feat: Add Codex OAuth FAST mode toggle (#2210)

* Add Codex OAuth FAST mode toggle

* fix(codex-oauth): default FAST mode to off to avoid surprise quota burn

service_tier="priority" consumes ChatGPT subscription quota at a higher
rate. Users must now opt in explicitly rather than inherit FAST mode
silently when this feature ships.

---------

Co-authored-by: Jason <farion1231@gmail.com>

src-tauri/src/provider.rs

@@ -262,6 +262,9 @@ pub struct ProviderMeta {
     /// conversions fall back to provider ID.
     #[serde(rename = "promptCacheKey", skip_serializing_if = "Option::is_none")]
     pub prompt_cache_key: Option<String>,
+    /// Codex OAuth FAST mode: inject `service_tier = "priority"` for ChatGPT Codex requests.
+    #[serde(rename = "codexFastMode", skip_serializing_if = "Option::is_none")]
+    pub codex_fast_mode: Option<bool>,
     /// 累加模式应用中，该 provider 是否已写入 live config。
     /// `None` 表示旧数据/未知状态，`Some(false)` 表示明确仅存在于数据库中。
     #[serde(rename = "liveConfigManaged", skip_serializing_if = "Option::is_none")]
@@ -277,6 +280,12 @@ pub struct ProviderMeta {
 }
 
 impl ProviderMeta {
+    /// Codex OAuth FAST mode 是否启用。默认关闭，因为 `service_tier="priority"`
+    /// 会按更高速率消耗 ChatGPT 订阅配额，用户需显式开启以换取更低延迟。
+    pub fn codex_fast_mode_enabled(&self) -> bool {
+        self.codex_fast_mode.unwrap_or(false)
+    }
+
     /// 解析指定托管认证供应商绑定的账号 ID。
     ///
     /// 新版优先读取 authBinding，旧版继续兼容 githubAccountId。

src-tauri/src/proxy/providers/claude.rs

@@ -151,10 +151,16 @@ pub fn transform_claude_request_for_api_format(
         "openai_responses" => {
             // Codex OAuth (ChatGPT Plus/Pro 反代) 需要在请求体里强制 store: false
             // + include: ["reasoning.encrypted_content"]，由 transform 层统一处理。
+            let codex_fast_mode = provider
+                .meta
+                .as_ref()
+                .map(|m| m.codex_fast_mode_enabled())
+                .unwrap_or(false);
             super::transform_responses::anthropic_to_responses(
                 body,
                 Some(cache_key),
                 is_codex_oauth,
+                codex_fast_mode,
             )
         }
         "openai_chat" => {
@@ -1330,6 +1336,43 @@ mod tests {
         assert_eq!(transformed["prompt_cache_key"], "explicit-cache-key");
     }
 
+    #[test]
+    fn test_transform_claude_request_for_api_format_codex_oauth_fast_mode_off() {
+        let provider = create_provider_with_meta(
+            json!({
+                "env": {
+                    "ANTHROPIC_BASE_URL": "https://chatgpt.com/backend-api/codex"
+                }
+            }),
+            ProviderMeta {
+                provider_type: Some("codex_oauth".to_string()),
+                codex_fast_mode: Some(false),
+                ..ProviderMeta::default()
+            },
+        );
+        let body = json!({
+            "model": "gpt-5.4",
+            "messages": [{ "role": "user", "content": "hello" }],
+            "max_tokens": 128
+        });
+
+        let transformed = transform_claude_request_for_api_format(
+            body,
+            &provider,
+            "openai_responses",
+            None,
+            None,
+        )
+        .unwrap();
+
+        assert_eq!(transformed["store"], json!(false));
+        assert!(transformed.get("service_tier").is_none());
+        assert_eq!(
+            transformed["include"],
+            json!(["reasoning.encrypted_content"])
+        );
+    }
+
     #[test]
     fn test_transform_claude_request_for_api_format_gemini_native() {
         let provider = create_provider_with_meta(

src-tauri/src/proxy/providers/transform_responses.rs

@@ -17,10 +17,13 @@ use serde_json::{json, Value};
 /// `is_codex_oauth`: 当目标后端是 ChatGPT Plus/Pro 反代 (`chatgpt.com/backend-api/codex`) 时为 true。
 /// 该后端强制要求 `store: false`，并要求 `include` 包含 `reasoning.encrypted_content`
 /// 以便在无服务端状态下保持多轮 reasoning 上下文。
+/// `codex_fast_mode`: 仅在 `is_codex_oauth` 为 true 时生效，控制是否注入
+/// `service_tier = "priority"`。
 pub fn anthropic_to_responses(
     body: Value,
     cache_key: Option<&str>,
     is_codex_oauth: bool,
+    codex_fast_mode: bool,
 ) -> Result<Value, ProxyError> {
     let mut result = json!({});
 
