add sandbox policy and sandbox name to codex.tool.call metrics (#10711)

This will give visibility into the comparative success rate of the
Windows sandbox implementations compared to other platforms.

codex-rs/core/src/exec.rs

@@ -128,6 +128,17 @@ pub enum SandboxType {
     WindowsRestrictedToken,
 }
 
+impl SandboxType {
+    pub(crate) fn as_metric_tag(self) -> &'static str {
+        match self {
+            SandboxType::None => "none",
+            SandboxType::MacosSeatbelt => "seatbelt",
+            SandboxType::LinuxSeccomp => "seccomp",
+            SandboxType::WindowsRestrictedToken => "windows_sandbox",
+        }
+    }
+}
+
 #[derive(Clone)]
 pub struct StdoutStream {
     pub sub_id: String,

codex-rs/core/src/tools/registry.rs

@@ -3,11 +3,15 @@ use std::sync::Arc;
 use std::time::Duration;
 
 use crate::client_common::tools::ToolSpec;
+use crate::exec::SandboxType;
 use crate::function_tool::FunctionCallError;
+use crate::protocol::SandboxPolicy;
+use crate::safety::get_platform_sandbox;
 use crate::tools::context::ToolInvocation;
 use crate::tools::context::ToolOutput;
 use crate::tools::context::ToolPayload;
 use async_trait::async_trait;
+use codex_protocol::config_types::WindowsSandboxLevel;
 use codex_protocol::models::ResponseInputItem;
 use codex_utils_readiness::Readiness;
 use tracing::warn;
@@ -73,19 +77,33 @@ impl ToolRegistry {
         let otel = invocation.turn.otel_manager.clone();
         let payload_for_response = invocation.payload.clone();
         let log_payload = payload_for_response.log_payload();
+        let metric_tags = [
+            (
+                "sandbox",
+                sandbox_tag(
+                    &invocation.turn.sandbox_policy,
+                    invocation.turn.windows_sandbox_level,
+                ),
+            ),
+            (
+                "sandbox_policy",
+                sandbox_policy_tag(&invocation.turn.sandbox_policy),
+            ),
+        ];
 
         let handler = match self.handler(tool_name.as_ref()) {
             Some(handler) => handler,
             None => {
                 let message =
                     unsupported_tool_call_message(&invocation.payload, tool_name.as_ref());
-                otel.tool_result(
+                otel.tool_result_with_tags(
                     tool_name.as_ref(),
                     &call_id_owned,
                     log_payload.as_ref(),
                     Duration::ZERO,
                     false,
                     &message,
+                    &metric_tags,
                 );
                 return Err(FunctionCallError::RespondToModel(message));
             }
@@ -93,13 +111,14 @@ impl ToolRegistry {
 
         if !handler.matches_kind(&invocation.payload) {
             let message = format!("tool {tool_name} invoked with incompatible payload");
-            otel.tool_result(
+            otel.tool_result_with_tags(
                 tool_name.as_ref(),
                 &call_id_owned,
                 log_payload.as_ref(),
                 Duration::ZERO,
                 false,
                 &message,
+                &metric_tags,
             );
             return Err(FunctionCallError::Fatal(message));
         }
@@ -107,10 +126,11 @@ impl ToolRegistry {
         let output_cell = tokio::sync::Mutex::new(None);
 
         let result = otel
-            .log_tool_result(
+            .log_tool_result_with_tags(
                 tool_name.as_ref(),
                 &call_id_owned,
                 log_payload.as_ref(),
+                &metric_tags,
                 || {
                     let handler = handler.clone();
                     let output_cell = &output_cell;
@@ -231,3 +251,29 @@ fn unsupported_tool_call_message(payload: &ToolPayload, tool_name: &str) -> Stri
         _ => format!("unsupported call: {tool_name}"),
     }
 }
+
+fn sandbox_tag(policy: &SandboxPolicy, windows_sandbox_level: WindowsSandboxLevel) -> &'static str {
+    if matches!(policy, SandboxPolicy::DangerFullAccess) {
+        return "none";
+    }
+    if matches!(policy, SandboxPolicy::ExternalSandbox { .. }) {
+        return "external";
+    }
+    if cfg!(target_os = "windows") && matches!(windows_sandbox_level, WindowsSandboxLevel::Elevated)
+    {
+        return "windows_elevated";
+    }
+
+    get_platform_sandbox(windows_sandbox_level != WindowsSandboxLevel::Disabled)
+        .map(SandboxType::as_metric_tag)
+        .unwrap_or("none")
+}
+
+fn sandbox_policy_tag(policy: &SandboxPolicy) -> &'static str {
+    match policy {
+        SandboxPolicy::ReadOnly => "read-only",
+        SandboxPolicy::WorkspaceWrite { .. } => "workspace-write",
+        SandboxPolicy::DangerFullAccess => "danger-full-access",
+        SandboxPolicy::ExternalSandbox { .. } => "external-sandbox",
+    }
+}

codex-rs/otel/src/traces/otel_manager.rs

@@ -566,11 +566,12 @@ impl OtelManager {
         );
     }
 
-    pub async fn log_tool_result<F, Fut, E>(
+    pub async fn log_tool_result_with_tags<F, Fut, E>(
         &self,
         tool_name: &str,
         call_id: &str,
         arguments: &str,
+        extra_tags: &[(&str, &str)],
         f: F,
     ) -> Result<(String, bool), E>
     where
@@ -587,13 +588,14 @@ impl OtelManager {
             Err(error) => (Cow::Owned(error.to_string()), false),
         };
 
-        self.tool_result(
+        self.tool_result_with_tags(
             tool_name,
             call_id,
             arguments,
             duration,
             success,
             output.as_ref(),
+            extra_tags,
         );
 
         result
@@ -619,7 +621,8 @@ impl OtelManager {
         );
     }
 
-    pub fn tool_result(
+    #[allow(clippy::too_many_arguments)]
+    pub fn tool_result_with_tags(
         &self,
         tool_name: &str,
         call_id: &str,
@@ -627,18 +630,15 @@ impl OtelManager {
         duration: Duration,
         success: bool,
         output: &str,
+        extra_tags: &[(&str, &str)],
     ) {
         let success_str = if success { "true" } else { "false" };
-        self.counter(
+        let mut tags = Vec::with_capacity(2 + extra_tags.len());
-            TOOL_CALL_COUNT_METRIC,
+        tags.push(("tool", tool_name));
-            1,
+        tags.push(("success", success_str));
-            &[("tool", tool_name), ("success", success_str)],
+        tags.extend_from_slice(extra_tags);
-        );
+        self.counter(TOOL_CALL_COUNT_METRIC, 1, &tags);
-        self.record_duration(
+        self.record_duration(TOOL_CALL_DURATION_METRIC, duration, &tags);
-            TOOL_CALL_DURATION_METRIC,
-            duration,
-            &[("tool", tool_name), ("success", success_str)],
-        );
         tracing::event!(
             tracing::Level::INFO,
             event.name = "codex.tool_result",

codex-rs/otel/tests/suite/runtime_summary.rs

@@ -35,13 +35,14 @@ fn runtime_metrics_summary_collects_tool_api_and_streaming_metrics() -> Result<(
 
     manager.reset_runtime_metrics();
 
-    manager.tool_result(
+    manager.tool_result_with_tags(
         "shell",
         "call-1",
         "{\"cmd\":\"echo\"}",
         Duration::from_millis(250),
         true,
         "ok",
+        &[],
     );
     manager.record_api_request(1, Some(200), None, Duration::from_millis(300));
     manager.record_websocket_request(Duration::from_millis(400), None);