support 'flex' tier in app-server in addition to 'fast' (#13391)

codex-rs/app-server-protocol/schema/json/ClientRequest.json

@@ -1728,7 +1728,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

codex-rs/app-server-protocol/schema/json/EventMsg.json

@@ -5460,7 +5460,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -12134,7 +12134,8 @@
       },
       "ServiceTier": {
         "enum": [
-          "fast"
+          "fast",
+          "flex"
         ],
         "type": "string"
       },

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

@@ -10770,7 +10770,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -707,7 +707,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -53,7 +53,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     }

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

@@ -744,7 +744,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -759,7 +759,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -744,7 +744,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -78,7 +78,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     }

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

@@ -744,7 +744,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

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

@@ -305,7 +305,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },

codex-rs/app-server-protocol/schema/typescript/ServiceTier.ts

@@ -2,4 +2,4 @@
 
 // This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
 
-export type ServiceTier = "fast";
+export type ServiceTier = "fast" | "flex";

codex-rs/app-server/README.md

@@ -180,6 +180,7 @@ Start a fresh thread when you need a new Codex conversation.
     "approvalPolicy": "never",
     "sandbox": "workspaceWrite",
     "personality": "friendly",
+    "serviceTier": "flex",
     "serviceName": "my_app_server_client", // optional metrics tag (`service_name`)
     // Experimental: requires opt-in
     "dynamicTools": [
@@ -208,6 +209,7 @@ Start a fresh thread when you need a new Codex conversation.
 ```
 
 Valid `personality` values are `"friendly"`, `"pragmatic"`, and `"none"`. When `"none"` is selected, the personality placeholder is replaced with an empty string.
+Valid `serviceTier` values are `"fast"` and `"flex"`.
 
 To continue a stored session, call `thread/resume` with the `thread.id` you previously recorded. The response shape matches `thread/start`, and no additional notifications are emitted. You can also pass the same configuration overrides supported by `thread/start`, such as `personality`:
 
@@ -415,6 +417,7 @@ You can optionally specify config overrides on the new turn. If specified, these
         "networkAccess": true
     },
     "model": "gpt-5.1-codex",
+    "serviceTier": "flex",
     "effort": "medium",
     "summary": "concise",
     "personality": "friendly",

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

@@ -12,6 +12,7 @@ use codex_app_server_protocol::ThreadStartedNotification;
 use codex_app_server_protocol::ThreadStatus;
 use codex_app_server_protocol::ThreadStatusChangedNotification;
 use codex_core::config::set_project_trust_level;
+use codex_protocol::config_types::ServiceTier;
 use codex_protocol::config_types::TrustLevel;
 use codex_protocol::openai_models::ReasoningEffort;
 use pretty_assertions::assert_eq;
@@ -180,6 +181,34 @@ model_reasoning_effort = "high"
     Ok(())
 }
 
+#[tokio::test]
+async fn thread_start_accepts_flex_service_tier() -> Result<()> {
+    let server = create_mock_responses_server_repeating_assistant("Done").await;
+
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let req_id = mcp
+        .send_thread_start_request(ThreadStartParams {
+            service_tier: Some(Some(ServiceTier::Flex)),
+            ..Default::default()
+        })
+        .await?;
+
+    let resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(req_id)),
+    )
+    .await??;
+    let ThreadStartResponse { service_tier, .. } = to_response::<ThreadStartResponse>(resp)?;
+
+    assert_eq!(service_tier, Some(ServiceTier::Flex));
+    Ok(())
+}
+
 #[tokio::test]
 async fn thread_start_accepts_metrics_service_name() -> Result<()> {
     let server = create_mock_responses_server_repeating_assistant("Done").await;

codex-rs/core/config.schema.json

@@ -1336,7 +1336,8 @@
     },
     "ServiceTier": {
       "enum": [
-        "fast"
+        "fast",
+        "flex"
       ],
       "type": "string"
     },
@@ -2146,7 +2147,7 @@
           "$ref": "#/definitions/ServiceTier"
         }
       ],
-      "description": "Optional explicit service tier preference for new turns."
+      "description": "Optional explicit service tier preference for new turns (`fast` or `flex`)."
     },
     "shell_environment_policy": {
       "allOf": [

codex-rs/core/src/client.rs

@@ -546,7 +546,8 @@ impl ModelClientSession {
             include,
             service_tier: match service_tier {
                 Some(ServiceTier::Fast) => Some("priority".to_string()),
-                _ => None,
+                Some(service_tier) => Some(service_tier.to_string()),
+                None => None,
             },
             prompt_cache_key,
             text,

codex-rs/core/src/client_common.rs

@@ -234,6 +234,7 @@ mod tests {
     use codex_api::common::OpenAiVerbosity;
     use codex_api::common::TextControls;
     use codex_api::create_text_param_for_request;
+    use codex_protocol::config_types::ServiceTier;
     use codex_protocol::models::FunctionCallOutputPayload;
