+                    实时连接
+                  </Badge>
+                </div>
+                <div class="mt-4 grid grid-cols-2 gap-3 text-sm">
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Nodes
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(currentTunnelNodes) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Proxy Connections
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(currentProxyConnections) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Active Streams
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(currentActiveStreams) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Service Up
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ gatewayMetrics?.serviceUp === 1 ? '在线' : '未知' }}
+                    </div>
+                  </div>
+                </div>
+              </section>
+
+              <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+                <div class="flex items-center justify-between gap-3">
+                  <h3 class="text-sm font-semibold">
+                    今日请求概况
+                  </h3>
+                  <Badge variant="outline">
+                    {{ systemStatus?.internal_gateway.status || 'gateway' }}
+                  </Badge>
+                </div>
+                <div class="mt-4 grid grid-cols-2 gap-3 text-sm">
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Requests
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatMetricNumber(systemStatus?.today_stats.requests) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Tokens
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ formatTokens(systemStatus?.today_stats.tokens) }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Cost
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ systemStatus?.today_stats.cost_usd || '-' }}
+                    </div>
+                  </div>
+                  <div>
+                    <div class="text-xs text-muted-foreground">
+                      Active Providers / Keys
+                    </div>
+                    <div class="mt-1 text-lg font-semibold">
+                      {{ providerAndKeySummary }}
+                    </div>
+                  </div>
+                </div>
+              </section>
+            </div>
+
+            <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+              <div class="flex items-center justify-between gap-3">
+                <h3 class="text-sm font-semibold">
+                  Fallback 统计
+                </h3>
+                <span class="text-xs text-muted-foreground">
+                  总计 {{ formatMetricNumber(gatewayMetrics?.fallbackTotal) }}
+                </span>
+              </div>
+
+              <div
+                v-if="!fallbackRows.length"
+                class="mt-4 rounded-lg border border-dashed border-border/70 px-3 py-4 text-sm text-muted-foreground"
+              >
+                当前没有记录到 fallback 计数。
+              </div>
+
+              <div
+                v-else
+                class="mt-4 space-y-3"
+              >
+                <div
+                  v-for="item in fallbackRows"
+                  :key="item.name"
+                  class="rounded-lg border border-border/60 bg-background/50 px-3 py-3"
+                >
+                  <div class="flex items-center justify-between gap-3">
+                    <span class="text-sm font-medium">{{ item.label }}</span>
+                    <span class="text-sm font-semibold">{{ formatMetricNumber(item.total) }}</span>
+                  </div>
+                  <div class="mt-2 h-2 overflow-hidden rounded-full bg-muted">
+                    <div
+                      class="h-full rounded-full bg-primary/80"
+                      :style="{ width: `${Math.max(item.ratio, 8)}%` }"
+                    />
+                  </div>
+                </div>
+              </div>
+            </section>
+          </div>
+
+          <div class="grid grid-cols-1 gap-4 xl:grid-cols-2">
+            <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+              <div class="flex items-center justify-between gap-3">
+                <h3 class="text-sm font-semibold">
+                  最近错误
+                </h3>
+                <span class="text-xs text-muted-foreground">
+                  {{ formatMetricNumber(resilienceStatus?.error_statistics.total_errors) }} / 24h
+                </span>
+              </div>
+
+              <div
+                v-if="!recentErrors.length"
+                class="mt-4 rounded-lg border border-dashed border-border/70 px-3 py-4 text-sm text-muted-foreground"
+              >
+                当前没有最近错误。
+              </div>
+
+              <div
+                v-else
+                class="mt-4 space-y-3"
+              >
+                <article
+                  v-for="item in recentErrors"
+                  :key="item.error_id"
+                  class="rounded-lg border border-border/60 bg-background/50 px-3 py-3"
+                >