@@ -125,10 +128,15 @@ pub fn anthropic_to_responses(
     //   （codex-rs 结构体根本没有这三个字段，OpenAI 自己的客户端不发它们）
     // - instructions / tools / parallel_tool_calls: 必填字段，缺则兜底默认值
     //   （cc-switch 的 transform 当前是"条件写入"，可能产生缺失）
+    // - service_tier: 仅在 FAST mode 开启时写入 "priority"
+    //   （与 OpenAI 官方 codex-rs 当前请求结构保持一致）
     // - stream: 必须永远 true（codex-rs 硬编码 true，且 cc-switch 的
     //   SSE 解析层只处理流式响应，强制覆盖避免客户端误传 false）
     if is_codex_oauth {
         result["store"] = json!(false);
+        if codex_fast_mode {
+            result["service_tier"] = json!("priority");
+        }
 
         const REASONING_MARKER: &str = "reasoning.encrypted_content";
         let mut includes: Vec<Value> = body
@@ -520,7 +528,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["model"], "gpt-4o");
         assert_eq!(result["max_output_tokens"], 1024);
         assert_eq!(result["input"][0]["role"], "user");
@@ -539,7 +547,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["instructions"], "You are a helpful assistant.");
         // system should not appear in input
         assert_eq!(result["input"].as_array().unwrap().len(), 1);
@@ -557,7 +565,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["instructions"], "Part 1\n\nPart 2");
     }
 
@@ -574,7 +582,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["tools"][0]["type"], "function");
         assert_eq!(result["tools"][0]["name"], "get_weather");
         assert!(result["tools"][0].get("parameters").is_some());
@@ -591,7 +599,7 @@ mod tests {
             "tool_choice": {"type": "any"}
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["tool_choice"], "required");
     }
 
@@ -604,7 +612,7 @@ mod tests {
             "tool_choice": {"type": "tool", "name": "get_weather"}
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["tool_choice"]["type"], "function");
         assert_eq!(result["tool_choice"]["name"], "get_weather");
     }
@@ -623,7 +631,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         let input_arr = result["input"].as_array().unwrap();
 
         // Should produce: assistant message (text) + function_call item
@@ -653,7 +661,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         let input_arr = result["input"].as_array().unwrap();
 
         // Should produce: function_call_output item (lifted)
@@ -677,7 +685,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         let input_arr = result["input"].as_array().unwrap();
 
         // thinking should be discarded, only text remains
@@ -700,7 +708,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         let content = result["input"][0]["content"].as_array().unwrap();
 
         assert_eq!(content[0]["type"], "input_text");
@@ -858,7 +866,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["model"], "o3-mini");
     }
 
@@ -870,7 +878,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, Some("my-provider-id"), false).unwrap();
+        let result = anthropic_to_responses(input, Some("my-provider-id"), false, true).unwrap();
         assert_eq!(result["prompt_cache_key"], "my-provider-id");
     }
 
@@ -888,7 +896,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert!(result["tools"][0].get("cache_control").is_none());
     }
 
@@ -905,7 +913,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert!(result["input"][0]["content"][0]
             .get("cache_control")
             .is_none());
@@ -967,7 +975,7 @@ mod tests {
             "max_tokens": 4096,
             "messages": [{"role": "user", "content": "Hello"}]
         });
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["max_output_tokens"], 4096);
         assert!(result.get("max_completion_tokens").is_none());
     }
@@ -981,7 +989,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "xhigh");
     }
 
@@ -995,7 +1003,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "low");
     }
 
@@ -1008,7 +1016,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "low");
     }
 
