mirror of
https://github.com/fawney19/Aether.git
synced 2026-09-02 01:10:23 +08:00
Merge pull request #374 from Entropy-Xu/codex/provider-performance-stats
feat(stats): 添加 Provider 性能统计分析
This commit is contained in:
@@ -495,6 +495,19 @@ pub(super) fn classify_admin_observability_family_route(
|
||||
"admin:stats",
|
||||
false,
|
||||
))
|
||||
} else if method == http::Method::GET
|
||||
&& matches!(
|
||||
normalized_path,
|
||||
"/api/admin/stats/performance/providers" | "/api/admin/stats/performance/providers/"
|
||||
)
|
||||
{
|
||||
Some(classified(
|
||||
"admin_proxy",
|
||||
"stats_manage",
|
||||
"provider_performance",
|
||||
"admin:stats",
|
||||
false,
|
||||
))
|
||||
} else if method == http::Method::GET
|
||||
&& matches!(
|
||||
normalized_path,
|
||||
|
||||
@@ -81,6 +81,25 @@ fn classifies_admin_stats_performance_percentiles_as_admin_proxy_route() {
|
||||
assert!(!decision.is_execution_runtime_candidate());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn classifies_admin_stats_provider_performance_as_admin_proxy_route() {
|
||||
let headers = headers(&[]);
|
||||
let uri: Uri = "/api/admin/stats/performance/providers"
|
||||
.parse()
|
||||
.expect("uri should parse");
|
||||
let decision =
|
||||
classify_control_route(&http::Method::GET, &uri, &headers).expect("route should classify");
|
||||
|
||||
assert_eq!(decision.route_class.as_deref(), Some("admin_proxy"));
|
||||
assert_eq!(decision.route_family.as_deref(), Some("stats_manage"));
|
||||
assert_eq!(decision.route_kind.as_deref(), Some("provider_performance"));
|
||||
assert_eq!(
|
||||
decision.auth_endpoint_signature.as_deref(),
|
||||
Some("admin:stats")
|
||||
);
|
||||
assert!(!decision.is_execution_runtime_candidate());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn classifies_admin_stats_cost_forecast_as_admin_proxy_route() {
|
||||
let headers = headers(&[]);
|
||||
|
||||
@@ -1029,6 +1029,21 @@ impl GatewayDataState {
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_provider_performance(
|
||||
&self,
|
||||
query: &aether_data_contracts::repository::usage::UsageProviderPerformanceQuery,
|
||||
) -> Result<
|
||||
aether_data_contracts::repository::usage::StoredUsageProviderPerformance,
|
||||
DataLayerError,
|
||||
> {
|
||||
match &self.usage_reader {
|
||||
Some(repository) => repository.summarize_usage_provider_performance(query).await,
|
||||
None => Ok(
|
||||
aether_data_contracts::repository::usage::StoredUsageProviderPerformance::default(),
|
||||
),
|
||||
}
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_cost_savings(
|
||||
&self,
|
||||
query: &aether_data_contracts::repository::usage::UsageCostSavingsSummaryQuery,
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
use super::range::build_comparison_range;
|
||||
use super::range::{build_comparison_range, parse_bounded_u32};
|
||||
use super::resolve_admin_usage_time_range;
|
||||
use crate::handlers::admin::request::{AdminAppState, AdminRequestContext};
|
||||
use crate::handlers::admin::shared::query_param_value;
|
||||
@@ -6,16 +6,18 @@ use crate::GatewayError;
|
||||
use aether_admin::observability::stats::{
|
||||
admin_stats_bad_request_response, admin_stats_comparison_empty_response,
|
||||
admin_stats_error_distribution_empty_response,
|
||||
admin_stats_performance_percentiles_empty_response, admin_stats_time_series_empty_response,
|
||||
admin_stats_performance_percentiles_empty_response,
|
||||
admin_stats_provider_performance_empty_response, admin_stats_time_series_empty_response,
|
||||
build_admin_stats_comparison_response_from_aggregates,
|
||||
build_admin_stats_error_distribution_response_from_summaries,
|
||||
build_admin_stats_performance_percentiles_response_from_summaries,
|
||||
build_admin_stats_provider_performance_response,
|
||||
