Merge branch 'fawney19:main' into main

This commit is contained in:
ZheFox
2026-05-19 00:50:52 +08:00
committed by GitHub
29 changed files with 795 additions and 79 deletions

View File

@@ -1059,6 +1059,71 @@ fn admin_usage_upstream_is_stream(item: &StoredRequestUsageAudit) -> bool {
.unwrap_or(item.is_stream)
}
fn admin_usage_metadata_string<'a>(
item: &'a StoredRequestUsageAudit,
key: &str,
) -> Option<&'a str> {
item.request_metadata
.as_ref()
.and_then(Value::as_object)
.and_then(|metadata| metadata.get(key))
.and_then(Value::as_str)
.map(str::trim)
.filter(|value| !value.is_empty())
}
fn infer_client_family_from_user_agent(user_agent: &str) -> Option<&'static str> {
let normalized = user_agent.trim().to_ascii_lowercase();
if normalized.is_empty() {
return None;
}
if normalized.starts_with("codex_vscode") {
return Some("codex_vscode");
}
if normalized.starts_with("codex") {
return Some("codex");
}
if normalized.contains("claude-code") || normalized.contains("claude_code") {
return Some("claude_code");
}
if normalized.contains("opencode") {
return Some("opencode");
}
if normalized.contains("geminicli") || normalized.contains("gemini-cli") {
return Some("gemini_cli");
}
if normalized.starts_with("openai/js") {
return Some("openai_js_sdk");
}
None
}
pub fn admin_usage_client_family(item: &StoredRequestUsageAudit) -> Option<&str> {
item.client_family
.as_deref()
.map(str::trim)
.filter(|value| !value.is_empty())
.or_else(|| {
item.request_metadata
.as_ref()
.and_then(Value::as_object)
.and_then(|metadata| {
metadata
.get("client_session_affinity")
.and_then(Value::as_object)
.and_then(|affinity| affinity.get("client_family"))
.and_then(Value::as_str)
.or_else(|| metadata.get("client_family").and_then(Value::as_str))
})
.map(str::trim)
.filter(|value| !value.is_empty())
})
.or_else(|| {
admin_usage_metadata_string(item, "user_agent")
.and_then(infer_client_family_from_user_agent)
})
}
fn admin_usage_active_request_json(
item: &StoredRequestUsageAudit,
api_key_name: Option<String>,
@@ -1091,6 +1156,11 @@ fn admin_usage_active_request_json(
"upstream_is_stream": upstream_is_stream,
"client_requested_stream": client_is_stream,
"client_is_stream": client_is_stream,
"client_family": admin_usage_client_family(item),
"client_ip": admin_usage_metadata_string(item, "client_ip"),
"user_agent": admin_usage_metadata_string(item, "user_agent"),
"request_path": admin_usage_metadata_string(item, "request_path"),
"request_path_and_query": admin_usage_metadata_string(item, "request_path_and_query"),
"has_fallback": admin_usage_has_fallback(item),
});
if let Some(api_format) = item.api_format.as_ref() {
@@ -1192,6 +1262,27 @@ pub fn admin_usage_record_json(
json!(client_is_stream),
);
object.insert("client_is_stream".to_string(), json!(client_is_stream));
maybe_insert_string_field(object, "client_family", admin_usage_client_family(item));
maybe_insert_string_field(
object,
"client_ip",
admin_usage_metadata_string(item, "client_ip"),
);
maybe_insert_string_field(
object,
"user_agent",
admin_usage_metadata_string(item, "user_agent"),
);
maybe_insert_string_field(
object,
"request_path",
admin_usage_metadata_string(item, "request_path"),
);
maybe_insert_string_field(
object,
"request_path_and_query",
admin_usage_metadata_string(item, "request_path_and_query"),
);
payload
}
@@ -2443,6 +2534,74 @@ mod tests {
assert_eq!(record["client_is_stream"], false);
}
#[test]
fn admin_usage_record_infers_client_family_from_user_agent() {
let item = StoredRequestUsageAudit {
request_metadata: Some(json!({
"client_ip": "192.168.0.28",
"user_agent": "codex_vscode/0.131.0-alpha.9 (Windows 10.0.26200; x86_64)"
})),
..sample_usage("completed", Some(200), None)
};
let record = admin_usage_record_json(
&item,
&BTreeMap::new(),
&BTreeMap::new(),
false,
false,
None,
);
let active = admin_usage_active_request_json(&item, None, None, None);
assert_eq!(record["client_family"], "codex_vscode");
assert_eq!(record["client_ip"], "192.168.0.28");
assert_eq!(active["client_family"], "codex_vscode");
}
#[test]
fn admin_usage_record_labels_openai_js_user_agent_as_sdk() {
let item = StoredRequestUsageAudit {
request_metadata: Some(json!({
"user_agent": "OpenAI/JS 6.34.0"
})),
..sample_usage("completed", Some(200), None)
};
let record = admin_usage_record_json(
&item,
&BTreeMap::new(),
&BTreeMap::new(),
false,
false,
None,
);
assert_eq!(record["client_family"], "openai_js_sdk");
}
#[test]
fn admin_usage_record_prefers_typed_client_family() {
let item = StoredRequestUsageAudit {
client_family: Some("codex".to_string()),
request_metadata: Some(json!({
"user_agent": "OpenAI/JS 6.34.0"
})),
..sample_usage("completed", Some(200), None)
};
let record = admin_usage_record_json(
&item,
&BTreeMap::new(),
&BTreeMap::new(),
false,
false,
None,
);
assert_eq!(record["client_family"], "codex");
}
#[test]
fn client_requested_stream_uses_stream_generate_content_path_over_stale_metadata_flag() {
let item = StoredRequestUsageAudit {

