[codex] Add use_responses_lite 'override' logic (#26487)

## Summary

- add a defaulted `ModelInfo.use_responses_lite` catalog field
- support serializing `reasoning.context` while preserving the existing
effort and summary path
- has not been turned on for any models yet

I've added an override to parallel tools if responses_lite is on. I've
also forced persistent reasoning when using responses_lite. It would be
ideal if we could centralize all the responses_lite plumbing, but I
think this is best for now to keep the plumbing & diffs small.

## Testing

- `cargo test -p codex-protocol
model_info_defaults_availability_nux_to_none_when_omitted`
- `RUST_MIN_STACK=8388608 cargo test -p codex-core
responses_lite_sets_all_turns_context_and_disables_parallel_tool_calls`
- `RUST_MIN_STACK=8388608 cargo test -p codex-core
configured_reasoning_summary_is_sent`
- `cargo check -p codex-core --tests`
- `RUST_MIN_STACK=8388608 cargo clippy -p codex-core --tests` (passes
with pre-existing warnings in `codex-code-mode` and
`codex-core-plugins`)

codex-rs/core/src/client.rs

@@ -44,6 +44,7 @@ use codex_api::RawMemory as ApiRawMemory;
 use codex_api::RealtimeCallClient as ApiRealtimeCallClient;
 use codex_api::RealtimeSessionConfig as ApiRealtimeSessionConfig;
 use codex_api::Reasoning;
+use codex_api::ReasoningContext;
 use codex_api::RequestTelemetry;
 use codex_api::ReqwestTransport;
 use codex_api::ResponseCreateWsRequest;
@@ -609,6 +610,7 @@ impl ModelClient {
             reasoning: effort.map(|effort| Reasoning {
                 effort: Some(effort),
                 summary: None,
+                context: None,
             }),
         };
 
@@ -727,6 +729,11 @@ impl ModelClient {
                 } else {
                     Some(summary)
                 },
+                // When Responses Lite is disabled, omit context so Responses uses the default,
+                // which is currently `current_turn`.
+                context: model_info
+                    .use_responses_lite
+                    .then_some(ReasoningContext::AllTurns),
             })
         } else {
             None
@@ -775,7 +782,7 @@ impl ModelClient {
             input,
             tools,
             tool_choice: "auto".to_string(),
-            parallel_tool_calls: prompt.parallel_tool_calls,
+            parallel_tool_calls: prompt.parallel_tool_calls && !model_info.use_responses_lite,
             reasoning,
             store: provider.is_azure_responses_endpoint(),
             stream: true,

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

@@ -231,6 +231,7 @@ fn remote_model_with_auto_review_override(slug: &str, review_model: &str) -> Mod
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: Some(review_model.to_string()),
         tool_mode: None,
         multi_agent_version: None,

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

@@ -1872,6 +1872,71 @@ async fn configured_reasoning_summary_is_sent() -> anyhow::Result<()> {
             .and_then(|value| value.as_str()),
         Some("concise")
     );
+    pretty_assertions::assert_eq!(
+        request_body
+            .get("reasoning")
+            .and_then(|reasoning| reasoning.get("context")),
+        None
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn responses_lite_sets_all_turns_context_and_disables_parallel_tool_calls()
+-> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = mount_sse_once(
+        &server,
+        sse(vec![ev_response_created("resp1"), ev_completed("resp1")]),
+    )
+    .await;
+
+    let mut model_catalog = bundled_models_response()
+        .unwrap_or_else(|err| panic!("bundled models.json should parse: {err}"));
+    let model = model_catalog
+        .models
+        .iter_mut()
+        .find(|model| model.slug == "gpt-5.4")
+        .expect("gpt-5.4 exists in bundled models.json");
+    model.use_responses_lite = true;
+    model.supports_parallel_tool_calls = true;
+
+    let TestCodex { codex, .. } = test_codex()
+        .with_model("gpt-5.4")
+        .with_config(move |config| {
+            config.model_catalog = Some(model_catalog);
+        })
+        .build(&server)
+        .await?;
+
+    codex
+        .submit(Op::UserInput {
+            environments: None,
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+                text_elements: Vec::new(),
+            }],
+            final_output_json_schema: None,
+            responsesapi_client_metadata: None,
+            additional_context: Default::default(),
+            thread_settings: Default::default(),
+        })
+        .await?;
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TurnComplete(_))).await;
+
+    let request_body = resp_mock.single_request().body_json();
+    pretty_assertions::assert_eq!(
+        request_body
+            .get("reasoning")
+            .and_then(|reasoning| reasoning.get("context"))
+            .and_then(|value| value.as_str()),
+        Some("all_turns")
+    );
+    pretty_assertions::assert_eq!(request_body.get("parallel_tool_calls"), Some(&json!(false)));
 
     Ok(())
 }

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

@@ -112,6 +112,7 @@ fn test_model_info(
         input_modalities,
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,
@@ -932,6 +933,7 @@ async fn model_switch_to_smaller_model_updates_token_context_window() -> Result<
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,

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

@@ -370,6 +370,7 @@ fn test_remote_model(slug: &str, priority: i32) -> ModelInfo {
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,

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

@@ -592,6 +592,7 @@ async fn remote_model_friendly_personality_instructions_with_feature() -> anyhow
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,
@@ -705,6 +706,7 @@ async fn user_turn_personality_remote_model_template_includes_update_message() -
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,

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

@@ -478,6 +478,7 @@ async fn remote_models_remote_model_uses_unified_exec() -> Result<()> {
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,
@@ -730,6 +731,7 @@ async fn remote_models_apply_remote_base_instructions() -> Result<()> {
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,
@@ -1216,6 +1218,7 @@ fn test_remote_model_with_policy(
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,

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

@@ -1349,6 +1349,7 @@ async fn stdio_image_responses_are_sanitized_for_text_only_model() -> anyhow::Re
                 input_modalities: vec![InputModality::Text],
                 used_fallback_model_metadata: false,
                 supports_search_tool: false,
+                use_responses_lite: false,
                 auto_review_model_override: None,
                 tool_mode: None,
                 multi_agent_version: None,

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

@@ -60,6 +60,7 @@ fn test_model_info(
         input_modalities: default_input_modalities(),
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,

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

@@ -1355,6 +1355,7 @@ async fn view_image_tool_returns_unsupported_message_for_text_only_model() -> an
         input_modalities: vec![InputModality::Text],
         used_fallback_model_metadata: false,
         supports_search_tool: false,
+        use_responses_lite: false,
         auto_review_model_override: None,
         tool_mode: None,
         multi_agent_version: None,