Expose response usage metadata in completion events (#41087)

## What changed

- Parse optional `usage_metadata.amount` values from Responses API completion events without converting their string representation.
- Propagate the metadata through SSE and WebSocket streams, regular turns, and remote compaction.
- Add `usageMetadata` to the app-server `rawResponse/completed` notification and its generated schemas.

## Testing

- Cover missing, null, zero, and high-precision amounts, including completion events emitted during compaction.

GitOrigin-RevId: b16d81e8350eab0dc7ff613d0dc9dbed73041713

codex-rs/app-server-protocol/schema/json/codex_app_server_protocol.schemas.json

@@ -17253,6 +17253,16 @@
                 "type": "null"
               }
             ]
+          },
+          "usageMetadata": {
+            "anyOf": [
+              {
+                "$ref": "#/definitions/v2/ResponseUsageMetadata"
+              },
+              {
+                "type": "null"
+              }
+            ]
           }
         },
         "required": [
@@ -18501,6 +18511,18 @@
           }
         ]
       },
+      "ResponseUsageMetadata": {
+        "description": "Usage metadata reported for one upstream response.",
+        "properties": {
+          "amount": {
+            "type": [
+              "string",
+              "null"
+            ]
+          }
+        },
+        "type": "object"
+      },
       "ResponsesApiWebSearchAction": {
         "oneOf": [
           {

codex-rs/app-server-protocol/schema/json/codex_app_server_protocol.v2.schemas.json

@@ -13386,6 +13386,16 @@
               "type": "null"
             }
           ]
+        },
+        "usageMetadata": {
+          "anyOf": [
+            {
+              "$ref": "#/definitions/ResponseUsageMetadata"
+            },
+            {
+              "type": "null"
+            }
+          ]
         }
       },
       "required": [
@@ -14634,6 +14644,18 @@
         }
       ]
     },
+    "ResponseUsageMetadata": {
+      "description": "Usage metadata reported for one upstream response.",
+      "properties": {
+        "amount": {
+          "type": [
+            "string",
+            "null"
+          ]
+        }
+      },
+      "type": "object"
+    },
     "ResponsesApiWebSearchAction": {
       "oneOf": [
         {

codex-rs/app-server-protocol/schema/json/v2/RawResponseCompletedNotification.json

@@ -1,6 +1,18 @@
 {
   "$schema": "http://json-schema.org/draft-07/schema#",
   "definitions": {
+    "ResponseUsageMetadata": {
+      "description": "Usage metadata reported for one upstream response.",
+      "properties": {
+        "amount": {
+          "type": [
+            "string",
+            "null"
+          ]
+        }
+      },
+      "type": "object"
+    },
     "TokenUsageBreakdown": {
       "properties": {
         "cacheWriteInputTokens": {
@@ -59,6 +71,16 @@
           "type": "null"
         }
       ]
+    },
+    "usageMetadata": {
+      "anyOf": [
+        {
+          "$ref": "#/definitions/ResponseUsageMetadata"
+        },
+        {
+          "type": "null"
+        }
+      ]
     }
   },
   "required": [

codex-rs/app-server-protocol/schema/typescript/v2/RawResponseCompletedNotification.ts

@@ -1,10 +1,11 @@
 // GENERATED CODE! DO NOT MODIFY BY HAND!
 
 // This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
+import type { ResponseUsageMetadata } from "./ResponseUsageMetadata";
 import type { TokenUsageBreakdown } from "./TokenUsageBreakdown";
 
 /**
  * Internal-only notification containing the exact usage from one upstream
  * Responses API completion.
  */
-export type RawResponseCompletedNotification = { threadId: string, turnId: string, responseId: string, usage: TokenUsageBreakdown | null, };
+export type RawResponseCompletedNotification = { threadId: string, turnId: string, responseId: string, usage: TokenUsageBreakdown | null, usageMetadata: ResponseUsageMetadata | null, };

codex-rs/app-server-protocol/schema/typescript/v2/ResponseUsageMetadata.ts

@@ -0,0 +1,8 @@
+// GENERATED CODE! DO NOT MODIFY BY HAND!
+
+// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
+
+/**
+ * Usage metadata reported for one upstream response.
+ */
+export type ResponseUsageMetadata = { amount: string | null, };

