perf(usage): heatmap 改为数据库端按天聚合查询

将 admin 和 user heatmap 从逐条加载 usage audit 记录后在应用层聚合,
改为通过 SQL GROUP BY DATE 在数据库端直接按天汇总, 大幅减少数据传输量。

新增 UsageDailyHeatmapQuery / StoredUsageDailySummary 类型,
在 trait、SQL、内存实现中均补齐 summarize_usage_daily_heatmap 方法。

同时优化 docker-compose: postgres 增加空闲事务超时与 keepalive 参数,
gateway 增加健康检查配置。

crates/aether-data-contracts/src/repository/usage/mod.rs

@@ -3,6 +3,6 @@ mod types;
 pub use types::{
     parse_usage_body_ref, usage_body_ref, StoredProviderApiKeyUsageSummary,
     StoredProviderUsageSummary, StoredProviderUsageWindow, StoredRequestUsageAudit,
-    UpsertUsageRecord, UsageAuditListQuery, UsageBodyField, UsageReadRepository, UsageRepository,
-    UsageWriteRepository,
+    StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery, UsageBodyField,
+    UsageDailyHeatmapQuery, UsageReadRepository, UsageRepository, UsageWriteRepository,
 };

crates/aether-data-contracts/src/repository/usage/types.rs

@@ -482,6 +482,25 @@ pub struct UsageAuditListQuery {
     pub model: Option<String>,
 }
 
+#[derive(Debug, Clone, PartialEq, Default, serde::Serialize, serde::Deserialize)]
+pub struct UsageDailyHeatmapQuery {
+    pub created_from_unix_secs: u64,
+    pub user_id: Option<String>,
+    /// When true, exclude rows with status in ('pending', 'streaming') (admin heatmap).
+    /// When false, only include rows with billing_status = 'settled' and total_cost_usd > 0 (user heatmap).
+    pub admin_mode: bool,
+}
+
+#[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize)]
+pub struct StoredUsageDailySummary {
+    /// Date as "YYYY-MM-DD"
+    pub date: String,
+    pub requests: u64,
+    pub total_tokens: u64,
+    pub total_cost_usd: f64,
+    pub actual_total_cost_usd: f64,
+}
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
 #[serde(rename_all = "snake_case")]
 pub enum UsageBodyField {
@@ -586,6 +605,11 @@ pub trait UsageReadRepository: Send + Sync {
         provider_id: &str,
         since_unix_secs: u64,
     ) -> Result<StoredProviderUsageSummary, crate::DataLayerError>;
+
+    async fn summarize_usage_daily_heatmap(
+        &self,
+        query: &UsageDailyHeatmapQuery,
+    ) -> Result<Vec<StoredUsageDailySummary>, crate::DataLayerError>;
 }
 
 #[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize)]

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

@@ -10,8 +10,8 @@ use serde_json::Value;
 use super::{
     strip_deprecated_usage_display_fields, usage_can_recover_terminal_failure,
     StoredProviderApiKeyUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
-    StoredRequestUsageAudit, UpsertUsageRecord, UsageAuditListQuery, UsageReadRepository,
-    UsageWriteRepository,
+    StoredRequestUsageAudit, StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery,
+    UsageDailyHeatmapQuery, UsageReadRepository, UsageWriteRepository,
 };
 use crate::DataLayerError;
 
@@ -335,6 +335,70 @@ impl UsageReadRepository for InMemoryUsageReadRepository {
 
         Ok(summary)
     }
+
+    async fn summarize_usage_daily_heatmap(
+        &self,
+        query: &UsageDailyHeatmapQuery,
+    ) -> Result<Vec<StoredUsageDailySummary>, DataLayerError> {
+        let items = self.by_request_id.read().expect("usage repository lock");
+        let mut daily = BTreeMap::<String, (u64, u64, f64, f64)>::new();
+        for item in items.values() {
+            if item.created_at_unix_ms < query.created_from_unix_secs {
+                continue;
+            }
+            if let Some(user_id) = &query.user_id {
+                if item.user_id.as_deref() != Some(user_id) {
+                    continue;
+                }
+            }
+            if query.admin_mode {
+                if item.status == "pending" || item.status == "streaming" {
+                    continue;
+                }
+            } else if item.billing_status != "settled" || item.total_cost_usd <= 0.0 {
+                continue;
+            }
+            let ts = i64::try_from(item.created_at_unix_ms).unwrap_or_default();
+            let Some(dt) = chrono::DateTime::<chrono::Utc>::from_timestamp(ts, 0) else {
+                continue;
+            };
+            let date_key = dt.date_naive().to_string();
+            let entry = daily.entry(date_key).or_insert((0, 0, 0.0, 0.0));
+            entry.0 += 1;
+            let cache_creation = if item.cache_creation_input_tokens == 0
+                && (item.cache_creation_ephemeral_5m_input_tokens
+                    + item.cache_creation_ephemeral_1h_input_tokens)
+                    > 0
+            {
+                item.cache_creation_ephemeral_5m_input_tokens
+                    + item.cache_creation_ephemeral_1h_input_tokens
+            } else {
+                item.cache_creation_input_tokens
+            };
+            entry.1 += item.input_tokens
+                + item.output_tokens
+                + cache_creation
+                + item.cache_read_input_tokens;
+            entry.2 += item.total_cost_usd;
+            entry.3 += item.actual_total_cost_usd;
+        }
+        let mut result: Vec<_> = daily
+            .into_iter()
+            .map(
+                |(date, (requests, total_tokens, total_cost_usd, actual_total_cost_usd))| {
+                    StoredUsageDailySummary {
+                        date,
+                        requests,
+                        total_tokens,
+                        total_cost_usd,
+                        actual_total_cost_usd,
+                    }
+                },
+            )
+            .collect();
+        result.sort_by(|a, b| a.date.cmp(&b.date));
+        Ok(result)
+    }
 }
 
