fix(coding-agent/modes): fixed model switches deferred during streaming to apply correctly on completion

- Fixed model switches deferred during streaming to apply correctly when the stream completes, preventing model changes from being lost.
- Added pendingModelSwitch field to track deferred model changes during active streams.
- Added flushPendingModelSwitch() method to apply deferred model switches after streaming completes.
- Modified switchToPlanModel() to defer model changes when a stream is active instead of applying immediately.
- Modified exitPlanMode() to defer model restoration when a stream is active instead of applying immediately.
- Modified EventController to call flushPendingModelSwitch() when stream completion is detected.

packages/coding-agent/CHANGELOG.md

@@ -1,7 +1,6 @@
 # Changelog
 
 ## [Unreleased]
-
 ### Added
 
 - Support for image file mentions—images are now automatically detected, resized, and attached when referenced with @filepath syntax
@@ -14,6 +13,7 @@
 
 ### Fixed
 
+- Fixed model switches deferred during streaming to apply correctly when the stream completes, preventing model changes from being lost
 - Fixed plan mode toggles during streaming to inject plan-mode context immediately, preventing file edits while in plan mode
 - Fixed plan mode model switches during streaming to defer model changes until the current turn completes
 

packages/coding-agent/src/modes/controllers/event-controller.ts

@@ -269,6 +269,7 @@ export class EventController {
 					this.ctx.streamingComponent = undefined;
 					this.ctx.streamingMessage = undefined;
 				}
+				await this.ctx.flushPendingModelSwitch();
 				this.ctx.pendingTools.clear();
 				this.ctx.ui.requestRender();
 				this.sendCompletionNotification();

packages/coding-agent/src/modes/interactive-mode.ts

@@ -136,6 +136,7 @@ export class InteractiveMode implements InteractiveModeContext {
 	private readonly changelogMarkdown: string | undefined;
 	private planModePreviousTools: string[] | undefined;
 	private planModePreviousModel: Model | undefined;
+	private pendingModelSwitch: Model | undefined;
 	private planModeHasEntered = false;
 	public readonly lspServers:
 		| Array<{ name: string; status: "ready" | "error"; fileTypes: string[]; error?: string }>
@@ -533,6 +534,10 @@ export class InteractiveMode implements InteractiveModeContext {
 			return;
 		}
 		this.planModePreviousModel = currentModel;
+		if (this.session.isStreaming) {
+			this.pendingModelSwitch = planModel;
+			return;
+		}
 		try {
 			await this.session.setModelTemporary(planModel);
 		} catch (error) {
@@ -542,6 +547,20 @@ export class InteractiveMode implements InteractiveModeContext {
 		}
 	}
 
+	/** Apply any deferred model switch after the current stream ends. */
+	async flushPendingModelSwitch(): Promise<void> {
+		const model = this.pendingModelSwitch;
+		if (!model) return;
+		this.pendingModelSwitch = undefined;
+		try {
+			await this.session.setModelTemporary(model);
+		} catch (error) {
+			this.showWarning(
+				`Failed to switch model after streaming: ${error instanceof Error ? error.message : String(error)}`,
+			);
+		}
+	}
+
 	private async enterPlanMode(options?: {
 		planFilePath?: string;
 		workflow?: "parallel" | "iterative";
@@ -588,7 +607,11 @@ export class InteractiveMode implements InteractiveModeContext {
 			await this.session.setActiveToolsByName(previousTools);
 		}
 		if (this.planModePreviousModel) {
-			await this.session.setModelTemporary(this.planModePreviousModel);
+			if (this.session.isStreaming) {
+				this.pendingModelSwitch = this.planModePreviousModel;
+			} else {
+				await this.session.setModelTemporary(this.planModePreviousModel);
+			}
 		}
 
 		this.session.setPlanModeState(undefined);