fix: プロバイダークールダウンの別名判定を修正

Codex CLI の利用可能モデルを再計測し、最小トークン消費だった gpt-5.2 を既定モデルに更新

状態ファイルと設定上のプロバイダー名を比較時に正規化し、gemini/agy と antigravity、apple-ai/apple_intelligence と apple-intelligence を同一プロバイダーとして降格

旧エイリアス経由でもクールダウン降格されることを確認する state テストを追加

README と AGENTS の既定モデル、測定結果、クールダウン別名挙動を更新

Author: owayo

AGENTS.md

@@ -124,13 +124,14 @@ Provider state file note:
 - `State::save()` writes to `~/.config/git-sc/.providers-state.tmp` first and then `rename(2)`s it onto the final path so concurrent `git-sc` invocations never read a half-written TOML file. On rename failure the temporary file is deleted before the error is returned.
 - The temporary file suffix combines PID, monotonic nanosecond timestamp, and a process-local `AtomicU64` counter, so multiple threads (or rapid consecutive saves) that happen to observe the same wall-clock nanosecond never share a tmp path. Without the counter, two concurrent threads could write to the same `*.tmp.PID.NANOS` file and the slower thread's `rename(2)` would fail with `ENOENT` after the faster thread already moved it.
 - `provider_cooldown_minutes` is converted to seconds with saturating arithmetic, so extremely large user-provided values do not panic in debug builds or wrap in release builds; they are treated as effectively indefinite cooldowns.
+- Cooldown reordering canonicalizes provider aliases before comparing state keys with configured providers, so `gemini`/`agy` remain tied to `antigravity` and legacy `apple-ai`/`apple_intelligence` keys remain tied to `apple-intelligence`.
 
 ### Agent Context
 
 When invoked from a coding agent, `App::run()` reads the `CLAW_HOOKS_AGENT_MESSAGE` environment variable and passes it to `AiService::build_prompt()` as `agent_context`. This context is injected into the AI prompt before the diff section, guiding the AI to reflect the developer's high-level intent in the commit message. The context is applied across standard generation and `--amend` / `--reword` / `--squash` / `--generate-for` workflows.
 
 Default Codex model note:
-- As of May 30, 2026 (JST), the default Codex model is `gpt-5.3-codex-spark` after re-running `codex debug models` and `echo "Hello" | codex exec -c model_reasoning_effort='medium' -m <model>` for each listed Codex model. Single-run token consumption (cold start): `gpt-5.3-codex-spark` (`8,737`), `gpt-5.5` (`11,798`), `codex-auto-review` (`11,942`), `gpt-5.3-codex` (`12,021`), `gpt-5.4-mini` (`12,136`), `gpt-5.4` (`15,533`), and `gpt-5.2` (`16,828`). `gpt-5.3-codex-spark` was selected because it had the lowest measured consumption among the currently available Codex models.
+- As of May 31, 2026 (JST), the default Codex model is `gpt-5.2` after re-running `codex debug models` and `echo "Hello" | codex exec -c model_reasoning_effort='medium' -m <model>` for each listed Codex model. Single-run token consumption: `gpt-5.2` (`11,668`), `gpt-5.5` (`11,799`), `gpt-5.4` (`11,950`), `codex-auto-review` (`11,950`), `gpt-5.3-codex` (`12,023`), `gpt-5.3-codex-spark` (`12,083`), and `gpt-5.4-mini` (`15,214`). `gpt-5.2` was selected because it had the lowest measured consumption among the currently available Codex models.
 
 ## Configuration Files
 

README.ja.md

@@ -265,7 +265,7 @@ codex_reasoning_effort = "low"
 # 注: Antigravity CLI (`agy`) はモデル選択フラグを持たないため、`gemini` 行は不要
 # (旧 git-sc が生成した `gemini = "..."` 行は後方互換のためパースされるが実行時には無視されます)
 [models]
-codex = "gpt-5.3-codex-spark"
+codex = "gpt-5.2"
 claude = "haiku"
 opencode = ""
 
@@ -291,7 +291,9 @@ provider_timeout_seconds = 60
 | `prefix_rules` | URLベースのプレフィックス形式 | `[]` |
 | `prefix_scripts` | 外部プレフィックススクリプト | `[]` |
 
-既存のグローバル設定ファイルは自動では書き換えられません。現在の Codex 既定モデルは `gpt-5.3-codex-spark` です。既存設定で使うには、`~/.config/git-sc/config.toml` の `models.codex` を更新してください。この既定値は 2026年5月30日 (JST) 時点の Codex モデル一覧で再検証済みです。
+既存のグローバル設定ファイルは自動では書き換えられません。現在の Codex 既定モデルは `gpt-5.2` です。既存設定で使うには、`~/.config/git-sc/config.toml` の `models.codex` を更新してください。この既定値は 2026年5月31日 (JST) 時点の Codex モデル一覧で再検証済みです。
+
+プロバイダーのクールダウン状態は、並び替え前に旧エイリアスを正規化します。そのため `gemini`/`agy` のクールダウンは `antigravity` に、旧 `apple-ai` / `apple_intelligence` キーは `apple-intelligence` に引き続き適用されます。
 
 ### prefix_type の値
 

README.md

@@ -266,7 +266,7 @@ codex_reasoning_effort = "low"
 # Note: Antigravity CLI (`agy`) does not accept a model flag, so no `gemini` line is needed here.
 # A legacy `gemini = "..."` value is parsed for backward compatibility but ignored at runtime.
 [models]
-codex = "gpt-5.3-codex-spark"
+codex = "gpt-5.2"
 claude = "haiku"
 opencode = ""
 