+                  <div class="flex items-start justify-between gap-4">
+                    <div>
+                      <div class="text-sm font-medium">
+                        {{ item.error_type }}
+                      </div>
+                      <div class="mt-1 text-xs text-muted-foreground">
+                        {{ item.operation }}
+                      </div>
+                    </div>
+                    <span class="shrink-0 text-xs text-muted-foreground">
+                      {{ formatDate(item.timestamp) }}
+                    </span>
+                  </div>
+
+                  <div class="mt-2 flex flex-wrap gap-2">
+                    <Badge variant="outline">
+                      HTTP {{ item.context.status_code ?? '-' }}
+                    </Badge>
+                    <Badge variant="outline">
+                      {{ item.context.provider_name || item.context.provider_id || '未知 Provider' }}
+                    </Badge>
+                    <Badge variant="outline">
+                      {{ item.context.api_format || item.context.model || '未知格式' }}
+                    </Badge>
+                  </div>
+
+                  <p
+                    v-if="item.context.error_message"
+                    class="mt-2 break-words text-xs text-muted-foreground"
+                  >
+                    {{ item.context.error_message }}
+                  </p>
+                </article>
+              </div>
+            </section>
+
+            <section class="rounded-xl border border-border/70 bg-card/60 p-4">
+              <div class="flex items-center justify-between gap-3">
+                <h3 class="text-sm font-semibold">
+                  熔断历史与建议
+                </h3>
+                <span class="text-xs text-muted-foreground">
+                  开路 {{ formatMetricNumber(resilienceStatus?.error_statistics.open_circuit_breakers) }}
+                </span>
+              </div>
+
+              <div
+                v-if="!circuitHistory.length"
+                class="mt-4 rounded-lg border border-dashed border-border/70 px-3 py-4 text-sm text-muted-foreground"
+              >
+                当前没有熔断事件。
+              </div>
+
+              <div
+                v-else
+                class="mt-4 space-y-3"
+              >
+                <article
+                  v-for="item in circuitHistory"
+                  :key="`${item.key_id}-${item.api_format}-${item.timestamp}`"
+                  class="rounded-lg border border-border/60 bg-background/50 px-3 py-3"
+                >
+                  <div class="flex items-start justify-between gap-4">
+                    <div>
+                      <div class="flex flex-wrap items-center gap-2">
+                        <span class="text-sm font-medium">
+                          {{ item.provider_name || item.provider_id }}
+                        </span>
+                        <Badge :variant="item.event === 'opened' ? 'destructive' : 'warning'">
+                          {{ item.event === 'opened' ? '已打开' : '半开' }}
+                        </Badge>
+                      </div>
+                      <div class="mt-1 text-xs text-muted-foreground">
+                        {{ item.key_name || item.key_id }} · {{ item.api_format || '未知格式' }}
+                      </div>
+                    </div>
+                    <span class="shrink-0 text-xs text-muted-foreground">
+                      {{ formatDate(item.timestamp) }}
+                    </span>
+                  </div>
+
+                  <div class="mt-2 text-xs text-muted-foreground">
+                    原因：{{ item.reason || '未提供' }}
+                  </div>
+                  <div class="mt-1 text-xs text-muted-foreground">
+                    恢复窗口：{{ item.recovery_seconds != null ? `${item.recovery_seconds}s` : '-' }}
+                  </div>
+                </article>
+              </div>
+
+              <div class="mt-4 border-t border-border/70 pt-4">
+                <h4 class="text-xs font-semibold uppercase tracking-wide text-muted-foreground">
+                  建议
+                </h4>
+                <ul
+                  v-if="resilienceRecommendations.length"
+                  class="mt-3 space-y-2 text-sm"
+                >
+                  <li
+                    v-for="item in resilienceRecommendations"
+                    :key="item"
+                    class="rounded-lg border border-border/60 bg-background/50 px-3 py-2"
+                  >
+                    {{ item }}
+                  </li>
+                </ul>
+                <div
+                  v-else
+                  class="mt-3 text-sm text-muted-foreground"
+                >
+                  当前没有额外运维建议。
+                </div>
+              </div>
+            </section>
+          </div>
+        </div>
+      </div>
+    </Card>
+
+    <div class="grid grid-cols-1 gap-4 lg:grid-cols-2">
       <Card class="p-4">
         <PercentileChart
           title="响应延迟百分位"
@@ -31,7 +491,7 @@
       </Card>
     </div>
 