View File

@@ -6,6 +6,10 @@ use chrono::{TimeZone, Utc};
use serde_json::{json, Value};
use std::collections::BTreeMap;
pub fn normalize_endpoint_api_format(api_format: &str) -> String {
aether_ai_formats::normalize_api_format_alias(api_format)
}
fn unix_secs_to_rfc3339(unix_secs: u64) -> Option<String> {
Utc.timestamp_opt(unix_secs as i64, 0)
.single()
@@ -45,7 +49,7 @@ pub fn key_api_formats_without_entry(
fn active_endpoint_api_formats(endpoints: &[StoredProviderCatalogEndpoint]) -> Vec<String> {
let mut formats = Vec::new();
for endpoint in endpoints.iter().filter(|endpoint| endpoint.is_active) {
let api_format = aether_ai_formats::normalize_api_format_alias(&endpoint.api_format);
let api_format = normalize_endpoint_api_format(&endpoint.api_format);
if !formats.iter().any(|existing| existing == &api_format) {
formats.push(api_format);
}
@@ -63,7 +67,7 @@ fn configured_key_api_formats(key: &StoredProviderCatalogKey) -> Vec<String> {
};
let mut normalized = Vec::new();
for api_format in formats.iter().filter_map(serde_json::Value::as_str) {
let api_format = aether_ai_formats::normalize_api_format_alias(api_format);
let api_format = normalize_endpoint_api_format(api_format);
if !normalized.iter().any(|existing| existing == &api_format) {
normalized.push(api_format);
}

View File

@@ -1,8 +1,8 @@
mod types;
pub use types::{
parse_usage_body_ref, usage_body_ref, PendingUsageCleanupSummary,
ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
parse_usage_body_ref, usage_body_ref, usage_request_metadata_client_family,
PendingUsageCleanupSummary, ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
StoredProviderApiKeyWindowUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
StoredRequestUsageAudit, StoredUsageAuditAggregation, StoredUsageAuditSummary,
StoredUsageBreakdownSummaryRow, StoredUsageCacheAffinityHitSummary,