 fn detach_usage_body(

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

@@ -4,8 +4,8 @@ mod sql;
 #[allow(unused_imports)]
 pub(crate) use aether_data_contracts::repository::usage::{
     StoredProviderApiKeyUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
-    StoredRequestUsageAudit, UpsertUsageRecord, UsageAuditListQuery, UsageReadRepository,
-    UsageRepository, UsageWriteRepository,
+    StoredRequestUsageAudit, StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery,
+    UsageDailyHeatmapQuery, UsageReadRepository, UsageRepository, UsageWriteRepository,
 };
 pub use memory::InMemoryUsageReadRepository;
 pub use sql::SqlxUsageReadRepository;

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

@@ -14,7 +14,8 @@ use uuid::Uuid;
 use super::{
     incoming_usage_can_recover_terminal_failure, strip_deprecated_usage_display_fields,
     StoredProviderApiKeyUsageSummary, StoredProviderUsageSummary, StoredRequestUsageAudit,
-    UpsertUsageRecord, UsageAuditListQuery, UsageReadRepository, UsageWriteRepository,
+    StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery, UsageDailyHeatmapQuery,
+    UsageReadRepository, UsageWriteRepository,
 };
 use crate::postgres::PostgresTransactionRunner;
 use crate::{error::SqlxResultExt, DataLayerError};
@@ -1209,6 +1210,67 @@ impl SqlxUsageReadRepository {
         Ok(items)
     }
 
+    pub async fn summarize_usage_daily_heatmap(
+        &self,
+        query: &UsageDailyHeatmapQuery,
+    ) -> Result<Vec<StoredUsageDailySummary>, DataLayerError> {
+        let mut sql = String::from(
+            r#"SELECT
+  DATE("usage".created_at) AS day,
+  COUNT(*)::BIGINT AS requests,
+  COALESCE(SUM("usage".input_tokens + "usage".output_tokens
+    + CASE
+        WHEN COALESCE("usage".cache_creation_input_tokens, 0) = 0
+             AND (COALESCE("usage".cache_creation_input_tokens_5m, 0) + COALESCE("usage".cache_creation_input_tokens_1h, 0)) > 0
+        THEN COALESCE("usage".cache_creation_input_tokens_5m, 0) + COALESCE("usage".cache_creation_input_tokens_1h, 0)
+        ELSE COALESCE("usage".cache_creation_input_tokens, 0)
+      END
+    + COALESCE("usage".cache_read_input_tokens, 0)), 0)::BIGINT AS total_tokens,
+  COALESCE(SUM(CAST("usage".total_cost_usd AS DOUBLE PRECISION)), 0) AS total_cost_usd,
+  COALESCE(SUM(CAST("usage".actual_total_cost_usd AS DOUBLE PRECISION)), 0) AS actual_total_cost_usd
+FROM "usage"
+WHERE "usage".created_at >= TO_TIMESTAMP($1::double precision)"#,
+        );
+        if query.admin_mode {
+            sql.push_str(" AND \"usage\".status NOT IN ('pending', 'streaming')");
+        } else {
+            sql.push_str(
+                " AND \"usage\".billing_status = 'settled' AND CAST(\"usage\".total_cost_usd AS DOUBLE PRECISION) > 0",
+            );
+        }
+        let mut bind_index = 2;
+        if query.user_id.is_some() {
+            sql.push_str(&format!(" AND \"usage\".user_id = ${bind_index}"));
+            bind_index += 1;
+        }
+        let _ = bind_index;
+        sql.push_str(" GROUP BY day ORDER BY day ASC");
+
+        let mut q = sqlx::query(&sql).bind(query.created_from_unix_secs as f64);
+        if let Some(user_id) = &query.user_id {
+            q = q.bind(user_id.clone());
+        }
+
+        let mut rows = q.fetch(&self.pool);
+        let mut items = Vec::new();
+        while let Some(row) = rows.try_next().await.map_postgres_err()? {
+            let day: chrono::NaiveDate = row.try_get("day").map_postgres_err()?;
+            let requests: i64 = row.try_get("requests").map_postgres_err()?;
+            let total_tokens: i64 = row.try_get("total_tokens").map_postgres_err()?;
+            let total_cost_usd: f64 = row.try_get("total_cost_usd").map_postgres_err()?;
+            let actual_total_cost_usd: f64 =
+                row.try_get("actual_total_cost_usd").map_postgres_err()?;
+            items.push(StoredUsageDailySummary {
+                date: day.to_string(),
+                requests: requests as u64,
+                total_tokens: total_tokens as u64,
+                total_cost_usd,
+                actual_total_cost_usd,
+            });
+        }
+        Ok(items)
+    }
+
     pub async fn list_recent_usage_audits(
         &self,
         user_id: Option<&str>,
@@ -1703,6 +1765,13 @@ impl UsageReadRepository for SqlxUsageReadRepository {
     ) -> Result<StoredProviderUsageSummary, DataLayerError> {
         Self::summarize_provider_usage_since(self, provider_id, since_unix_secs).await
     }
+
+    async fn summarize_usage_daily_heatmap(
+        &self,
+        query: &UsageDailyHeatmapQuery,
+    ) -> Result<Vec<StoredUsageDailySummary>, DataLayerError> {
+        Self::summarize_usage_daily_heatmap(self, query).await
+    }
 }
 
 #[async_trait]