build_admin_stats_time_series_response_from_summaries, AdminStatsAggregate,
|
||||
AdminStatsComparisonType, AdminStatsGranularity, AdminStatsTimeRange, AdminStatsUsageFilter,
|
||||
};
|
||||
use aether_data_contracts::repository::usage::{
|
||||
UsageAuditSummaryQuery, UsageErrorDistributionQuery, UsagePerformancePercentilesQuery,
|
||||
UsageTimeSeriesGranularity, UsageTimeSeriesQuery,
|
||||
UsageProviderPerformanceQuery, UsageTimeSeriesGranularity, UsageTimeSeriesQuery,
|
||||
};
|
||||
use axum::{body::Body, http, response::Response};
|
||||
|
||||
@@ -173,6 +175,56 @@ pub(super) async fn maybe_build_local_admin_stats_analytics_response(
|
||||
));
|
||||
}
|
||||
|
||||
if request_context.route_kind() == Some("provider_performance")
|
||||
&& request_context.method() == http::Method::GET
|
||||
&& matches!(
|
||||
request_context.path(),
|
||||
"/api/admin/stats/performance/providers" | "/api/admin/stats/performance/providers/"
|
||||
)
|
||||
{
|
||||
let time_range = match resolve_admin_usage_time_range(request_context.query_string()) {
|
||||
Ok(value) => value,
|
||||
Err(detail) => return Ok(Some(admin_stats_bad_request_response(detail))),
|
||||
};
|
||||
let granularity =
|
||||
match query_param_value(request_context.query_string(), "granularity").as_deref() {
|
||||
None | Some("day") => UsageTimeSeriesGranularity::Day,
|
||||
Some("hour") => UsageTimeSeriesGranularity::Hour,
|
||||
Some(_) => {
|
||||
return Ok(Some(admin_stats_bad_request_response(
|
||||
"granularity must be one of: day, hour".to_string(),
|
||||
)));
|
||||
}
|
||||
};
|
||||
let limit = match query_param_value(request_context.query_string(), "limit")
|
||||
.map(|value| parse_bounded_u32("limit", &value, 1, 20))
|
||||
.transpose()
|
||||
{
|
||||
Ok(value) => value.unwrap_or(8) as usize,
|
||||
Err(detail) => return Ok(Some(admin_stats_bad_request_response(detail))),
|
||||
};
|
||||
if !state.has_usage_data_reader() {
|
||||
return Ok(Some(admin_stats_provider_performance_empty_response()));
|
||||
}
|
||||
|
||||
let Some((created_from_unix_secs, created_until_unix_secs)) = time_range.to_unix_bounds()
|
||||
else {
|
||||
return Ok(Some(admin_stats_provider_performance_empty_response()));
|
||||
};
|
||||
let performance = state
|
||||
.summarize_usage_provider_performance(&UsageProviderPerformanceQuery {
|
||||
created_from_unix_secs,
|
||||
created_until_unix_secs,
|
||||
granularity,
|
||||
tz_offset_minutes: time_range.tz_offset_minutes,
|
||||
limit,
|
||||
})
|
||||
.await?;
|
||||
return Ok(Some(build_admin_stats_provider_performance_response(
|
||||
&performance,
|
||||
)));
|
||||
}
|
||||
|
||||
if request_context.route_kind() == Some("time_series")
|
||||
&& request_context.method() == http::Method::GET
|
||||
&& matches!(
|
||||
|
||||
@@ -168,6 +168,16 @@ impl<'a> AdminAppState<'a> {
|
||||
.await
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_provider_performance(
|
||||
&self,
|
||||
query: &aether_data_contracts::repository::usage::UsageProviderPerformanceQuery,
|
||||
) -> Result<
|
||||
aether_data_contracts::repository::usage::StoredUsageProviderPerformance,
|
||||
GatewayError,
|
||||
> {
|
||||
self.app.summarize_usage_provider_performance(query).await
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_cost_savings(
|
||||
&self,
|
||||
query: &aether_data_contracts::repository::usage::UsageCostSavingsSummaryQuery,
|
||||
|
||||
@@ -251,6 +251,16 @@ impl AppState {
|
||||
.map_err(|err| GatewayError::Internal(err.to_string()))
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_provider_performance(
|
||||
&self,
|
||||
query: &usage::UsageProviderPerformanceQuery,
|
||||
) -> Result<usage::StoredUsageProviderPerformance, GatewayError> {
|
||||
self.data
|
||||
.summarize_usage_provider_performance(query)
|
||||
.await
|