View File

@@ -34,6 +34,8 @@ pub struct StoredRequestUsageAudit {
pub provider_endpoint_kind: Option<String>,
pub has_format_conversion: bool,
pub is_stream: bool,
#[serde(default, skip_serializing_if = "Option::is_none")]
pub client_family: Option<String>,
pub input_tokens: u64,
pub output_tokens: u64,
pub total_tokens: u64,
@@ -211,6 +213,7 @@ impl StoredRequestUsageAudit {
provider_endpoint_kind,
has_format_conversion,
is_stream,
client_family: None,
input_tokens: parse_u64(input_tokens, "usage.input_tokens")?,
output_tokens: parse_u64(output_tokens, "usage.output_tokens")?,
total_tokens: parse_u64(total_tokens, "usage.total_tokens")?,
@@ -311,6 +314,10 @@ impl StoredRequestUsageAudit {
.filter(|value| !value.is_empty())
}
pub fn request_metadata_client_family(&self) -> Option<&str> {
usage_request_metadata_client_family(self.request_metadata.as_ref())
}
fn billing_snapshot_resolved_number(&self, key: &str) -> Option<f64> {
self.request_metadata_object()
.and_then(|metadata| metadata.get("billing_snapshot"))
@@ -1794,6 +1801,18 @@ pub struct UsageCleanupPreviewCounts {
pub log: u64,
}
pub fn usage_request_metadata_client_family(value: Option<&Value>) -> Option<&str> {
let metadata = value.and_then(Value::as_object)?;
metadata
.get("client_session_affinity")
.and_then(Value::as_object)
.and_then(|affinity| affinity.get("client_family"))
.and_then(Value::as_str)
.or_else(|| metadata.get("client_family").and_then(Value::as_str))
.map(str::trim)
.filter(|value| !value.is_empty())
}
fn parse_u64(value: i32, field_name: &str) -> Result<u64, crate::DataLayerError> {
u64::try_from(value).map_err(|_| {
crate::DataLayerError::UnexpectedValue(format!("invalid {field_name}: {value}"))

View File

@@ -29,11 +29,12 @@ use serde_json::Value;
use super::{
api_key_usage_contribution, provider_api_key_usage_contribution,
strip_deprecated_usage_display_fields, usage_can_recover_terminal_failure,
ApiKeyUsageContribution, ApiKeyUsageDelta, ProviderApiKeyUsageContribution,
ProviderApiKeyUsageDelta, ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
StoredProviderApiKeyWindowUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
StoredRequestUsageAudit, StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery,
UsageDailyHeatmapQuery, UsageReadRepository, UsageWriteRepository,
usage_request_metadata_client_family, ApiKeyUsageContribution, ApiKeyUsageDelta,
ProviderApiKeyUsageContribution, ProviderApiKeyUsageDelta, ProviderApiKeyWindowUsageRequest,
StoredProviderApiKeyUsageSummary, StoredProviderApiKeyWindowUsageSummary,
StoredProviderUsageSummary, StoredProviderUsageWindow, StoredRequestUsageAudit,
StoredUsageDailySummary, UpsertUsageRecord, UsageAuditListQuery, UsageDailyHeatmapQuery,
UsageReadRepository, UsageWriteRepository,
};
use crate::repository::auth::InMemoryAuthApiKeySnapshotRepository;
use crate::repository::provider_catalog::InMemoryProviderCatalogReadRepository;
@@ -56,6 +57,7 @@ impl InMemoryUsageReadRepository {
let mut by_request_id = BTreeMap::new();
for mut item in items {
hydrate_legacy_body_refs(&mut item);
hydrate_client_family(&mut item);
by_request_id.insert(item.request_id.clone(), item);
}
Self {
@@ -75,6 +77,7 @@ impl InMemoryUsageReadRepository {
let mut detached_bodies = BTreeMap::new();
for mut item in items {
hydrate_legacy_body_refs(&mut item);
hydrate_client_family(&mut item);
let request_id = item.request_id.clone();
if let Some(body_ref) = detach_usage_body(
&request_id,
@@ -2546,6 +2549,13 @@ fn hydrate_legacy_body_refs(item: &mut StoredRequestUsageAudit) {
}
}
fn hydrate_client_family(item: &mut StoredRequestUsageAudit) {
if item.client_family.is_none() {
item.client_family = usage_request_metadata_client_family(item.request_metadata.as_ref())
.map(ToOwned::to_owned);
}
}
fn persisted_usage_body_ref(
incoming_ref: Option<&str>,
incoming_body: Option<&Value>,
@@ -2851,6 +2861,13 @@ impl UsageWriteRepository for InMemoryUsageReadRepository {
})
},
),
client_family: usage_request_metadata_client_family(request_metadata.as_ref())
.map(ToOwned::to_owned)
.or_else(|| {
existing
.as_ref()
.and_then(|existing| existing.client_family.clone())
}),
request_metadata,
created_at_unix_ms,
updated_at_unix_secs: usage.updated_at_unix_secs,
@@ -4399,6 +4416,7 @@ mod tests {
provider_endpoint_kind: Some("chat".to_string()),
has_format_conversion: false,
is_stream: false,
client_family: None,
input_tokens: 10,
output_tokens: 20,
total_tokens: 30,