codex-rs/app-server-protocol/schema/typescript/v2/index.ts

@@ -408,6 +408,7 @@ export type { RemoteControlEnableParams } from "./RemoteControlEnableParams";
 export type { RemoteControlStatusChangedNotification } from "./RemoteControlStatusChangedNotification";
 export type { RequestPermissionProfile } from "./RequestPermissionProfile";
 export type { ResidencyRequirement } from "./ResidencyRequirement";
+export type { ResponseUsageMetadata } from "./ResponseUsageMetadata";
 export type { ReviewDelivery } from "./ReviewDelivery";
 export type { ReviewStartParams } from "./ReviewStartParams";
 export type { ReviewStartResponse } from "./ReviewStartResponse";

codex-rs/app-server-protocol/src/protocol/v2/thread.rs

@@ -1841,6 +1841,23 @@ pub struct RawResponseCompletedNotification {
     pub turn_id: String,
     pub response_id: String,
     pub usage: Option<TokenUsageBreakdown>,
+    pub usage_metadata: Option<ResponseUsageMetadata>,
+}
+
+/// Usage metadata reported for one upstream response.
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ResponseUsageMetadata {
+    pub amount: Option<String>,
+}
+
+impl From<codex_protocol::ResponseUsageMetadata> for ResponseUsageMetadata {
+    fn from(value: codex_protocol::ResponseUsageMetadata) -> Self {
+        Self {
+            amount: value.amount,
+        }
+    }
 }
 
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]

codex-rs/app-server/src/bespoke_event_handling.rs

@@ -1118,6 +1118,7 @@ pub(crate) async fn apply_bespoke_event_handling(
                 turn_id: event_turn_id,
                 response_id: raw_response_completed_event.response_id,
                 usage: raw_response_completed_event.token_usage.map(Into::into),
+                usage_metadata: raw_response_completed_event.usage_metadata.map(Into::into),
             };
             outgoing
                 .send_server_notification(ServerNotification::RawResponseCompleted(notification))

codex-rs/app-server/tests/suite/v2/compaction.rs

@@ -15,6 +15,7 @@ use codex_app_server_protocol::ItemStartedNotification;
 use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::RawResponseCompletedNotification;
 use codex_app_server_protocol::RequestId;
+use codex_app_server_protocol::ResponseUsageMetadata;
 use codex_app_server_protocol::ThreadCompactStartParams;
 use codex_app_server_protocol::ThreadCompactStartResponse;
 use codex_app_server_protocol::ThreadItem;
@@ -244,9 +245,11 @@ async fn thread_compact_start_triggers_compaction_and_returns_empty_response() -
     skip_if_no_network!(Ok(()));
 
     let server = responses::start_mock_server().await;
+    let mut completed = responses::ev_completed_with_tokens("r1", /*total_tokens*/ 200);
+    completed["response"]["usage_metadata"] = serde_json::json!({ "amount": "0.125" });
     let sse = responses::sse(vec![
         responses::ev_assistant_message("m1", "MANUAL_COMPACT_SUMMARY"),
-        responses::ev_completed_with_tokens("r1", /*total_tokens*/ 200),
+        completed,
     ]);
     responses::mount_sse_sequence(&server, vec![sse]).await;
 
@@ -300,6 +303,9 @@ async fn thread_compact_start_triggers_compaction_and_returns_empty_response() -
             thread_id,
             turn_id: started.turn_id,
             response_id: "r1".to_string(),
+            usage_metadata: Some(ResponseUsageMetadata {
+                amount: Some("0.125".to_string()),
+            }),
             usage: Some(TokenUsageBreakdown {
                 total_tokens: 200,
                 input_tokens: 200,

codex-rs/app-server/tests/suite/v2/turn_start.rs

@@ -799,27 +799,43 @@ async fn turn_start_sends_service_tier_id_to_model_request() -> Result<()> {
     Ok(())
 }
 
