Update realtime websocket API (#13265)

- migrate the realtime websocket transport to the new session and
handoff flow
- make the realtime model configurable in config.toml and use API-key
auth for the websocket

---------

Co-authored-by: Codex <noreply@openai.com>

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

@@ -435,12 +435,16 @@ pub struct Config {
     pub realtime_audio: RealtimeAudioConfig,
 
     /// Experimental / do not use. Overrides only the realtime conversation
-    /// websocket transport base URL (the `Op::RealtimeConversation` `/ws`
+    /// websocket transport base URL (the `Op::RealtimeConversation`
+    /// `/v1/realtime`
     /// connection) without changing normal provider HTTP requests.
     pub experimental_realtime_ws_base_url: Option<String>,
+    /// Experimental / do not use. Selects the realtime websocket model/snapshot
+    /// used for the `Op::RealtimeConversation` connection.
+    pub experimental_realtime_ws_model: Option<String>,
     /// Experimental / do not use. Overrides only the realtime conversation
-    /// websocket transport backend prompt (the `Op::RealtimeConversation`
-    /// `/ws` session.create backend_prompt) without changing normal prompts.
+    /// websocket transport instructions (the `Op::RealtimeConversation`
+    /// `/ws` session.update instructions) without changing normal prompts.
     pub experimental_realtime_ws_backend_prompt: Option<String>,
     /// When set, restricts ChatGPT login to a specific workspace identifier.
     pub forced_chatgpt_workspace_id: Option<String>,
@@ -1188,12 +1192,16 @@ pub struct ConfigToml {
     pub audio: Option<RealtimeAudioToml>,
 
     /// Experimental / do not use. Overrides only the realtime conversation
-    /// websocket transport base URL (the `Op::RealtimeConversation` `/ws`
+    /// websocket transport base URL (the `Op::RealtimeConversation`
+    /// `/v1/realtime`
     /// connection) without changing normal provider HTTP requests.
     pub experimental_realtime_ws_base_url: Option<String>,
+    /// Experimental / do not use. Selects the realtime websocket model/snapshot
+    /// used for the `Op::RealtimeConversation` connection.
+    pub experimental_realtime_ws_model: Option<String>,
     /// Experimental / do not use. Overrides only the realtime conversation
-    /// websocket transport backend prompt (the `Op::RealtimeConversation`
-    /// `/ws` session.create backend_prompt) without changing normal prompts.
+    /// websocket transport instructions (the `Op::RealtimeConversation`
+    /// `/ws` session.update instructions) without changing normal prompts.
     pub experimental_realtime_ws_backend_prompt: Option<String>,
     pub projects: Option<HashMap<String, ProjectConfig>>,
 
@@ -2182,6 +2190,7 @@ impl Config {
                     speaker: audio.speaker,
                 }),
             experimental_realtime_ws_base_url: cfg.experimental_realtime_ws_base_url,
+            experimental_realtime_ws_model: cfg.experimental_realtime_ws_model,
             experimental_realtime_ws_backend_prompt: cfg.experimental_realtime_ws_backend_prompt,
             forced_chatgpt_workspace_id,
             forced_login_method,
@@ -4924,6 +4933,7 @@ model_verbosity = "high"
                 chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
                 realtime_audio: RealtimeAudioConfig::default(),
                 experimental_realtime_ws_base_url: None,
+                experimental_realtime_ws_model: None,
                 experimental_realtime_ws_backend_prompt: None,
                 base_instructions: None,
                 developer_instructions: None,
@@ -5052,6 +5062,7 @@ model_verbosity = "high"
             chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
             realtime_audio: RealtimeAudioConfig::default(),
             experimental_realtime_ws_base_url: None,
+            experimental_realtime_ws_model: None,
             experimental_realtime_ws_backend_prompt: None,
             base_instructions: None,
             developer_instructions: None,
@@ -5178,6 +5189,7 @@ model_verbosity = "high"
             chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
             realtime_audio: RealtimeAudioConfig::default(),
             experimental_realtime_ws_base_url: None,
+            experimental_realtime_ws_model: None,
             experimental_realtime_ws_backend_prompt: None,
             base_instructions: None,
             developer_instructions: None,
@@ -5290,6 +5302,7 @@ model_verbosity = "high"
             chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
             realtime_audio: RealtimeAudioConfig::default(),
             experimental_realtime_ws_base_url: None,
+            experimental_realtime_ws_model: None,
             experimental_realtime_ws_backend_prompt: None,
             base_instructions: None,
             developer_instructions: None,
@@ -6135,6 +6148,34 @@ experimental_realtime_ws_backend_prompt = "prompt from config"
         Ok(())
     }
 
+    #[test]
+    fn experimental_realtime_ws_model_loads_from_config_toml() -> std::io::Result<()> {
+        let cfg: ConfigToml = toml::from_str(
+            r#"
+experimental_realtime_ws_model = "realtime-test-model"
+"#,
+        )
+        .expect("TOML deserialization should succeed");
+
+        assert_eq!(
+            cfg.experimental_realtime_ws_model.as_deref(),
+            Some("realtime-test-model")
+        );
+
+        let codex_home = TempDir::new()?;
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert_eq!(
+            config.experimental_realtime_ws_model.as_deref(),
+            Some("realtime-test-model")
+        );
+        Ok(())
+    }
+
     #[test]
     fn realtime_audio_loads_from_config_toml() -> std::io::Result<()> {
         let cfg: ConfigToml = toml::from_str(