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
+    <div class="grid grid-cols-1 gap-4 lg:grid-cols-2">
       <Card class="p-4">
         <ErrorDistributionChart
           title="错误分布"
@@ -39,7 +499,7 @@
           :loading="errorLoading"
         />
       </Card>
-      <Card class="p-4 space-y-3">
+      <Card class="space-y-3 p-4">
         <h3 class="text-sm font-semibold">
           错误趋势
         </h3>
@@ -58,7 +518,7 @@
       </Card>
     </div>
 
-    <Card class="p-4 space-y-3">
+    <Card class="space-y-3 p-4">
       <h3 class="text-sm font-semibold">
         提供商健康度
       </h3>
@@ -69,39 +529,68 @@
         <LoadingState />
       </div>
       <div
-        v-else
-        class="grid grid-cols-1 sm:grid-cols-2 lg:grid-cols-3 gap-3 text-sm"
+        v-else-if="providerStatus.length"
+        class="grid grid-cols-1 gap-3 text-sm sm:grid-cols-2 lg:grid-cols-3"
       >
         <div
           v-for="provider in providerStatus"
           :key="provider.name"
-          class="p-3 border rounded-lg"
+          class="rounded-lg border p-3"
         >
           <div class="flex items-center justify-between">
             <span class="font-medium">{{ provider.name }}</span>
             <span class="text-xs text-muted-foreground">{{ provider.requests }} 请求</span>
           </div>
-          <div class="text-xs text-muted-foreground mt-1">
+          <div class="mt-1 text-xs text-muted-foreground">
             状态: {{ provider.status }}
           </div>
         </div>
       </div>
+      <div
+        v-else
+        class="rounded-lg border border-dashed border-border/70 px-3 py-4 text-sm text-muted-foreground"
+      >
+        当前没有提供商状态数据。
+      </div>
     </Card>
   </div>
 </template>
 
 <script setup lang="ts">
-import { ref, computed, onMounted, onUnmounted, watch } from 'vue'
-import Card from '@/components/ui/card.vue'
-import { LoadingState, TimeRangePicker } from '@/components/common'
-import { ErrorDistributionChart, PercentileChart } from '@/components/stats'
-import LineChart from '@/components/charts/LineChart.vue'
+import { computed, onMounted, onUnmounted, ref, watch } from 'vue'
+import {
+  Activity,
+  AlertTriangle,
+  Cable,
+  GitBranch,
+  ShieldCheck,
+  Workflow,
+} from 'lucide-vue-next'
 import { adminApi, type ErrorDistributionResponse, type PercentileItem } from '@/api/admin'
 import { dashboardApi, type ProviderStatus } from '@/api/dashboard'
+import {
+  monitoringApi,
+  type AdminMonitoringCircuitHistoryItem,
+  type AdminMonitoringResilienceStatus,
+  type AdminMonitoringSystemStatus,
+  type GatewayMetricsSummary,
+} from '@/api/monitoring'
+import LineChart from '@/components/charts/LineChart.vue'
+import { LoadingState, TimeRangePicker } from '@/components/common'
+import { ErrorDistributionChart, PercentileChart } from '@/components/stats'
+import Badge from '@/components/ui/badge.vue'
+import Card from '@/components/ui/card.vue'
+import RefreshButton from '@/components/ui/refresh-button.vue'
+import { useToast } from '@/composables/useToast'
 import { getDateRangeFromPeriod } from '@/features/usage/composables'
 import type { DateRangeParams } from '@/features/usage/types'
+import { formatDate, formatNumber, formatTokens } from '@/utils/format'
+import { log } from '@/utils/logger'
+
+const LIVE_REFRESH_INTERVAL_MS = 10_000
 
 const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod('last30days'))
+const { error: showError } = useToast()
 
 const percentiles = ref<PercentileItem[]>([])
 const percentileLoading = ref(false)
@@ -112,12 +601,25 @@ const errorLoading = ref(false)
 