+#[test_case(None, json!(null); "without_usage_metadata")]
+#[test_case(Some(json!({})), json!({ "amount": null }); "without_amount")]
+#[test_case(Some(json!({ "amount": null })), json!({ "amount": null }); "null_amount")]
+#[test_case(Some(json!({ "amount": "0" })), json!({ "amount": "0" }); "zero_amount")]
+#[test_case(
+    Some(json!({ "amount": "0.12345678901234567890" })),
+    json!({ "amount": "0.12345678901234567890" });
+    "exact_amount"
+)]
 #[tokio::test]
-async fn turn_start_emits_raw_response_completed_with_upstream_usage() -> Result<()> {
+async fn turn_start_emits_raw_response_completed_with_upstream_usage(
+    upstream_metadata: Option<Value>,
+    expected_metadata: Value,
+) -> Result<()> {
     let server = responses::start_mock_server().await;
+    let mut completed = json!({
+        "type": "response.completed",
+        "response": {
+            "id": "resp-1",
+            "usage": {
+                "input_tokens": 30,
+                "input_tokens_details": { "cached_tokens": 11 },
+                "output_tokens": 7,
+                "output_tokens_details": { "reasoning_tokens": 3 },
+                "total_tokens": 37
+            }
+        }
+    });
+    if let Some(metadata) = upstream_metadata {
+        completed["response"]["usage_metadata"] = metadata;
+    }
     let body = responses::sse(vec![
         responses::ev_response_created("resp-1"),
         responses::ev_assistant_message("msg-1", "Done"),
-        json!({
+        completed,
-            "type": "response.completed",
-            "response": {
-                "id": "resp-1",
-                "usage": {
-                    "input_tokens": 30,
-                    "input_tokens_details": { "cached_tokens": 11 },
-                    "output_tokens": 7,
-                    "output_tokens_details": { "reasoning_tokens": 3 },
-                    "total_tokens": 37
-                }
-            }
-        }),
     ]);
-    responses::mount_sse_once(&server, body).await;
+    let response_mock = responses::mount_sse_once(&server, body).await;
 
     let codex_home = TempDir::new()?;
     MockResponsesConfig::new(&server.uri()).write(codex_home.path())?;
@@ -856,6 +872,13 @@ async fn turn_start_emits_raw_response_completed_with_upstream_usage() -> Result
         mcp.read_stream_until_notification_message("rawResponse/completed"),
     )
     .await??;
+    assert_eq!(
+        notification
+            .params
+            .as_ref()
+            .and_then(|params| params.get("usageMetadata")),
+        Some(&expected_metadata),
+    );
     let notification: codex_app_server_protocol::ServerNotification = notification.try_into()?;
     let codex_app_server_protocol::ServerNotification::RawResponseCompleted(notification) =
         notification
@@ -869,6 +892,7 @@ async fn turn_start_emits_raw_response_completed_with_upstream_usage() -> Result
             thread_id: thread.id,
             turn_id: turn.id,
             response_id: "resp-1".to_string(),
+            usage_metadata: serde_json::from_value(expected_metadata)?,
             usage: Some(TokenUsageBreakdown {
                 total_tokens: 37,
                 input_tokens: 30,
@@ -880,6 +904,7 @@ async fn turn_start_emits_raw_response_completed_with_upstream_usage() -> Result
         }
     );
 
+    response_mock.single_request();
     Ok(())
 }
 

codex-rs/codex-api/src/common.rs

@@ -1,4 +1,5 @@
 use crate::error::ApiError;
+use codex_protocol::ResponseUsageMetadata;
 use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
 use codex_protocol::config_types::Verbosity as VerbosityConfig;
 use codex_protocol::models::ResponseItem;
@@ -113,6 +114,7 @@ pub enum ResponseEvent {
     Completed {
         response_id: String,
         token_usage: Option<TokenUsage>,
+        usage_metadata: Option<ResponseUsageMetadata>,
         /// Did the model affirmatively end its turn? Some providers do not set this,
         /// so we rely on fallback logic when this is `None`.
         end_turn: Option<bool>,

