Add exclude output to streamed Responses requests

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

@@ -179,6 +179,8 @@ pub struct ResponsesApiRequest {
     pub store: bool,
     pub stream: bool,
     pub include: Vec<String>,
+    #[serde(skip_serializing_if = "Vec::is_empty")]
+    pub exclude: Vec<String>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub service_tier: Option<String>,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -203,6 +205,7 @@ impl From<&ResponsesApiRequest> for ResponseCreateWsRequest {
             store: request.store,
             stream: request.stream,
             include: request.include.clone(),
+            exclude: request.exclude.clone(),
             service_tier: request.service_tier.clone(),
             prompt_cache_key: request.prompt_cache_key.clone(),
             text: request.text.clone(),
@@ -227,6 +230,8 @@ pub struct ResponseCreateWsRequest {
     pub store: bool,
     pub stream: bool,
     pub include: Vec<String>,
+    #[serde(skip_serializing_if = "Vec::is_empty")]
+    pub exclude: Vec<String>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub service_tier: Option<String>,
     #[serde(skip_serializing_if = "Option::is_none")]

codex-rs/codex-api/tests/clients.rs

@@ -331,6 +331,7 @@ async fn streaming_client_retries_on_transport_error() -> Result<()> {
         store: false,
         stream: true,
         include: Vec::new(),
+        exclude: Vec::new(),
         service_tier: None,
         prompt_cache_key: None,
         text: None,
@@ -432,6 +433,7 @@ async fn azure_default_store_attaches_ids_and_headers() -> Result<()> {
         store: true,
         stream: true,
         include: Vec::new(),
+        exclude: Vec::new(),
         service_tier: None,
         prompt_cache_key: None,
         text: None,

codex-rs/core/src/client.rs

@@ -760,6 +760,7 @@ impl ModelClient {
             store: provider.is_azure_responses_endpoint(),
             stream: true,
             include,
+            exclude: vec!["output".to_string()],
             service_tier,
             prompt_cache_key,
             text,

codex-rs/core/src/client_common_tests.rs

@@ -22,6 +22,7 @@ fn serializes_text_verbosity_when_set() {
         store: false,
         stream: true,
         include: vec![],
+        exclude: vec![],
         prompt_cache_key: None,
         service_tier: None,
         text: Some(TextControls {
@@ -69,6 +70,7 @@ fn serializes_text_schema_with_strict_format() {
         store: false,
         stream: true,
         include: vec![],
+        exclude: vec![],
         prompt_cache_key: None,
         service_tier: None,
         text: Some(text_controls),
@@ -130,6 +132,7 @@ fn omits_text_when_not_set() {
         store: false,
         stream: true,
         include: vec![],
+        exclude: vec![],
         prompt_cache_key: None,
         service_tier: None,
         text: None,
@@ -140,6 +143,30 @@ fn omits_text_when_not_set() {
     assert!(v.get("text").is_none());
 }
 
+#[test]
+fn omits_exclude_when_not_set() {
+    let req = ResponsesApiRequest {
+        model: "gpt-5.1".to_string(),
+        instructions: "i".to_string(),
+        input: vec![],
+        tools: vec![],
+        tool_choice: "auto".to_string(),
+        parallel_tool_calls: true,
+        reasoning: None,
+        store: false,
+        stream: true,
+        include: vec![],
+        exclude: vec![],
+        prompt_cache_key: None,
+        service_tier: None,
+        text: None,
+        client_metadata: None,
+    };
+
+    let v = serde_json::to_value(&req).expect("json");
+    assert!(v.get("exclude").is_none());
+}
+
 #[test]
 fn serializes_flex_service_tier_when_set() {
     let req = ResponsesApiRequest {
@@ -153,6 +180,7 @@ fn serializes_flex_service_tier_when_set() {
         store: false,
         stream: true,
         include: vec![],
+        exclude: vec![],
         prompt_cache_key: None,
         service_tier: Some(ServiceTier::Flex.to_string()),
         text: None,

codex-rs/core/src/client_tests.rs

@@ -23,6 +23,8 @@ use codex_model_provider_info::create_oss_provider_with_base_url;
 use codex_otel::SessionTelemetry;
 use codex_protocol::SessionId;
 use codex_protocol::ThreadId;
+use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
+use codex_protocol::models::BaseInstructions;
 use codex_protocol::models::ContentItem;
 use codex_protocol::models::ResponseItem;
 use codex_protocol::openai_models::ModelInfo;
@@ -123,6 +125,43 @@ fn test_session_telemetry() -> SessionTelemetry {
     )
 }
 
+#[tokio::test]
+async fn build_responses_request_excludes_completed_output() {
+    let session = test_model_client(SessionSource::Cli).new_session();
+    let prompt = super::Prompt {
+        input: Vec::new(),
+        tools: Vec::new(),
+        parallel_tool_calls: false,
+        base_instructions: BaseInstructions {
+            text: "base instructions".to_string(),
+        },
+        personality: None,
+        output_schema: None,
+        output_schema_strict: false,
+    };
+    let provider = session
+        .client
+        .state
+        .provider
+        .api_provider()
+        .await
+        .expect("resolve api provider");
+
+    let request = session
+        .client
+        .build_responses_request(
+            &provider,
+            &prompt,
+            &test_model_info(),
+            /*effort*/ None,
+            ReasoningSummaryConfig::None,
+            /*service_tier*/ None,
+        )
+        .expect("build responses request");
+
+    assert_eq!(request.exclude, vec!["output".to_string()]);
+}
+
 #[derive(Default)]
 struct TagCollectorVisitor {
     tags: BTreeMap<String, String>,

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

@@ -68,6 +68,10 @@ async fn responses_mode_stream_cli() {
 
     let request = resp_mock.single_request();
     assert_eq!(request.path(), "/v1/responses");
+    assert_eq!(
+        request.body_json()["exclude"],
+        serde_json::json!(["output"])
+    );
 
     // TODO(jif) fix
     // // Verify a new session rollout was created and is discoverable via list_conversations