@@ -292,7 +292,9 @@ provider_timeout_seconds = 60
 | `prefix_rules` | URL-based prefix format | `[]` |
 | `prefix_scripts` | External prefix scripts | `[]` |
 
-Existing global config files are not rewritten automatically. The current Codex default is `gpt-5.3-codex-spark`; to use it in an existing setup, update `models.codex` in `~/.config/git-sc/config.toml`. This default was revalidated against the available Codex model list on May 30, 2026 (JST).
+Existing global config files are not rewritten automatically. The current Codex default is `gpt-5.2`; to use it in an existing setup, update `models.codex` in `~/.config/git-sc/config.toml`. This default was revalidated against the available Codex model list on May 31, 2026 (JST).
+
+Provider cooldown state normalizes legacy aliases before reordering providers, so `gemini`/`agy` cooldown entries still apply to `antigravity`, and legacy `apple-ai` / `apple_intelligence` entries still apply to `apple-intelligence`.
 
 ### prefix_type Values
 

src/config.rs

@@ -30,7 +30,7 @@ fn default_gemini_model() -> String {
 }
 
 fn default_codex_model() -> String {
-    "gpt-5.3-codex-spark".to_string()
+    "gpt-5.2".to_string()
 }
 
 fn default_claude_model() -> String {
@@ -604,8 +604,8 @@ codex_reasoning_effort = "high"
     #[test]
     fn test_default_codex_model_uses_latest_lowest_token_model() {
         // codex debug models の単発計測結果に基づく現在の既定モデルを固定する。
-        assert_eq!(default_codex_model(), "gpt-5.3-codex-spark");
-        assert_eq!(Config::default().models.codex, "gpt-5.3-codex-spark");
+        assert_eq!(default_codex_model(), "gpt-5.2");
+        assert_eq!(Config::default().models.codex, "gpt-5.2");
     }
 
     #[test]
@@ -1917,7 +1917,7 @@ gemini = "gemini-2.5-pro"
 
         assert_eq!(config.models.gemini, default_gemini_model());
         assert_eq!(config.models.codex, default_codex_model());
-        assert!(content.contains(r#"codex = "gpt-5.3-codex-spark""#));
+        assert!(content.contains(r#"codex = "gpt-5.2""#));
         assert_eq!(config.models.claude, default_claude_model());
         assert_eq!(config.models.opencode, default_opencode_model());
     }

src/state.rs

@@ -131,6 +131,21 @@ impl State {
         cooldown_minutes.saturating_mul(60)
     }
 
+    /// 状態ファイルと設定ファイルのプロバイダー名を比較用の正規名にそろえる。
+    ///
+    /// 設定ファイルには後方互換エイリアスが残ることがあるため、保存済みの失敗キーと
+    /// 現在の provider 配列を同じプロバイダーとして扱えるようにする。
+    fn canonical_provider_key(provider: &str) -> String {
+        let lower = provider.to_lowercase();
+        match lower.as_str() {
+            "agy" | "gemini" | "antigravity" => "antigravity".to_string(),
+            "apple-ai" | "apple_intelligence" | "apple-intelligence" => {
+                "apple-intelligence".to_string()
+            }
+            _ => lower,
+        }
+    }
+
     /// プロバイダーの失敗を記録
     pub fn record_failure(&mut self, provider: &str) {
         self.provider_failures.insert(
@@ -171,16 +186,20 @@ impl State {
     /// 降格されたプロバイダーは末尾に移動
     pub fn reorder_providers(&self, providers: Vec<String>, cooldown_minutes: u64) -> Vec<String> {
         let demoted = self.get_demoted_providers(cooldown_minutes);
+        let demoted: Vec<String> = demoted
+            .iter()
+            .map(|p| Self::canonical_provider_key(p))
+            .collect();
 
         let mut normal: Vec<String> = providers
             .iter()
-            .filter(|p| !demoted.contains(&p.to_lowercase()))
+            .filter(|p| !demoted.contains(&Self::canonical_provider_key(p)))
             .cloned()
             .collect();
 
         let mut demoted_providers: Vec<String> = providers
             .iter()
-            .filter(|p| demoted.contains(&p.to_lowercase()))
+            .filter(|p| demoted.contains(&Self::canonical_provider_key(p)))
             .cloned()
             .collect();
 
@@ -284,6 +303,50 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_reorder_providers_demotes_antigravity_when_config_uses_gemini_alias() {
+        let mut state = State::default();
+        state.record_failure("antigravity");
+
+        let providers = vec![
+            "gemini".to_string(),
+            "codex".to_string(),
+            "claude".to_string(),
+        ];
+
+        let reordered = state.reorder_providers(providers, 60);
+        assert_eq!(
+            reordered,
+            vec![
+                "codex".to_string(),
+                "claude".to_string(),
+                "gemini".to_string(),
+            ]
+        );
+    }
+
+    #[test]
+    fn test_reorder_providers_demotes_apple_intelligence_from_legacy_state_key() {
+        let mut state = State::default();
+        state.record_failure("apple-ai");
+
+        let providers = vec![
+            "opencode".to_string(),
+            "apple-intelligence".to_string(),
+            "codex".to_string(),
+        ];
+
+        let reordered = state.reorder_providers(providers, 60);
+        assert_eq!(
+            reordered,
+            vec![
+                "opencode".to_string(),
+                "codex".to_string(),
+                "apple-intelligence".to_string(),
+            ]
+        );
+    }
+
     #[test]
     fn test_reorder_providers_multiple_demoted() {
         let mut state = State::default();