codex-rs/codex-api/src/sse/responses.rs

@@ -8,6 +8,7 @@ use crate::safety_buffering::treatment_from_headers;
 use crate::telemetry::SseTelemetry;
 use codex_client::ByteStream;
 use codex_client::StreamResponse;
+use codex_protocol::ResponseUsageMetadata;
 use codex_protocol::models::ResponseItem;
 use codex_protocol::protocol::MisalignmentErrorDetails;
 use codex_protocol::protocol::ModelVerification;
@@ -118,6 +119,7 @@ struct ResponseCompleted {
     id: String,
     #[serde(default)]
     usage: Option<ResponseCompletedUsage>,
+    usage_metadata: Option<ResponseUsageMetadata>,
     #[serde(default)]
     end_turn: Option<bool>,
 }
@@ -481,6 +483,7 @@ pub fn process_responses_event(
                         return Ok(Some(ResponseEvent::Completed {
                             response_id: resp.id,
                             token_usage: resp.usage.map(Into::into),
+                            usage_metadata: resp.usage_metadata,
                             end_turn: resp.end_turn,
                         }));
                     }
@@ -864,10 +867,12 @@ mod tests {
             Ok(ResponseEvent::Completed {
                 response_id,
                 token_usage,
+                usage_metadata,
                 end_turn,
             }) => {
                 assert_eq!(response_id, "resp1");
                 assert!(token_usage.is_none());
+                assert!(usage_metadata.is_none());
                 assert!(end_turn.is_none());
             }
             other => panic!("unexpected third event: {other:?}"),
@@ -1060,10 +1065,12 @@ mod tests {
             Ok(ResponseEvent::Completed {
                 response_id,
                 token_usage,
+                usage_metadata,
                 end_turn,
             }) => {
                 assert_eq!(response_id, "resp1");
                 assert!(token_usage.is_none());
+                assert!(usage_metadata.is_none());
                 assert!(end_turn.is_none());
             }
             other => panic!("unexpected event: {other:?}"),
@@ -1570,6 +1577,7 @@ mod tests {
             ResponseEvent::Completed {
                 response_id,
                 token_usage: None,
+                usage_metadata: None,
                 end_turn: None,
             } if response_id == "resp-1"
         );
@@ -1607,6 +1615,7 @@ mod tests {
             ResponseEvent::Completed {
                 response_id,
                 token_usage: None,
+                usage_metadata: None,
                 end_turn: None,
             } if response_id == "resp-1"
         );
@@ -1642,6 +1651,7 @@ mod tests {
             ResponseEvent::Completed {
                 response_id,
                 token_usage: None,
+                usage_metadata: None,
                 end_turn: None,
             } if response_id == "resp-1"
         );
@@ -1677,6 +1687,7 @@ mod tests {
             ResponseEvent::Completed {
                 response_id,
                 token_usage: None,
+                usage_metadata: None,
                 end_turn: None,
             } if response_id == "resp-1"
         );

codex-rs/codex-api/tests/sse_end_to_end.rs

@@ -113,6 +113,7 @@ async fn responses_stream_parses_items_and_completed_end_to_end() -> Result<()>
         "type": "response.completed",
         "response": {
             "id": "resp1",
+            "usage_metadata": { "amount": "0.12345678901234567890" },
             "usage": {
                 "input_tokens": 10,
                 "output_tokens": 5,
@@ -165,9 +166,16 @@ async fn responses_stream_parses_items_and_completed_end_to_end() -> Result<()>
         ResponseEvent::Completed {
             response_id,
             token_usage,
+            usage_metadata,
             end_turn,
         } => {
             assert_eq!(response_id, "resp1");
+            assert_eq!(
+                usage_metadata,
+                &Some(codex_protocol::ResponseUsageMetadata {
+                    amount: Some("0.12345678901234567890".to_string()),
+                })
+            );
             assert_eq!(
                 token_usage.as_ref().map(|usage| usage.total_tokens),
                 Some(15)

codex-rs/core/src/client.rs

@@ -2184,6 +2184,7 @@ where
                 Ok(ResponseEvent::Completed {
                     response_id,
                     token_usage,
+                    usage_metadata,
                     end_turn,
                 }) => {
                     feedback_tags!(last_model_response_id = &response_id);
@@ -2206,6 +2207,7 @@ where
                         .send(Ok(ResponseEvent::Completed {
                             response_id,
                             token_usage,
+                            usage_metadata,
                             end_turn,
                         }))
                         .await