||||
.map_err(|err| GatewayError::Internal(err.to_string()))
|
||||
}
|
||||
|
||||
pub(crate) async fn summarize_usage_cost_savings(
|
||||
&self,
|
||||
query: &usage::UsageCostSavingsSummaryQuery,
|
||||
|
||||
@@ -853,6 +853,187 @@ async fn gateway_handles_admin_stats_performance_percentiles_locally_without_usa
|
||||
upstream_handle.abort();
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn gateway_handles_admin_stats_provider_performance_locally_with_trusted_admin_principal() {
|
||||
let (_upstream_url, upstream_hits, upstream_handle) =
|
||||
start_stats_upstream("/api/admin/stats/performance/providers").await;
|
||||
|
||||
let mut usage = (1..=10)
|
||||
.map(|index| {
|
||||
let mut row = sample_usage_row(
|
||||
&format!("usage-provider-perf-a-{index}"),
|
||||
&format!("req-provider-perf-a-{index}"),
|
||||
Some("user-1"),
|
||||
Some("key-1"),
|
||||
Some("primary"),
|
||||
"OpenAI",
|
||||
"gpt-5",
|
||||
10,
|
||||
10,
|
||||
0.01,
|
||||
0.01,
|
||||
DAY_1_UNIX_SECS + i64::from(index),
|
||||
);
|
||||
row.response_time_ms = Some((index * 100) as u64);
|
||||
row.first_byte_time_ms = Some((index * 10) as u64);
|
||||
row
|
||||
})
|
||||
.collect::<Vec<_>>();
|
||||
let mut failed = sample_usage_row(
|
||||
"usage-provider-perf-failed",
|
||||
"req-provider-perf-failed",
|
||||
Some("user-1"),
|
||||
Some("key-1"),
|
||||
Some("primary"),
|
||||
"OpenAI",
|
||||
"gpt-5",
|
||||
10,
|
||||
99,
|
||||
0.01,
|
||||
0.01,
|
||||
DAY_1_UNIX_SECS + 20,
|
||||
);
|
||||
failed.status = "failed".to_string();
|
||||
failed.status_code = Some(500);
|
||||
failed.error_message = Some("upstream failed".to_string());
|
||||
usage.push(failed);
|
||||
|
||||
let mut provider_b = sample_usage_row(
|
||||
"usage-provider-perf-b",
|
||||
"req-provider-perf-b",
|
||||
Some("user-1"),
|
||||
Some("key-1"),
|
||||
Some("primary"),
|
||||
"Anthropic",
|
||||
"claude-sonnet",
|
||||
10,
|
||||
20,
|
||||
0.01,
|
||||
0.01,
|
||||
DAY_1_UNIX_SECS + 30,
|
||||
);
|
||||
provider_b.provider_id = Some("provider-2".to_string());
|
||||
provider_b.response_time_ms = Some(1000);
|
||||
provider_b.first_byte_time_ms = None;
|
||||
usage.push(provider_b);
|
||||
|
||||
let mut unknown_provider = sample_usage_row(
|
||||
"usage-provider-perf-unknown",
|
||||
"req-provider-perf-unknown",
|
||||
Some("user-1"),
|
||||
Some("key-1"),
|
||||
Some("primary"),
|
||||
"unknown",
|
||||
"gpt-5",
|
||||
10,
|
||||
999,
|
||||
0.01,
|
||||
0.01,
|
||||
DAY_1_UNIX_SECS + 40,
|
||||
);
|
||||
unknown_provider.provider_id = None;
|
||||
usage.push(unknown_provider);
|
||||
|
||||
let usage_repository = Arc::new(InMemoryUsageReadRepository::seed(usage));
|
||||
let gateway = build_router_with_state(
|
||||
AppState::new()
|
||||
.expect("gateway should build")
|
||||
.with_data_state_for_tests(GatewayDataState::with_usage_reader_for_tests(
|
||||
usage_repository,
|
||||
)),
|
||||
);
|
||||
let (gateway_url, gateway_handle) = start_server(gateway).await;
|
||||
|
||||
let response = admin_request(reqwest::Client::new().get(format!(
|
||||
"{gateway_url}/api/admin/stats/performance/providers?start_date=2024-03-21&end_date=2024-03-21&granularity=hour&limit=2&tz_offset_minutes=0"
|
||||
)))
|
||||
.send()
|
||||
.await
|
||||
.expect("request should succeed");
|
||||
|
||||
assert_eq!(response.status(), StatusCode::OK);
|
||||
let payload: serde_json::Value = response.json().await.expect("json body should parse");
|
||||
assert_eq!(payload["summary"]["request_count"], 12);
|
||||
assert_eq!(payload["summary"]["success_rate"], 91.67);
|
||||
assert_eq!(payload["summary"]["avg_output_tps"], 18.46);
|
||||
assert_eq!(payload["summary"]["avg_first_byte_time_ms"], 55.0);
|
||||
assert_eq!(payload["summary"]["avg_response_time_ms"], 590.91);
|
||||
|
||||