View File

@@ -363,7 +363,8 @@ mod sqlite;
#[allow(unused_imports)]
pub(crate) use aether_data_contracts::repository::usage::{
PendingUsageCleanupSummary, ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
usage_request_metadata_client_family, PendingUsageCleanupSummary,
ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
StoredProviderApiKeyWindowUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
StoredRequestUsageAudit, StoredUsageAuditAggregation, StoredUsageAuditSummary,
StoredUsageBreakdownSummaryRow, StoredUsageCacheAffinityHitSummary,

View File

@@ -6,8 +6,8 @@ use sqlx::{mysql::MySqlRow, Row};
use super::{
provider_api_key_usage_is_error, provider_api_key_usage_is_success,
strip_deprecated_usage_display_fields, usage_can_recover_terminal_failure,
InMemoryUsageReadRepository, PendingUsageCleanupSummary, StoredRequestUsageAudit,
UpsertUsageRecord, UsageWriteRepository,
usage_request_metadata_client_family, InMemoryUsageReadRepository, PendingUsageCleanupSummary,
StoredRequestUsageAudit, UpsertUsageRecord, UsageWriteRepository,
};
use crate::driver::mysql::MysqlPool;
use crate::error::SqlResultExt;
@@ -850,6 +850,8 @@ fn map_usage_row(row: &MySqlRow) -> Result<StoredRequestUsageAudit, DataLayerErr
.map(|raw| serde_json::from_str(&raw))
.transpose()
.map_err(|err| DataLayerError::UnexpectedValue(err.to_string()))?;
audit.client_family = usage_request_metadata_client_family(audit.request_metadata.as_ref())
.map(ToOwned::to_owned);
let upstream_is_stream = row
.try_get::<Option<bool>, _>("upstream_is_stream")
.map_sql_err()?;

View File

@@ -8591,6 +8591,9 @@ fn map_usage_row(
.try_get::<f64, _>("cache_read_cost_usd")
.map_postgres_err()?;
usage.output_price_per_1m = row.try_get("output_price_per_1m").map_postgres_err()?;
usage.client_family = row
.try_get::<Option<String>, _>("client_family")
.map_postgres_err()?;
usage.request_headers = row.try_get("request_headers").map_postgres_err()?;
let request_body = usage_json_column(
row,

View File

@@ -93,6 +93,10 @@ SELECT
"usage".first_byte_time_ms,
"usage".status,
COALESCE(usage_settlement_snapshots.billing_status, "usage".billing_status) AS billing_status,
COALESCE(
NULLIF(BTRIM("usage".request_metadata->'client_session_affinity'->>'client_family'), ''),
NULLIF(BTRIM("usage".request_metadata->>'client_family'), '')
) AS client_family,
COALESCE(usage_http_audits.request_headers, "usage".request_headers) AS request_headers,
"usage".request_body,
"usage".request_body_compressed,