 const providerStatus = ref<ProviderStatus[]>([])
 const providerLoading = ref(false)
+
+const systemStatus = ref<AdminMonitoringSystemStatus | null>(null)
+const resilienceStatus = ref<AdminMonitoringResilienceStatus | null>(null)
+const circuitHistory = ref<AdminMonitoringCircuitHistoryItem[]>([])
+const gatewayMetrics = ref<GatewayMetricsSummary | null>(null)
+const liveLoading = ref(false)
+const liveRefreshing = ref(false)
+const liveReady = ref(false)
+const liveLoadError = ref<string | null>(null)
+const liveLastUpdatedAt = ref<string | null>(null)
+
 let percentilesRequestId = 0
 let errorsRequestId = 0
 let providersRequestId = 0
+let liveRequestId = 0
 let loadAllPromise: Promise<void> | null = null
 let hasPendingLoadAll = false
 let loadAllDebounceTimer: ReturnType<typeof setTimeout> | null = null
+let liveRefreshTimer: ReturnType<typeof setInterval> | null = null
 
 function buildTimeRangeParams() {
   return {
@@ -129,6 +631,18 @@ function buildTimeRangeParams() {
   }
 }
 
+function formatMetricNumber(value: number | null | undefined): string {
+  if (value == null || Number.isNaN(value)) {
+    return '-'
+  }
+
+  if (!Number.isInteger(value)) {
+    return value.toFixed(2)
+  }
+
+  return formatNumber(value)
+}
+
 async function loadPercentiles() {
   const requestId = ++percentilesRequestId
   percentileLoading.value = true
@@ -136,6 +650,10 @@ async function loadPercentiles() {
     const data = await adminApi.getPercentiles(buildTimeRangeParams())
     if (requestId !== percentilesRequestId) return
     percentiles.value = data
+  } catch (error) {
+    if (requestId !== percentilesRequestId) return
+    percentiles.value = []
+    log.error('加载延迟百分位失败', error)
   } finally {
     if (requestId === percentilesRequestId) {
       percentileLoading.value = false
@@ -151,6 +669,11 @@ async function loadErrors() {
     if (requestId !== errorsRequestId) return
     errorDistribution.value = response.distribution
     errorTrend.value = response.trend
+  } catch (error) {
+    if (requestId !== errorsRequestId) return
+    errorDistribution.value = []
+    errorTrend.value = []
+    log.error('加载错误分布失败', error)
   } finally {
     if (requestId === errorsRequestId) {
       errorLoading.value = false
@@ -165,6 +688,10 @@ async function loadProviders() {
     const data = await dashboardApi.getProviderStatus()
     if (requestId !== providersRequestId) return
     providerStatus.value = data
+  } catch (error) {
+    if (requestId !== providersRequestId) return
+    providerStatus.value = []
+    log.error('加载提供商状态失败', error)
   } finally {
     if (requestId === providersRequestId) {
       providerLoading.value = false
@@ -172,6 +699,82 @@ async function loadProviders() {
   }
 }
 
+async function loadLiveData(options: { silent?: boolean } = {}) {
+  const requestId = ++liveRequestId
+  const initialLoad = !liveReady.value
+
+  if (initialLoad) {
+    liveLoading.value = true
+  } else {
+    liveRefreshing.value = true
+  }
+
+  const results = await Promise.allSettled([
+    monitoringApi.getSystemStatus(),
+    monitoringApi.getResilienceStatus(),
+    monitoringApi.getCircuitHistory(8),
+    monitoringApi.getGatewayMetricsSummary(),
+  ])
+
+  if (requestId !== liveRequestId) {
+    return
+  }
+
+  const failedScopes: string[] = []
+  let successCount = 0
+
+  const [systemResult, resilienceResult, circuitResult, metricsResult] = results
+
+  if (systemResult.status === 'fulfilled') {
+    systemStatus.value = systemResult.value
+    successCount += 1
+  } else {
+    failedScopes.push('系统状态')
+    log.error('加载系统状态失败', systemResult.reason)
+  }
+
+  if (resilienceResult.status === 'fulfilled') {
+    resilienceStatus.value = resilienceResult.value
+    successCount += 1
+  } else {
+    failedScopes.push('韧性状态')
+    log.error('加载韧性状态失败', resilienceResult.reason)
+  }
+
+  if (circuitResult.status === 'fulfilled') {
+    circuitHistory.value = circuitResult.value.items
+    successCount += 1
+  } else {
+    failedScopes.push('熔断历史')
+    log.error('加载熔断历史失败', circuitResult.reason)
+  }
+
+  if (metricsResult.status === 'fulfilled') {
+    gatewayMetrics.value = metricsResult.value
+    successCount += 1
+  } else {
+    failedScopes.push('网关指标')
+    log.error('加载网关指标失败', metricsResult.reason)
+  }
+
+  liveReady.value = successCount > 0
+  if (successCount > 0) {
+    liveLastUpdatedAt.value = new Date().toISOString()
+  }
+  liveLoadError.value = failedScopes.length
+    ? `部分实时数据加载失败：${failedScopes.join('、')}`
+    : null
+
+  if (failedScopes.length && !options.silent) {
+    showError(liveLoadError.value ?? '实时性能数据加载失败')
+  }
+
+  if (requestId === liveRequestId) {
+    liveLoading.value = false
+    liveRefreshing.value = false
+  }
+}
+
 const errorTrendChartData = computed(() => ({
   labels: errorTrend.value.map(item => item.date),
   datasets: [
@@ -185,11 +788,147 @@ const errorTrendChartData = computed(() => ({
   ]
 }))
 