codex-rs/core/src/client_tests.rs

@@ -728,6 +728,7 @@ async fn response_stream_records_last_model_feedback_ids() {
         Ok(ResponseEvent::Completed {
             response_id: "resp-123".to_string(),
             token_usage: None,
+            usage_metadata: None,
             end_turn: Some(true),
         }),
     ]);

codex-rs/core/src/compact.rs

@@ -774,6 +774,7 @@ async fn drain_to_completed(
             Ok(ResponseEvent::Completed {
                 response_id,
                 token_usage,
+                usage_metadata,
                 ..
             }) => {
                 sess.send_event(
@@ -781,6 +782,7 @@ async fn drain_to_completed(
                     EventMsg::RawResponseCompleted(RawResponseCompletedEvent {
                         response_id,
                         token_usage: token_usage.clone(),
+                        usage_metadata,
                     }),
                 )
                 .await;

codex-rs/core/src/compact_remote_v2.rs

@@ -38,6 +38,7 @@ use codex_context_fragments::to_annotated_content;
 use codex_features::Feature;
 use codex_history::CodexHarnessMetadata;
 use codex_history::ResponseItemEnvelope;
+use codex_protocol::ResponseUsageMetadata;
 use codex_protocol::error::CodexErr;
 use codex_protocol::error::CodexErrorDetails;
 use codex_protocol::error::Result as CodexResult;
@@ -361,6 +362,7 @@ struct RemoteCompactionV2Output {
     compaction_output: ResponseItem,
     response_id: String,
     token_usage: Option<TokenUsage>,
+    usage_metadata: Option<ResponseUsageMetadata>,
 }
 
 async fn run_remote_compaction_request_v2(
@@ -422,6 +424,7 @@ async fn collect_compaction_output(
     let mut saw_completed = false;
     let mut completed_response_id = None;
     let mut completed_token_usage = None;
+    let mut completed_usage_metadata = None;
     while let Some(event) = stream.next().await {
         match event? {
             ResponseEvent::OutputItemDone(item) => {
@@ -436,11 +439,13 @@ async fn collect_compaction_output(
             ResponseEvent::Completed {
                 response_id,
                 token_usage,
+                usage_metadata,
                 ..
             } => {
                 saw_completed = true;
                 completed_response_id = Some(response_id);
                 completed_token_usage = token_usage;
+                completed_usage_metadata = usage_metadata;
                 break;
             }
             _ => {}
@@ -469,6 +474,7 @@ async fn collect_compaction_output(
         compaction_output,
         response_id,
         token_usage: completed_token_usage,
+        usage_metadata: completed_usage_metadata,
     })
 }
 
@@ -1161,6 +1167,9 @@ mod tests {
                     total_tokens: 123_498,
                     codex_rollout_budget_units: None,
                 }),
+                usage_metadata: Some(codex_protocol::ResponseUsageMetadata {
+                    amount: Some("0.125".to_string()),
+                }),
                 end_turn: Some(true),
             }),
         ]);
@@ -1169,6 +1178,12 @@ mod tests {
             .await
             .expect("compaction should be collected");
 
+        assert_eq!(
+            output.usage_metadata,
+            Some(codex_protocol::ResponseUsageMetadata {
+                amount: Some("0.125".to_string()),
+            }),
+        );
         assert_eq!(output.compaction_output, compaction);
         assert_eq!(output.response_id, "resp-compact");
         assert_eq!(

codex-rs/core/src/compact_remote_v2_attempt.rs

@@ -119,6 +119,7 @@ pub(super) async fn run_remote_compact_v2_attempt(
         compaction_output,
         response_id,
         token_usage,
+        usage_metadata,
     } = compaction_output_result?;
     // TODO: Emit this before compaction output validation so malformed completed
     // responses still surface their raw upstream usage.
@@ -127,6 +128,7 @@ pub(super) async fn run_remote_compact_v2_attempt(
         EventMsg::RawResponseCompleted(RawResponseCompletedEvent {
             response_id,
             token_usage: token_usage.clone(),
+            usage_metadata,
         }),
     )
     .await;

