feat(gateway/data): 新增 usage 关键词搜索、缓存命中摘要、结算成本摘要及 dashboard 聚合查询能力

- 新增 UsageAuditKeywordSearchQuery,支持多关键词、用户名、API Key 交叉过滤
- 新增 UsageCacheHitSummaryQuery/StoredUsageCacheHitSummary,统计请求缓存命中率
- 新增 UsageSettledCostSummaryQuery/StoredUsageSettledCostSummary,汇总结算成本
- 新增 UsageDashboardSummaryQuery、UsageBreakdownSummaryQuery 等 dashboard 聚合类型
- SQL 层实现对应查询方法,含 list_by_ids、list_usage_audits_by_keyword_search、count_usage_audits_by_keyword_search
- 将上述能力通过 GatewayDataState / AdminAppState 暴露给 handler 层
- user_me_usage 及 dashboard_filters 切换为新查询接口,移除旧的内存过滤逻辑
- 同步更新 memory 层及测试
This commit is contained in:
fawney19
2026-04-17 01:00:38 +08:00
parent 6e5af5ef70
commit 1e0bc61526
48 changed files with 5241 additions and 1162 deletions
@@ -4,7 +4,6 @@ use super::route_filters::{
parse_admin_monitoring_limit, parse_admin_monitoring_offset,
parse_admin_monitoring_username_filter,
};
use super::usage_helpers::admin_monitoring_usage_is_error;
use crate::constants::INTERNAL_GATEWAY_PATH_PREFIXES;
use crate::handlers::admin::request::{AdminAppState, AdminRequestContext};
use crate::query::monitoring as monitoring_query;
@@ -16,7 +15,9 @@ use aether_admin::observability::monitoring::{
build_admin_monitoring_system_status_payload_response,
build_admin_monitoring_user_behavior_payload_response,
};
use aether_data_contracts::repository::usage::UsageAuditListQuery;
use aether_data_contracts::repository::usage::{
UsageAuditSummaryQuery, UsageMonitoringErrorCountQuery,
};
use axum::{body::Body, response::Response};
pub(super) async fn build_admin_monitoring_audit_logs_response(
@@ -196,30 +197,27 @@ pub(super) async fn build_admin_monitoring_system_status_response(
.saturating_add(standalone_api_key_summary.active);
let today_usage = state
.list_usage_audits(&UsageAuditListQuery {
created_from_unix_secs: Some(today_start.timestamp().max(0) as u64),
..Default::default()
.summarize_usage_audits(&UsageAuditSummaryQuery {
created_from_unix_secs: today_start.timestamp().max(0) as u64,
created_until_unix_secs: now_unix_secs.saturating_add(1),
user_id: None,
provider_name: None,
model: None,
})
.await?;
let today_requests = today_usage.len();
let today_tokens = today_usage
.iter()
.map(|item| item.total_tokens)
.sum::<u64>();
let today_cost = today_usage
.iter()
.map(|item| item.total_cost_usd)
.sum::<f64>();
let today_requests = usize::try_from(today_usage.total_requests).unwrap_or(usize::MAX);
let today_tokens = today_usage.recorded_total_tokens;
let today_cost = today_usage.total_cost_usd;
let recent_errors = state
.list_usage_audits(&UsageAuditListQuery {
created_from_unix_secs: Some(recent_error_from.timestamp().max(0) as u64),
..Default::default()
})
.await?
.into_iter()
.filter(admin_monitoring_usage_is_error)
.count();
let recent_errors = usize::try_from(
state
.count_monitoring_usage_errors(&UsageMonitoringErrorCountQuery {
created_from_unix_secs: recent_error_from.timestamp().max(0) as u64,
created_until_unix_secs: now_unix_secs.saturating_add(1),
})
.await?,
)
.unwrap_or(usize::MAX);
let tunnel = state.tunnel.stats();
Ok(build_admin_monitoring_system_status_payload_response(
@@ -8,7 +8,7 @@ use crate::handlers::admin::observability::stats::round_to;
use crate::handlers::admin::request::AdminAppState;
use crate::scheduler::affinity::SCHEDULER_AFFINITY_TTL;
use crate::GatewayError;
use aether_data_contracts::repository::usage::UsageAuditListQuery;
use aether_data_contracts::repository::usage::UsageCacheHitSummaryQuery;
async fn count_admin_monitoring_cache_affinity_entries(state: &AdminAppState<'_>) -> usize {
list_admin_monitoring_cache_affinity_records(state)
@@ -325,27 +325,26 @@ pub(super) async fn build_admin_monitoring_cache_snapshot(
.unwrap_or_else(|| "provider".to_string());
let now = chrono::Utc::now();
let usage = if state.has_usage_data_reader() {
let usage_summary = if state.has_usage_data_reader() {
state
.list_usage_audits(&UsageAuditListQuery {
created_from_unix_secs: Some(
(now - chrono::Duration::hours(24)).timestamp().max(0) as u64,
),
..Default::default()
.summarize_usage_cache_hit_summary(&UsageCacheHitSummaryQuery {
created_from_unix_secs: (now - chrono::Duration::hours(24)).timestamp().max(0)
as u64,
created_until_unix_secs: now.timestamp().max(0) as u64,
user_id: None,
})
.await?
} else {
Vec::new()
Default::default()
};
let cache_hits = usage
.iter()
.filter(|item| item.cache_read_input_tokens > 0)
.count();
let cache_misses = usage.len().saturating_sub(cache_hits);
let cache_hit_rate = if usage.is_empty() {
let cache_hits = usage_summary.cache_hit_requests as usize;
let cache_misses = usage_summary
.total_requests
.saturating_sub(usage_summary.cache_hit_requests) as usize;
let cache_hit_rate = if usage_summary.total_requests == 0 {
0.0
} else {
round_to(cache_hits as f64 / usage.len() as f64, 4)
round_to(cache_hits as f64 / usage_summary.total_requests as f64, 4)
};
let total_affinities = count_admin_monitoring_cache_affinity_entries(state).await;
let storage_type = if state.redis_kv_runner().is_some() {
@@ -1,9 +1,8 @@
use super::super::usage_helpers::admin_monitoring_usage_is_error;
use crate::handlers::admin::request::AdminAppState;
use crate::handlers::admin::shared::{provider_key_health_summary, unix_secs_to_rfc3339};
use crate::GatewayError;
use aether_data_contracts::repository::{
provider_catalog::StoredProviderCatalogKey, usage::UsageAuditListQuery,
provider_catalog::StoredProviderCatalogKey, usage::UsageMonitoringErrorListQuery,
};
use serde_json::json;
use std::collections::BTreeMap;
@@ -152,14 +151,12 @@ pub(super) async fn build_admin_monitoring_resilience_snapshot(
}
let mut recent_usage_errors = state
.list_usage_audits(&UsageAuditListQuery {
created_from_unix_secs: Some(recent_error_from.timestamp().max(0) as u64),
..Default::default()
.list_monitoring_usage_errors(&UsageMonitoringErrorListQuery {
created_from_unix_secs: recent_error_from.timestamp().max(0) as u64,
created_until_unix_secs: (now.timestamp().max(0) as u64).saturating_add(1),
limit: None,
})
.await?
.into_iter()
.filter(admin_monitoring_usage_is_error)
.collect::<Vec<_>>();
.await?;
recent_usage_errors
.sort_by(|left, right| right.created_at_unix_ms.cmp(&left.created_at_unix_ms));