+const recentErrors = computed(() => resilienceStatus.value?.recent_errors ?? [])
+const resilienceRecommendations = computed(() => resilienceStatus.value?.recommendations ?? [])
+
+const healthStatusVariant = computed<'success' | 'warning' | 'destructive' | 'outline'>(() => {
+  switch (resilienceStatus.value?.status) {
+    case 'healthy':
+      return 'success'
+    case 'degraded':
+      return 'warning'
+    case 'critical':
+      return 'destructive'
+    default:
+      return 'outline'
+  }
+})
+
+const healthStatusText = computed(() => {
+  if (!resilienceStatus.value) {
+    return '健康状态未知'
+  }
+
+  const statusMap: Record<string, string> = {
+    healthy: '系统健康',
+    degraded: '系统降级',
+    critical: '系统告警',
+  }
+
+  return `${statusMap[resilienceStatus.value.status] ?? resilienceStatus.value.status} · ${resilienceStatus.value.health_score}/100`
+})
+
+const metricsAvailabilityText = computed(() => (
+  gatewayMetrics.value ? 'Prometheus 在线' : 'Prometheus 暂不可达'
+))
+
+const distributedGateVariant = computed<'warning' | 'outline'>(() => (
+  gatewayMetrics.value?.distributed.unavailable ? 'warning' : 'outline'
+))
+
+const distributedGateText = computed(() => (
+  gatewayMetrics.value?.distributed.unavailable ? '不可用' : '在线'
+))
+
+const liveLastUpdatedLabel = computed(() => (
+  liveLastUpdatedAt.value ? formatDate(liveLastUpdatedAt.value) : '尚未刷新'
+))
+
+const currentActiveStreams = computed(() => (
+  gatewayMetrics.value?.tunnel.activeStreams ?? systemStatus.value?.tunnel.active_streams ?? null
+))
+
+const currentProxyConnections = computed(() => (
+  gatewayMetrics.value?.tunnel.proxyConnections ?? systemStatus.value?.tunnel.proxy_connections ?? null
+))
+
+const currentTunnelNodes = computed(() => (
+  gatewayMetrics.value?.tunnel.nodes ?? systemStatus.value?.tunnel.nodes ?? null
+))
+
+const providerAndKeySummary = computed(() => {
+  if (!systemStatus.value) {
+    return '-'
+  }
+  return `${systemStatus.value.providers.active}/${systemStatus.value.providers.total} · ${systemStatus.value.api_keys.active}/${systemStatus.value.api_keys.total}`
+})
+
+const fallbackRows = computed(() => {
+  const items = gatewayMetrics.value?.fallbacks ?? []
+  const maxValue = Math.max(...items.map(item => item.total), 0)
+
+  return items
+    .filter(item => item.total > 0)
+    .sort((left, right) => right.total - left.total)
+    .map(item => ({
+      ...item,
+      ratio: maxValue > 0 ? item.total / maxValue * 100 : 0,
+    }))
+})
+
+const liveSummaryCards = computed(() => [
+  {
+    title: '系统健康',
+    value: resilienceStatus.value ? `${resilienceStatus.value.health_score}/100` : '-',
+    hint: `${healthStatusText.value} · 开路 ${formatMetricNumber(resilienceStatus.value?.error_statistics.open_circuit_breakers)}`,
+    icon: ShieldCheck,
+    iconClass: 'text-emerald-500',
+  },
+  {
+    title: '最近 1 小时错误',
+    value: formatMetricNumber(systemStatus.value?.recent_errors),
+    hint: `24h 总错误 ${formatMetricNumber(resilienceStatus.value?.error_statistics.total_errors)}`,
+    icon: AlertTriangle,
+    iconClass: 'text-yellow-500',
+  },
+  {
+    title: '当前活跃流',
+    value: formatMetricNumber(currentActiveStreams.value),
+    hint: `代理连接 ${formatMetricNumber(currentProxyConnections.value)}`,
+    icon: Cable,
+    iconClass: 'text-sky-500',
+  },
+  {
+    title: '本地 In Flight',
+    value: formatMetricNumber(gatewayMetrics.value?.local.inFlight),
+    hint: `剩余 permit ${formatMetricNumber(gatewayMetrics.value?.local.availablePermits)}`,
+    icon: Activity,
+    iconClass: 'text-blue-500',
+  },
+  {
+    title: '分布式 In Flight',
+    value: gatewayMetrics.value?.distributed.unavailable
+      ? '不可用'
+      : formatMetricNumber(gatewayMetrics.value?.distributed.inFlight),
+    hint: gatewayMetrics.value?.distributed.unavailable
+      ? '检查 Redis gate 状态'
+      : `剩余 permit ${formatMetricNumber(gatewayMetrics.value?.distributed.availablePermits)}`,
+    icon: Workflow,
+    iconClass: 'text-violet-500',
+  },
+  {
+    title: 'Fallback 累计',
+    value: formatMetricNumber(gatewayMetrics.value?.fallbackTotal),
+    hint: `今日请求 ${formatMetricNumber(systemStatus.value?.today_stats.requests)}`,
+    icon: GitBranch,
+    iconClass: 'text-rose-500',
+  },
+])
+
+const isRefreshing = computed(() => (
+  liveLoading.value ||
+  liveRefreshing.value ||
+  percentileLoading.value ||
+  errorLoading.value ||
+  providerLoading.value
+))
+
 async function loadAll() {
   if (loadAllPromise) {
     hasPendingLoadAll = true
     return loadAllPromise
   }
+
   loadAllPromise = Promise.all([loadPercentiles(), loadErrors(), loadProviders()])
     .then(() => undefined)
     .finally(() => {
@@ -199,13 +938,19 @@ async function loadAll() {
         void loadAll()
       }
     })
+
   return loadAllPromise
 }
 