assert_eq!(payload["providers"].as_array().map(Vec::len), Some(2));
|
||||
assert_eq!(payload["providers"][0]["provider_id"], "provider-1");
|
||||
assert_eq!(payload["providers"][0]["provider"], "OpenAI");
|
||||
assert_eq!(payload["providers"][0]["request_count"], 11);
|
||||
assert_eq!(payload["providers"][0]["success_count"], 10);
|
||||
assert_eq!(payload["providers"][0]["error_count"], 1);
|
||||
assert_eq!(payload["providers"][0]["success_rate"], 90.91);
|
||||
assert_eq!(payload["providers"][0]["output_tokens"], 199);
|
||||
assert_eq!(payload["providers"][0]["avg_output_tps"], 18.18);
|
||||
assert_eq!(payload["providers"][0]["avg_first_byte_time_ms"], 55.0);
|
||||
assert_eq!(payload["providers"][0]["avg_response_time_ms"], 550.0);
|
||||
assert_eq!(payload["providers"][0]["p90_response_time_ms"], 910);
|
||||
assert_eq!(payload["providers"][0]["p90_first_byte_time_ms"], 91);
|
||||
assert_eq!(payload["providers"][0]["tps_sample_count"], 10);
|
||||
assert_eq!(payload["providers"][0]["first_byte_sample_count"], 10);
|
||||
|
||||
assert_eq!(payload["providers"][1]["provider_id"], "provider-2");
|
||||
assert_eq!(payload["providers"][1]["provider"], "Anthropic");
|
||||
assert_eq!(payload["providers"][1]["avg_output_tps"], 20.0);
|
||||
assert_eq!(
|
||||
payload["providers"][1]["avg_first_byte_time_ms"],
|
||||
serde_json::Value::Null
|
||||
);
|
||||
assert_eq!(
|
||||
payload["providers"][1]["p90_response_time_ms"],
|
||||
serde_json::Value::Null
|
||||
);
|
||||
|
||||
assert_eq!(payload["timeline"].as_array().map(Vec::len), Some(2));
|
||||
assert_eq!(payload["timeline"][0]["date"], "2024-03-21T05:00:00+00:00");
|
||||
assert_eq!(payload["timeline"][0]["provider_id"], "provider-1");
|
||||
assert_eq!(payload["timeline"][0]["avg_output_tps"], 18.18);
|
||||
assert_eq!(payload["timeline"][0]["success_rate"], 90.91);
|
||||
assert_eq!(payload["timeline"][1]["provider_id"], "provider-2");
|
||||
assert_eq!(
|
||||
payload["timeline"][1]["avg_first_byte_time_ms"],
|
||||
serde_json::Value::Null
|
||||
);
|
||||
assert_eq!(*upstream_hits.lock().expect("mutex should lock"), 0);
|
||||
|
||||
gateway_handle.abort();
|
||||
upstream_handle.abort();
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn gateway_returns_empty_admin_stats_provider_performance_without_usage_reader() {
|
||||
let (_upstream_url, upstream_hits, upstream_handle) =
|
||||
start_stats_upstream("/api/admin/stats/performance/providers").await;
|
||||
|
||||
let gateway = build_router_with_state(AppState::new().expect("gateway should build"));
|
||||
let (gateway_url, gateway_handle) = start_server(gateway).await;
|
||||
|
||||
let response = admin_request(reqwest::Client::new().get(format!(
|
||||
"{gateway_url}/api/admin/stats/performance/providers?start_date=2024-03-21&end_date=2024-03-21&limit=2"
|
||||
)))
|
||||
.send()
|
||||
.await
|
||||
.expect("request should succeed");
|
||||
|
||||
assert_eq!(response.status(), StatusCode::OK);
|
||||
let payload: serde_json::Value = response.json().await.expect("json body should parse");
|
||||
assert_eq!(payload["summary"]["request_count"], 0);
|
||||
assert_eq!(payload["summary"]["success_rate"], 0.0);
|
||||
assert_eq!(
|
||||
payload["summary"]["avg_output_tps"],
|
||||
serde_json::Value::Null
|
||||
);
|
||||
assert_eq!(payload["providers"].as_array().map(Vec::len), Some(0));
|
||||
assert_eq!(payload["timeline"].as_array().map(Vec::len), Some(0));
|
||||
assert_eq!(*upstream_hits.lock().expect("mutex should lock"), 0);
|
||||
|
||||
gateway_handle.abort();
|
||||
upstream_handle.abort();
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn gateway_handles_admin_stats_time_series_locally_with_trusted_admin_principal() {
|
||||
let (upstream_url, upstream_hits, upstream_handle) =
|
||||
|
||||
Reference in New Issue
Block a user