Add git workspace metadata to turn analytics

Emit the existing per-turn git workspace metadata as a custom analytics fact and attach it to codex_turn_event params when the reducer has enough turn lifecycle state. Wait for the in-flight turn metadata enrichment before normal turn completion so short turns can still report workspace data.

This only changes the Codex client-side event payload; backend schema/table ingestion needs the matching field before downstream consumers can query it.

Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
Ningyi Xie
2026-04-24 23:25:16 -07:00
co-authored by Codex
parent f5497f4d65
commit cc694d4907
7 changed files with 191 additions and 4 deletions
@@ -48,6 +48,8 @@ use crate::facts::SkillInvokedInput;
use crate::facts::SubAgentThreadStartedInput;
use crate::facts::ThreadInitializationMode;
use crate::facts::TrackEventsContext;
use crate::facts::TurnGitMetadataFact;
use crate::facts::TurnGitWorkspaceMetadata;
use crate::facts::TurnResolvedConfigFact;
use crate::facts::TurnStatus;
use crate::facts::TurnSteerRequestError;
@@ -105,6 +107,7 @@ use codex_utils_absolute_path::test_support::PathBufExt;
use codex_utils_absolute_path::test_support::test_path_buf;
use pretty_assertions::assert_eq;
use serde_json::json;
use std::collections::BTreeMap;
use std::collections::HashSet;
use std::path::PathBuf;
use std::sync::Arc;
@@ -1771,6 +1774,7 @@ fn turn_event_serializes_expected_shape() {
subagent_tool_call_count: None,
web_search_count: None,
image_generation_count: None,
git_workspaces: None,
input_tokens: None,
cached_input_tokens: None,
output_tokens: None,
@@ -1832,6 +1836,7 @@ fn turn_event_serializes_expected_shape() {
"subagent_tool_call_count": null,
"web_search_count": null,
"image_generation_count": null,
"git_workspaces": null,
"input_tokens": null,
"cached_input_tokens": null,
"output_tokens": null,
@@ -2142,6 +2147,77 @@ async fn turn_lifecycle_emits_turn_event() {
assert_eq!(payload["event_params"]["total_tokens"], json!(321));
}
#[tokio::test]
async fn turn_lifecycle_includes_git_metadata_when_recorded() {
let mut reducer = AnalyticsReducer::default();
let mut out = Vec::new();
ingest_turn_prerequisites(
&mut reducer,
&mut out,
/*include_initialize*/ true,
/*include_resolved_config*/ true,
/*include_started*/ true,
/*include_token_usage*/ false,
)
.await;
let mut associated_remote_urls = BTreeMap::new();
associated_remote_urls.insert(
"origin".to_string(),
"git@github.com:openai/codex.git".to_string(),
);
let mut git_workspaces = BTreeMap::new();
git_workspaces.insert(
"/workspace/codex".to_string(),
TurnGitWorkspaceMetadata {
associated_remote_urls: Some(associated_remote_urls),
latest_git_commit_hash: Some("abc123".to_string()),
has_changes: Some(true),
},
);
reducer
.ingest(
AnalyticsFact::Custom(CustomAnalyticsFact::TurnGitMetadata(Box::new(
TurnGitMetadataFact {
turn_id: "turn-2".to_string(),
thread_id: "thread-2".to_string(),
git_workspaces,
},
))),
&mut out,
)
.await;
assert!(out.is_empty());
reducer
.ingest(
AnalyticsFact::Notification(Box::new(sample_turn_completed_notification(
"thread-2",
"turn-2",
AppServerTurnStatus::Completed,
/*codex_error_info*/ None,
))),
&mut out,
)
.await;
assert_eq!(out.len(), 1);
let payload = serde_json::to_value(&out[0]).expect("serialize turn event");
assert_eq!(
payload["event_params"]["git_workspaces"],
json!({
"/workspace/codex": {
"associated_remote_urls": {
"origin": "git@github.com:openai/codex.git"
},
"latest_git_commit_hash": "abc123",
"has_changes": true
}
})
);
}
#[tokio::test]
async fn accepted_steers_increment_turn_steer_count() {
let mut reducer = AnalyticsReducer::default();
+7
View File
@@ -18,6 +18,7 @@ use crate::facts::SkillInvocation;
use crate::facts::SkillInvokedInput;
use crate::facts::SubAgentThreadStartedInput;
use crate::facts::TrackEventsContext;
use crate::facts::TurnGitMetadataFact;
use crate::facts::TurnResolvedConfigFact;
use crate::facts::TurnTokenUsageFact;
use crate::reducer::AnalyticsReducer;
@@ -231,6 +232,12 @@ impl AnalyticsEventsClient {
)));
}
pub fn track_turn_git_metadata(&self, fact: TurnGitMetadataFact) {