+async function handleManualRefresh() {
+  await Promise.allSettled([loadLiveData(), loadAll()])
+}
+
 function scheduleLoadAll() {
   if (loadAllDebounceTimer) {
     clearTimeout(loadAllDebounceTimer)
   }
+
   loadAllDebounceTimer = setTimeout(() => {
     loadAllDebounceTimer = null
     void loadAll()
@@ -215,7 +960,11 @@ function scheduleLoadAll() {
 watch(timeRange, scheduleLoadAll, { deep: true })
 
 onMounted(() => {
+  void loadLiveData()
   void loadAll()
+  liveRefreshTimer = setInterval(() => {
+    void loadLiveData({ silent: true })
+  }, LIVE_REFRESH_INTERVAL_MS)
 })
 
 onUnmounted(() => {
@@ -223,10 +972,17 @@ onUnmounted(() => {
     clearTimeout(loadAllDebounceTimer)
     loadAllDebounceTimer = null
   }
+
+  if (liveRefreshTimer) {
+    clearInterval(liveRefreshTimer)
+    liveRefreshTimer = null
+  }
+
   hasPendingLoadAll = false
   loadAllPromise = null
   percentilesRequestId += 1
   errorsRequestId += 1
   providersRequestId += 1
+  liveRequestId += 1
 })
 </script>

frontend/vite.config.ts

@@ -73,6 +73,11 @@ export default defineConfig(({ mode }) => {
           changeOrigin: true,
           secure: false,
         },
+        '/_gateway/': {
+          target: gatewayTarget,
+          changeOrigin: true,
+          secure: false,
+        },
       },
     },
     preview: {