codex-rs/core/src/session/turn.rs

@@ -2565,6 +2565,7 @@ async fn try_run_sampling_request(
             ResponseEvent::Completed {
                 response_id,
                 token_usage,
+                usage_metadata,
                 end_turn,
             } => {
                 sess.services
@@ -2589,6 +2590,7 @@ async fn try_run_sampling_request(
                     EventMsg::RawResponseCompleted(RawResponseCompletedEvent {
                         response_id,
                         token_usage: token_usage.clone(),
+                        usage_metadata,
                     }),
                 )
                 .await;

codex-rs/core/tests/suite/client_websockets.rs

@@ -152,6 +152,53 @@ fn websocket_connection_metadata(harness: &WebsocketTestHarness) -> CodexRespons
     )
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn responses_websocket_preserves_credit_usage_metadata() {
+    skip_if_no_network!();
+
+    let mut completed = ev_completed("resp-1");
+    completed["response"]["usage_metadata"] = json!({ "amount": "0.12345678901234567890" });
+    let server =
+        start_websocket_server(vec![vec![vec![ev_response_created("resp-1"), completed]]]).await;
+    let harness = websocket_harness_for_codex_backend(&server).await;
+    let mut client_session = harness.client.new_session();
+    let prompt = prompt_with_input(vec![message_item("hello")]);
+    let responses_metadata = turn_metadata(&harness, /*turn_id*/ None);
+    let mut stream = client_session
+        .stream(
+            &prompt,
+            &harness.model_info,
+            &harness.session_telemetry,
+            harness.effort.clone(),
+            harness.summary,
+            /*service_tier*/ None,
+            &responses_metadata,
+            &InferenceTraceContext::disabled(),
+        )
+        .await
+        .expect("websocket stream failed");
+
+    let mut usage_metadata = None;
+    while let Some(event) = stream.next().await {
+        if let ResponseEvent::Completed {
+            usage_metadata: metadata,
+            ..
+        } = event.expect("websocket stream failed")
+        {
+            usage_metadata = metadata;
+            break;
+        }
+    }
+    assert_eq!(
+        usage_metadata,
+        Some(codex_protocol::ResponseUsageMetadata {
+            amount: Some("0.12345678901234567890".to_string()),
+        }),
+    );
+    assert_eq!(server.single_connection().len(), 1);
+    server.shutdown().await;
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn responses_websocket_streams_request() {
     skip_if_no_network!();

codex-rs/protocol/src/lib.rs

@@ -2,6 +2,7 @@ pub mod account;
 mod agent_path;
 pub mod auth;
 mod response_item_id;
+mod response_usage;
 mod sanitized_git_url;
 mod session_id;
 mod thread_id;
@@ -10,6 +11,7 @@ pub use agent_path::AgentPath;
 pub use permission_profile_intersection::PermissionIntersectionError;
 pub use permission_profile_intersection::intersect_effective_permission_profiles;
 pub use response_item_id::ResponseItemId;
+pub use response_usage::ResponseUsageMetadata;
 pub use sanitized_git_url::SanitizedGitUrl;
 pub use session_id::SessionId;
 pub use thread_id::RolloutId;

codex-rs/protocol/src/protocol.rs

@@ -1882,13 +1882,14 @@ pub struct RawResponseItemEvent {
     pub item: ResponseItem,
 }
 
-/// Exact usage reported by one upstream Responses API completion.
+/// Exact usage and metadata reported by one upstream Responses API completion.
 ///
 /// Unlike TokenCountEvent, this is not accumulated, estimated, or replayed.
 #[derive(Debug, Clone, Deserialize, Serialize, TS, JsonSchema)]
 pub struct RawResponseCompletedEvent {
     pub response_id: String,
     pub token_usage: Option<TokenUsage>,
+    pub usage_metadata: Option<crate::ResponseUsageMetadata>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize, TS, JsonSchema)]

codex-rs/protocol/src/response_usage.rs

@@ -0,0 +1,12 @@
+//! Per-response usage metadata reported by the upstream service, without aggregation.
+
+use schemars::JsonSchema;
+use serde::Deserialize;
+use serde::Serialize;
+use ts_rs::TS;
+
+/// Usage metadata reported for one upstream response.
+#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, Eq, TS, JsonSchema)]
+pub struct ResponseUsageMetadata {
+    pub amount: Option<String>,
+}