self.record_fact(AnalyticsFact::Custom(CustomAnalyticsFact::TurnGitMetadata(
Box::new(fact),
)));
}
pub fn track_plugin_installed(&self, plugin: PluginTelemetryMetadata) {
self.record_fact(AnalyticsFact::Custom(
CustomAnalyticsFact::PluginStateChanged(PluginStateChangedInput {
+3
View File
@@ -14,6 +14,7 @@ use crate::facts::PluginState;
use crate::facts::SubAgentThreadStartedInput;
use crate::facts::ThreadInitializationMode;
use crate::facts::TrackEventsContext;
use crate::facts::TurnGitWorkspaceMetadata;
use crate::facts::TurnStatus;
use crate::facts::TurnSteerRejectionReason;
use crate::facts::TurnSteerResult;
@@ -35,6 +36,7 @@ use codex_protocol::protocol::HookSource;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::TokenUsage;
use serde::Serialize;
use std::collections::BTreeMap;
#[derive(Clone, Copy, Debug, Serialize)]
#[serde(rename_all = "snake_case")]
@@ -492,6 +494,7 @@ pub(crate) struct CodexTurnEventParams {
pub(crate) subagent_tool_call_count: Option<usize>,
pub(crate) web_search_count: Option<usize>,
pub(crate) image_generation_count: Option<usize>,
pub(crate) git_workspaces: Option<BTreeMap<String, TurnGitWorkspaceMetadata>>,
pub(crate) input_tokens: Option<i64>,
pub(crate) cached_input_tokens: Option<i64>,
pub(crate) output_tokens: Option<i64>,
+20
View File
@@ -23,7 +23,9 @@ use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SkillScope;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::TokenUsage;
use serde::Deserialize;
use serde::Serialize;
use std::collections::BTreeMap;
use std::path::PathBuf;
#[derive(Clone)]
@@ -89,6 +91,23 @@ pub struct TurnTokenUsageFact {
pub token_usage: TokenUsage,
}
#[derive(Clone, Debug, Deserialize, PartialEq, Eq, Serialize)]
pub struct TurnGitWorkspaceMetadata {
#[serde(default, skip_serializing_if = "Option::is_none")]
pub associated_remote_urls: Option<BTreeMap<String, String>>,
#[serde(default, skip_serializing_if = "Option::is_none")]
pub latest_git_commit_hash: Option<String>,
#[serde(default, skip_serializing_if = "Option::is_none")]
pub has_changes: Option<bool>,
}
#[derive(Clone)]
pub struct TurnGitMetadataFact {
pub turn_id: String,
pub thread_id: String,
pub git_workspaces: BTreeMap<String, TurnGitWorkspaceMetadata>,
}
#[derive(Clone, Copy, Debug, Serialize)]
#[serde(rename_all = "snake_case")]
pub enum TurnStatus {
@@ -298,6 +317,7 @@ pub(crate) enum CustomAnalyticsFact {
GuardianReview(Box<GuardianReviewEventParams>),
TurnResolvedConfig(Box<TurnResolvedConfigFact>),
TurnTokenUsage(Box<TurnTokenUsageFact>),
TurnGitMetadata(Box<TurnGitMetadataFact>),
SkillInvoked(SkillInvokedInput),
AppMentioned(AppMentionedInput),
AppUsed(AppUsedInput),
+2
View File
@@ -34,6 +34,8 @@ pub use facts::SkillInvocation;
pub use facts::SubAgentThreadStartedInput;
pub use facts::ThreadInitializationMode;
pub use facts::TrackEventsContext;
pub use facts::TurnGitMetadataFact;
pub use facts::TurnGitWorkspaceMetadata;
pub use facts::TurnResolvedConfigFact;
pub use facts::TurnStatus;
pub use facts::TurnSteerRejectionReason;
+38
View File
@@ -41,6 +41,8 @@ use crate::facts::PluginUsedInput;
use crate::facts::SkillInvokedInput;
use crate::facts::SubAgentThreadStartedInput;
use crate::facts::ThreadInitializationMode;
use crate::facts::TurnGitMetadataFact;
use crate::facts::TurnGitWorkspaceMetadata;
use crate::facts::TurnResolvedConfigFact;
use crate::facts::TurnStatus;
use crate::facts::TurnSteerRejectionReason;
@@ -66,6 +68,7 @@ use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SkillScope;
use codex_protocol::protocol::TokenUsage;
use sha1::Digest;
use std::collections::BTreeMap;
use std::collections::HashMap;
use std::path::Path;
@@ -149,6 +152,7 @@ struct TurnState {
resolved_config: Option<TurnResolvedConfigFact>,
started_at: Option<u64>,
token_usage: Option<TokenUsage>,
git_workspaces: Option<BTreeMap<String, TurnGitWorkspaceMetadata>>,
completed: Option<CompletedTurnState>,
steer_count: usize,
}
@@ -211,6 +215,9 @@ impl AnalyticsReducer {
CustomAnalyticsFact::TurnTokenUsage(input) => {
self.ingest_turn_token_usage(*input, out);
}
CustomAnalyticsFact::TurnGitMetadata(input) => {