     use pretty_assertions::assert_eq;
 
@@ -342,6 +343,31 @@ mod tests {
         assert!(v.get("text").is_none());
     }
 
+    #[test]
+    fn serializes_flex_service_tier_when_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![],
+            prompt_cache_key: None,
+            service_tier: Some(ServiceTier::Flex.to_string()),
+            text: None,
+        };
+
+        let v = serde_json::to_value(&req).expect("json");
+        assert_eq!(
+            v.get("service_tier").and_then(|tier| tier.as_str()),
+            Some("flex")
+        );
+    }
+
     #[test]
     fn reserializes_shell_outputs_for_function_and_custom_tool_calls() {
         let raw_output = r#"{"output":"hello","metadata":{"exit_code":0,"duration_seconds":0.5}}"#;

codex-rs/core/src/config/mod.rs

@@ -189,7 +189,7 @@ pub struct Config {
     /// Optional override of model selection.
     pub model: Option<String>,
 
-    /// Effective service tier preference for new turns.
+    /// Effective service tier preference for new turns (`fast` or `flex`).
     pub service_tier: Option<ServiceTier>,
 
     /// Model used specifically for review sessions.
@@ -1191,7 +1191,7 @@ pub struct ConfigToml {
     /// Optionally specify a personality for the model
     pub personality: Option<Personality>,
 
-    /// Optional explicit service tier preference for new turns.
+    /// Optional explicit service tier preference for new turns (`fast` or `flex`).
     pub service_tier: Option<ServiceTier>,
 
     /// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
@@ -1984,15 +1984,14 @@ impl Config {
         let forced_login_method = cfg.forced_login_method;
 
         let model = model.or(config_profile.model).or(cfg.model);
-        let service_tier = if features.enabled(Feature::FastMode) {
-            service_tier_override.unwrap_or_else(|| {
-                config_profile
-                    .service_tier
-                    .or(cfg.service_tier)
-                    .filter(|tier| matches!(tier, ServiceTier::Fast))
-            })
-        } else {
-            None
+        let service_tier = service_tier_override
+            .unwrap_or_else(|| config_profile.service_tier.or(cfg.service_tier));
+        let service_tier = match service_tier {
+            Some(ServiceTier::Fast) if features.enabled(Feature::FastMode) => {
+                Some(ServiceTier::Fast)
+            }
+            Some(ServiceTier::Flex) => Some(ServiceTier::Flex),
+            _ => None,
         };
 
         let compact_prompt = compact_prompt.or(cfg.compact_prompt).and_then(|value| {

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

@@ -22,6 +22,7 @@ use codex_protocol::user_input::UserInput;
 use core_test_support::responses::ev_completed_with_tokens;
 use core_test_support::responses::ev_response_created;
 use core_test_support::responses::mount_models_once;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::sse_completed;
@@ -247,6 +248,25 @@ async fn service_tier_change_is_applied_on_next_http_turn() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn flex_service_tier_is_applied_to_http_turn() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let resp_mock = mount_sse_once(&server, sse_completed("resp-1")).await;
+
+    let test = test_codex().build(&server).await?;
+
+    test.submit_turn_with_service_tier("flex turn", Some(ServiceTier::Flex))
+        .await?;
+
+    let request = resp_mock.single_request();
+    let body = request.body_json();
+    assert_eq!(body["service_tier"].as_str(), Some("flex"));
+
+    Ok(())
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn model_change_from_image_to_text_strips_prior_image_content() -> Result<()> {
     skip_if_no_network!(Ok(()));

codex-rs/protocol/src/config_types.rs

@@ -118,6 +118,7 @@ pub enum WebSearchMode {
 #[strum(serialize_all = "lowercase")]
 pub enum ServiceTier {
     Fast,
+    Flex,
 }
 
 #[derive(Debug, Serialize, Deserialize, Clone, Copy, PartialEq, Eq, Display, JsonSchema, TS)]

codex-rs/tui/src/chatwidget.rs

@@ -3681,7 +3681,7 @@ impl ChatWidget {
                 self.open_model_popup();
             }
             SlashCommand::Fast => {
-                let next_tier = if self.config.service_tier.is_some() {
+                let next_tier = if matches!(self.config.service_tier, Some(ServiceTier::Fast)) {
                     None
                 } else {
                     Some(ServiceTier::Fast)
@@ -3976,7 +3976,8 @@ impl ChatWidget {
                     "on" => self.set_service_tier_selection(Some(ServiceTier::Fast)),
                     "off" => self.set_service_tier_selection(None),
                     "status" => {
-                        let status = if self.config.service_tier.is_some() {
+                        let status = if matches!(self.config.service_tier, Some(ServiceTier::Fast))
+                        {
                             "on"
                         } else {
                             "off"
@@ -4342,7 +4343,7 @@ impl ChatWidget {
             .personality
             .filter(|_| self.config.features.enabled(Feature::Personality))
             .filter(|_| self.current_model_supports_personality());
-        let service_tier = self.fast_mode_enabled().then_some(self.config.service_tier);
+        let service_tier = self.config.service_tier.map(Some);
         let op = Op::UserTurn {
             items,
             cwd: self.config.cwd.clone(),