View File

@@ -93,6 +93,10 @@ SELECT
"usage".first_byte_time_ms,
"usage".status,
COALESCE(usage_settlement_snapshots.billing_status, "usage".billing_status) AS billing_status,
COALESCE(
NULLIF(BTRIM("usage".request_metadata->'client_session_affinity'->>'client_family'), ''),
NULLIF(BTRIM("usage".request_metadata->>'client_family'), '')
) AS client_family,
COALESCE(usage_http_audits.request_headers, "usage".request_headers) AS request_headers,
"usage".request_body,
"usage".request_body_compressed,

View File

@@ -93,6 +93,10 @@ SELECT
"usage".first_byte_time_ms,
"usage".status,
COALESCE(usage_settlement_snapshots.billing_status, "usage".billing_status) AS billing_status,
COALESCE(
NULLIF(BTRIM("usage".request_metadata->'client_session_affinity'->>'client_family'), ''),
NULLIF(BTRIM("usage".request_metadata->>'client_family'), '')
) AS client_family,
NULL::json AS request_headers,
NULL::json AS request_body,
NULL::bytea AS request_body_compressed,
@@ -106,9 +110,21 @@ SELECT
NULL::json AS client_response_body,
NULL::bytea AS client_response_body_compressed,
CASE
WHEN ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
WHEN NULLIF(BTRIM("usage".request_metadata->>'client_ip'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'user_agent'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'request_path'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'request_path_and_query'), '') IS NOT NULL
OR ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
OR ("usage".request_metadata->>'upstream_is_stream') IN ('true', 'false')
THEN jsonb_build_object(
THEN jsonb_strip_nulls(jsonb_build_object(
'client_ip',
NULLIF(BTRIM("usage".request_metadata->>'client_ip'), ''),
'user_agent',
NULLIF(BTRIM("usage".request_metadata->>'user_agent'), ''),
'request_path',
NULLIF(BTRIM("usage".request_metadata->>'request_path'), ''),
'request_path_and_query',
NULLIF(BTRIM("usage".request_metadata->>'request_path_and_query'), ''),
'client_requested_stream',
CASE
WHEN ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
@@ -121,7 +137,7 @@ SELECT
THEN ("usage".request_metadata->>'upstream_is_stream')::boolean
ELSE NULL
END
)::json
))::json
ELSE NULL::json
END AS request_metadata,
NULL::varchar AS http_request_body_ref,

View File

@@ -93,6 +93,10 @@ SELECT
"usage".first_byte_time_ms,
"usage".status,
COALESCE(usage_settlement_snapshots.billing_status, "usage".billing_status) AS billing_status,
COALESCE(
NULLIF(BTRIM("usage".request_metadata->'client_session_affinity'->>'client_family'), ''),
NULLIF(BTRIM("usage".request_metadata->>'client_family'), '')
) AS client_family,
NULL::json AS request_headers,
NULL::json AS request_body,
NULL::bytea AS request_body_compressed,
@@ -106,9 +110,21 @@ SELECT
NULL::json AS client_response_body,
NULL::bytea AS client_response_body_compressed,
CASE
WHEN ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
WHEN NULLIF(BTRIM("usage".request_metadata->>'client_ip'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'user_agent'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'request_path'), '') IS NOT NULL
OR NULLIF(BTRIM("usage".request_metadata->>'request_path_and_query'), '') IS NOT NULL
OR ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
OR ("usage".request_metadata->>'upstream_is_stream') IN ('true', 'false')
THEN jsonb_build_object(
THEN jsonb_strip_nulls(jsonb_build_object(
'client_ip',
NULLIF(BTRIM("usage".request_metadata->>'client_ip'), ''),
'user_agent',
NULLIF(BTRIM("usage".request_metadata->>'user_agent'), ''),
'request_path',
NULLIF(BTRIM("usage".request_metadata->>'request_path'), ''),
'request_path_and_query',
NULLIF(BTRIM("usage".request_metadata->>'request_path_and_query'), ''),
'client_requested_stream',
CASE
WHEN ("usage".request_metadata->>'client_requested_stream') IN ('true', 'false')
@@ -121,7 +137,7 @@ SELECT
THEN ("usage".request_metadata->>'upstream_is_stream')::boolean
ELSE NULL
END
)::json
))::json
ELSE NULL::json
END AS request_metadata,
NULL::varchar AS http_request_body_ref,

