Add Guardian v2 approval review metrics (#39224)

## What changed

- Pass metrics bound to the active turn's model to approval-review contributors.
- Add counter support to `ExtensionMetrics` while preserving host session attribution.
- Record Guardian v2 tool-call score lag and count stale-score fallbacks with the `score_lag` reason.

## Testing

- Cover counter forwarding and session metadata tags.
- Verify Guardian v2 metric samples for current and stale scores.

GitOrigin-RevId: 0ec34736d29d2e37ed217a27e773b4f6230ae38a
This commit is contained in:
Dylan Hurd
2026-08-18 16:27:54 +00:00
committed by copyberry
parent 2a30972fcb
commit e7e13c68e2
15 changed files with 230 additions and 16 deletions
+3
View File
@@ -330,6 +330,9 @@ async fn run_guardian_review(
&session.services.session_extension_data,
&session.services.thread_extension_data,
&action.to_string(),
Some(crate::session::extension_metrics::from_session_telemetry(
turn.session_telemetry.clone(),
)),
)
.await
{
+1
View File
@@ -2999,6 +2999,7 @@ async fn escalated_retry_bypasses_extension_approval_and_runs_guardian() -> anyh
_session_store: &'a codex_extension_api::ExtensionData,
_thread_store: &'a codex_extension_api::ExtensionData,
_prompt: &'a str,
_extension_metrics: Option<Arc<dyn codex_extension_api::ExtensionMetrics>>,
) -> codex_extension_api::ExtensionFuture<'a, Option<ReviewDecision>> {
Box::pin(async move { Some(ReviewDecision::Approved) })
}
@@ -8,12 +8,16 @@ struct SessionTelemetryExtensionMetrics {
}
impl ExtensionMetrics for SessionTelemetryExtensionMetrics {
fn counter(&self, name: &str, inc: i64, tags: &[(&str, &str)]) {
self.session_telemetry.counter(name, inc, tags);
}
fn histogram(&self, name: &str, value: i64, tags: &[(&str, &str)]) {
self.session_telemetry.histogram(name, value, tags);
}
}
pub(super) fn from_session_telemetry(
pub(crate) fn from_session_telemetry(
session_telemetry: SessionTelemetry,
) -> Arc<dyn ExtensionMetrics> {
Arc::new(SessionTelemetryExtensionMetrics { session_telemetry })
+1 -1
View File
@@ -204,7 +204,7 @@ use codex_protocol::exec_output::StreamOutput;
mod code_mode_warning;
pub(crate) mod context_window;
mod environment;
mod extension_metrics;
pub(crate) mod extension_metrics;
mod handlers;
mod inject;
mod input_queue;
+23
View File
@@ -402,8 +402,31 @@ fn extension_metrics_preserve_session_metadata_tags() {
],
);
extension_metrics.counter(
"codex.test.extension.counter",
/*inc*/ 2,
&[("component", "skills"), ("model", "extension-model")],
);
let snapshot = metrics.snapshot().expect("metrics snapshot");
let attributes = single_histogram_attributes(&snapshot, "codex.test.extension");
let counter = find_metric(&snapshot, "codex.test.extension.counter");
let AggregatedMetrics::U64(MetricData::Sum(sum)) = counter.data() else {
panic!("expected counter");
};
let points = sum.data_points().collect::<Vec<_>>();
assert_eq!(points.len(), 1);
assert_eq!(points[0].value(), 2);
assert_eq!(
points[0]
.attributes()
.map(|attribute| (
attribute.key.as_str().to_string(),
attribute.value.as_str().to_string(),
))
.collect::<BTreeMap<_, _>>(),
attributes,
);
assert_eq!(
attributes,
BTreeMap::from([
+3
View File
@@ -6,6 +6,7 @@ use anyhow::Result;
use codex_extension_api::ApprovalReviewContributor;
use codex_extension_api::ExtensionData;
use codex_extension_api::ExtensionFuture;
use codex_extension_api::ExtensionMetrics;
use codex_extension_api::ExtensionRegistryBuilder;
use codex_features::Feature;
use codex_login::CodexAuth;
@@ -64,8 +65,10 @@ impl ApprovalReviewContributor for ApprovedReviewContributor {
_session_store: &'a ExtensionData,
_thread_store: &'a ExtensionData,
prompt: &'a str,
extension_metrics: Option<Arc<dyn ExtensionMetrics>>,
) -> ExtensionFuture<'a, Option<ReviewDecision>> {
Box::pin(async move {
assert!(extension_metrics.is_some());
assert!(prompt.contains("\"tool\":\"request_permissions\""));
Some(ReviewDecision::Approved)
})