self.ingest_turn_git_metadata(*input, out);
}
CustomAnalyticsFact::SkillInvoked(input) => {
self.ingest_skill_invoked(input, out).await;
}
@@ -350,6 +357,7 @@ impl AnalyticsReducer {
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
@@ -372,6 +380,7 @@ impl AnalyticsReducer {
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
@@ -380,6 +389,31 @@ impl AnalyticsReducer {
self.maybe_emit_turn_event(&turn_id, out);
}
fn ingest_turn_git_metadata(
&mut self,
input: TurnGitMetadataFact,
out: &mut Vec<TrackEventRequest>,
) {
if input.git_workspaces.is_empty() {
return;
}
let turn_id = input.turn_id.clone();
let turn_state = self.turns.entry(turn_id.clone()).or_insert(TurnState {
connection_id: None,
thread_id: None,
num_input_images: None,
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
turn_state.thread_id = Some(input.thread_id);
turn_state.git_workspaces = Some(input.git_workspaces);
self.maybe_emit_turn_event(&turn_id, out);
}
async fn ingest_skill_invoked(
&mut self,
input: SkillInvokedInput,
@@ -533,6 +567,7 @@ impl AnalyticsReducer {
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
@@ -615,6 +650,7 @@ impl AnalyticsReducer {
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
@@ -634,6 +670,7 @@ impl AnalyticsReducer {
resolved_config: None,
started_at: None,
token_usage: None,
git_workspaces: None,
completed: None,
steer_count: 0,
});
@@ -933,6 +970,7 @@ fn codex_turn_event_params(
subagent_tool_call_count: None,
web_search_count: None,
image_generation_count: None,
git_workspaces: turn_state.git_workspaces.clone(),
input_tokens: token_usage
.as_ref()
.map(|token_usage| token_usage.input_tokens),
+45 -4
View File
@@ -5,6 +5,7 @@ mod review;
mod undo;
mod user_shell;
use std::collections::BTreeMap;
use std::sync::Arc;
use std::time::Duration;
use std::time::Instant;
@@ -31,6 +32,8 @@ use crate::session::turn_context::TurnContext;
use crate::state::ActiveTurn;
use crate::state::RunningTask;
use crate::state::TaskKind;
use codex_analytics::TurnGitMetadataFact;
use codex_analytics::TurnGitWorkspaceMetadata;
use codex_analytics::TurnTokenUsageFact;
use codex_login::AuthManager;
use codex_models_manager::manager::SharedModelsManager;
@@ -151,6 +154,44 @@ fn bool_tag(value: bool) -> &'static str {
if value { "true" } else { "false" }
}
fn git_workspaces_from_turn_metadata(
metadata: serde_json::Value,
) -> Option<BTreeMap<String, TurnGitWorkspaceMetadata>> {
let workspaces = metadata.get("workspaces")?.as_object()?;
let git_workspaces = workspaces
.iter()
.filter_map(|(workspace_path, workspace_metadata)| {
serde_json::from_value::<TurnGitWorkspaceMetadata>(workspace_metadata.clone())
.ok()
.map(|metadata| (workspace_path.clone(), metadata))
})
.collect::<BTreeMap<_, _>>();
if git_workspaces.is_empty() {
None
} else {
Some(git_workspaces)
}
}
fn track_turn_git_metadata_analytics(sess: &Session, turn_context: &TurnContext) {
if !sess.enabled(Feature::GeneralAnalytics) {
return;
}
let Some(metadata) = turn_context.turn_metadata_state.current_meta_value() else {
return;
};
let Some(git_workspaces) = git_workspaces_from_turn_metadata(metadata) else {
return;
};
sess.services
.analytics_events_client
.track_turn_git_metadata(TurnGitMetadataFact {
turn_id: turn_context.sub_id.clone(),
thread_id: sess.conversation_id.to_string(),
git_workspaces,
});
}
/// Thin wrapper that exposes the parts of [`Session`] task runners need.
#[derive(Clone)]
pub(crate) struct SessionTaskContext {
@@ -539,10 +580,6 @@ impl Session {
turn_context: Arc<TurnContext>,
last_agent_message: Option<String>,
) {
turn_context
.turn_metadata_state
.cancel_git_enrichment_task();
let mut pending_input = Vec::<ResponseInputItem>::new();
let mut should_clear_active_turn = false;
let mut token_usage_at_turn_start = None;
@@ -691,6 +728,10 @@ impl Session {
{
warn!("failed to apply goal runtime turn-finished event: {err}");
}
track_turn_git_metadata_analytics(self, turn_context.as_ref());
turn_context
.turn_metadata_state
.cancel_git_enrichment_task();
let event = EventMsg::TurnComplete(TurnCompleteEvent {
turn_id: turn_context.sub_id.clone(),
last_agent_message,