View File

@@ -633,6 +633,14 @@ fn usage_sql_uses_json_null_placeholders_for_usage_payload_columns() {
super::LIST_USAGE_AUDITS_PREFIX,
super::LIST_RECENT_USAGE_AUDITS_PREFIX,
] {
assert!(sql.contains("jsonb_strip_nulls(jsonb_build_object("));
assert!(sql.contains("'client_ip'"));
assert!(sql.contains("request_metadata->>'client_ip'"));
assert!(sql.contains("'user_agent'"));
assert!(sql.contains("request_metadata->>'user_agent'"));
assert!(sql.contains("AS client_family"));
assert!(sql.contains("request_metadata->'client_session_affinity'->>'client_family'"));
assert!(sql.contains("request_metadata->>'client_family'"));
assert!(sql.contains("CAST(\"usage\".input_tokens AS INTEGER) AS input_tokens"));
assert!(sql.contains(
"usage_settlement_snapshots.billing_input_tokens AS settlement_billing_input_tokens"

View File

@@ -8,7 +8,8 @@ use sqlx::{sqlite::SqliteRow, QueryBuilder, Row, Sqlite};
use super::{
strip_deprecated_usage_display_fields, usage_can_recover_terminal_failure,
PendingUsageCleanupSummary, ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
usage_request_metadata_client_family, PendingUsageCleanupSummary,
ProviderApiKeyWindowUsageRequest, StoredProviderApiKeyUsageSummary,
StoredProviderApiKeyWindowUsageSummary, StoredProviderUsageSummary, StoredRequestUsageAudit,
StoredUsageAuditAggregation, StoredUsageAuditSummary, StoredUsageBreakdownSummaryRow,
StoredUsageCacheAffinityHitSummary, StoredUsageCacheAffinityIntervalRow,
@@ -3812,6 +3813,8 @@ fn map_usage_row(row: &SqliteRow) -> Result<StoredRequestUsageAudit, DataLayerEr
.map(|raw| serde_json::from_str(&raw))
.transpose()
.map_err(|err| DataLayerError::UnexpectedValue(err.to_string()))?;
audit.client_family = usage_request_metadata_client_family(audit.request_metadata.as_ref())
.map(ToOwned::to_owned);
let upstream_is_stream = row
.try_get::<Option<i64>, _>("upstream_is_stream")
.map_sql_err()?

View File

@@ -72,8 +72,10 @@ fn copy_allowed_metadata_fields(source: &Map<String, Value>, target: &mut Map<St
copy_non_empty_string(source, target, "trace_id");
copy_non_empty_string(source, target, "client_ip");
copy_non_empty_string(source, target, "user_agent");
copy_non_empty_string(source, target, "client_family");
copy_bool(source, target, "client_requested_stream");
copy_bool(source, target, "upstream_is_stream");
copy_non_null_value(source, target, "client_session_affinity");
copy_bool(source, target, "api_key_is_standalone");
copy_non_empty_string(source, target, "request_path");
copy_non_empty_string(source, target, "request_query_string");
@@ -110,8 +112,10 @@ fn move_allowed_metadata_fields(mut source: Map<String, Value>, target: &mut Map
remove_non_empty_string(&mut source, target, "trace_id");
remove_non_empty_string(&mut source, target, "client_ip");
remove_non_empty_string(&mut source, target, "user_agent");
remove_non_empty_string(&mut source, target, "client_family");
remove_bool(&mut source, target, "client_requested_stream");
remove_bool(&mut source, target, "upstream_is_stream");
remove_non_null_value(&mut source, target, "client_session_affinity");
remove_bool(&mut source, target, "api_key_is_standalone");
remove_non_empty_string(&mut source, target, "request_path");
remove_non_empty_string(&mut source, target, "request_query_string");