@@ -1021,7 +1029,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "medium");
     }
 
@@ -1034,7 +1042,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "high");
     }
 
@@ -1047,7 +1055,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert_eq!(result["reasoning"]["effort"], "xhigh");
     }
 
@@ -1060,7 +1068,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
         assert!(result.get("reasoning").is_none());
     }
 
@@ -1074,10 +1082,11 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         // store 必须显式为 false（ChatGPT 后端拒绝 true）
         assert_eq!(result["store"], json!(false));
+        assert_eq!(result["service_tier"], json!("priority"));
 
         // include 必须包含 reasoning.encrypted_content（无服务端状态下保持多轮 reasoning）
         assert_eq!(result["include"], json!(["reasoning.encrypted_content"]));
@@ -1093,9 +1102,10 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
 
         assert!(result.get("store").is_none());
+        assert!(result.get("service_tier").is_none());
         assert!(result.get("include").is_none());
     }
 
@@ -1109,7 +1119,7 @@ mod tests {
             "include": ["something.else", "reasoning.encrypted_content"]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
         let includes = result["include"]
             .as_array()
             .expect("include should be array");
@@ -1130,6 +1140,21 @@ mod tests {
         assert_eq!(marker_count, 1, "marker 不应被重复添加（idempotent 失败）");
     }
 
+    #[test]
+    fn test_anthropic_to_responses_codex_oauth_fast_mode_can_be_disabled() {
+        let input = json!({
+            "model": "gpt-5-codex",
+            "max_tokens": 1024,
+            "messages": [{"role": "user", "content": "Hello"}]
+        });
+
+        let result = anthropic_to_responses(input, None, true, false).unwrap();
+
+        assert_eq!(result["store"], json!(false));
+        assert!(result.get("service_tier").is_none());
+        assert_eq!(result["include"], json!(["reasoning.encrypted_content"]));
+    }
+
     #[test]
     fn test_anthropic_to_responses_codex_oauth_strips_max_output_tokens() {
         // ChatGPT Plus/Pro 反代不接受 max_output_tokens（OpenAI 官方 codex-rs 的
@@ -1141,7 +1166,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert!(
             result.get("max_output_tokens").is_none(),
@@ -1159,7 +1184,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
 
         assert_eq!(result["max_output_tokens"], json!(1024));
     }
@@ -1177,7 +1202,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert!(
             result.get("temperature").is_none(),
@@ -1195,7 +1220,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert!(
             result.get("top_p").is_none(),
@@ -1212,7 +1237,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert_eq!(
             result["instructions"],
@@ -1246,7 +1271,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert_eq!(
             result["instructions"],
@@ -1270,7 +1295,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, true).unwrap();
+        let result = anthropic_to_responses(input, None, true, true).unwrap();
 
         assert_eq!(
             result["stream"],
@@ -1291,7 +1316,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
 
         assert_eq!(result["temperature"], json!(0.7));
         assert_eq!(result["top_p"], json!(0.9));
@@ -1307,7 +1332,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_responses(input, None, false).unwrap();
+        let result = anthropic_to_responses(input, None, false, true).unwrap();
 
         assert!(
             result.get("parallel_tool_calls").is_none(),

src-tauri/src/services/stream_check.rs

@@ -360,10 +360,20 @@ impl StreamCheckService {
             .as_ref()
             .and_then(|m| m.provider_type.as_deref())
             == Some("codex_oauth");
+        let codex_fast_mode = provider
+            .meta
+            .as_ref()
+            .map(|m| m.codex_fast_mode_enabled())
+            .unwrap_or(false);
 
         let body = if is_openai_responses {
-            anthropic_to_responses(anthropic_body, Some(&provider.id), is_codex_oauth)
-                .map_err(|e| AppError::Message(format!("Failed to build test request: {e}")))?
+            anthropic_to_responses(
+                anthropic_body,
+                Some(&provider.id),
+                is_codex_oauth,
+                codex_fast_mode,
+            )
+            .map_err(|e| AppError::Message(format!("Failed to build test request: {e}")))?
         } else if is_gemini_native {
             anthropic_to_gemini(anthropic_body)
                 .map_err(|e| AppError::Message(format!("Failed to build test request: {e}")))?