fix(ai): honor codex transport option

closes #4083

packages/agent/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Changed
+
+- Changed the default agent transport to `auto` so providers can use their best available transport by default ([#4083](https://github.com/badlogic/pi-mono/issues/4083)).
+
 ## [0.72.0] - 2026-05-01
 
 ### Added

packages/agent/src/agent.ts

@@ -201,7 +201,7 @@ export class Agent {
 		this.followUpQueue = new PendingMessageQueue(options.followUpMode ?? "one-at-a-time");
 		this.sessionId = options.sessionId;
 		this.thinkingBudgets = options.thinkingBudgets;
-		this.transport = options.transport ?? "sse";
+		this.transport = options.transport ?? "auto";
 		this.maxRetryDelayMs = options.maxRetryDelayMs;
 		this.toolExecution = options.toolExecution ?? "parallel";
 	}

packages/ai/CHANGELOG.md

@@ -23,6 +23,7 @@
 
 ### Fixed
 
+- Fixed OpenAI Codex Responses `streamSimple()` to honor the configured transport instead of always using SSE, and made `auto` the default transport with cached WebSocket context when available ([#4083](https://github.com/badlogic/pi-mono/issues/4083)).
 - Fixed Xiaomi MiMo model catalog to use the Token Plan Anthropic endpoint instead of the direct API ([#3912](https://github.com/badlogic/pi-mono/issues/3912)).
 
 ## [0.71.1] - 2026-05-01

packages/ai/src/providers/openai-codex-responses.ts

@@ -165,7 +165,7 @@ export const streamOpenAICodexResponses: StreamFunction<"openai-codex-responses"
 				websocketRequestId,
 			);
 			const bodyJson = JSON.stringify(body);
-			const transport = options?.transport || "sse";
+			const transport = options?.transport || "auto";
 
 			if (transport !== "sse") {
 				let websocketStarted = false;
@@ -992,7 +992,7 @@ async function processWebSocketStream(
 ): Promise<void> {
 	const { socket, entry, reused, release } = await acquireWebSocket(url, headers, options?.sessionId, options?.signal);
 	let keepConnection = true;
-	const useCachedContext = options?.transport === "websocket-cached";
+	const useCachedContext = options?.transport === "websocket-cached" || options?.transport === "auto";
 	// ChatGPT Codex Responses rejects `store: true` ("Store must be set to false").
 	// WebSocket continuation still works via connection-scoped previous_response_id state.
 	const fullBody = body;

packages/ai/src/providers/simple-options.ts

@@ -6,6 +6,7 @@ export function buildBaseOptions(model: Model<Api>, options?: SimpleStreamOption
 		maxTokens: options?.maxTokens ?? (model.maxTokens > 0 ? Math.min(model.maxTokens, 32000) : undefined),
 		signal: options?.signal,
 		apiKey: apiKey || options?.apiKey,
+		transport: options?.transport,
 		cacheRetention: options?.cacheRetention,
 		sessionId: options?.sessionId,
 		headers: options?.headers,

packages/ai/test/openai-codex-stream.test.ts

@@ -451,6 +451,7 @@ describe("openai-codex streaming", () => {
 			provider: "openai-codex",
 			baseUrl: "https://chatgpt.com/backend-api",
 			reasoning: true,
+			thinkingLevelMap: { xhigh: "xhigh" },
 			input: ["text"],
 			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
 			contextWindow: 400000,
@@ -751,6 +752,113 @@ describe("openai-codex streaming", () => {
 		const streamResult = streamOpenAICodexResponses(model, context, { apiKey: token });
 		await streamResult.result();
 	});
+	it("forwards auto transport from streamSimple options and uses cached websocket context", async () => {
+		const token = mockToken();
+		const sentBodies: unknown[] = [];
+
+		global.fetch = vi.fn(async () => new Response("unexpected fetch", { status: 500 })) as typeof fetch;
+
+		class MockWebSocket {
+			private listeners = new Map<string, Set<(event: unknown) => void>>();
+
+			constructor(_url: string, _protocols?: string | string[] | { headers?: Record<string, string> }) {
+				queueMicrotask(() => this.dispatch("open", {}));
+			}
+
+			addEventListener(type: string, listener: (event: unknown) => void): void {
+				let listeners = this.listeners.get(type);
+				if (!listeners) {
+					listeners = new Set();
+					this.listeners.set(type, listeners);
+				}
+				listeners.add(listener);
+			}
+
+			removeEventListener(type: string, listener: (event: unknown) => void): void {
+				this.listeners.get(type)?.delete(listener);
+			}
+
+			send(data: string): void {
+				sentBodies.push(JSON.parse(data));
+				const events = [
+					{
+						type: "response.output_item.added",
+						item: { type: "message", id: "msg_1", role: "assistant", status: "in_progress", content: [] },
+					},
+					{ type: "response.content_part.added", part: { type: "output_text", text: "" } },
+					{ type: "response.output_text.delta", delta: "Hello" },
+					{
+						type: "response.output_item.done",
+						item: {
+							type: "message",
+							id: "msg_1",
+							role: "assistant",
+							status: "completed",
+							content: [{ type: "output_text", text: "Hello" }],
+						},
+					},
+					{
+						type: "response.completed",
+						response: {
+							status: "completed",
+							usage: {
+								input_tokens: 5,
+								output_tokens: 3,
+								total_tokens: 8,
+								input_tokens_details: { cached_tokens: 0 },
+							},
+						},
+					},
+				];
+				queueMicrotask(() => {
+					for (const event of events) {
+						this.dispatch("message", { data: JSON.stringify(event) });
+					}
+				});
+			}
+
+			close(): void {}
+
+			private dispatch(type: string, event: unknown): void {
+				for (const listener of this.listeners.get(type) ?? []) {
+					listener(event);
+				}
+			}
+		}
+
+		globalThis.WebSocket = MockWebSocket as unknown as typeof WebSocket;
+
+		const model: Model<"openai-codex-responses"> = {
+			id: "gpt-5.1-codex",
+			name: "GPT-5.1 Codex",
+			api: "openai-codex-responses",
+			provider: "openai-codex",
+			baseUrl: "https://chatgpt.com/backend-api",
+			reasoning: true,
+			input: ["text"],
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+			contextWindow: 400000,
+			maxTokens: 128000,
+		};
+		const context: Context = {
+			systemPrompt: "You are a helpful assistant.",
+			messages: [{ role: "user", content: "Say hello", timestamp: 1 }],
+		};
+
+		await streamSimpleOpenAICodexResponses(model, context, {
+			apiKey: token,
+			sessionId: "session-auto",
+			transport: "auto",
+		}).result();
+
+		expect(sentBodies).toHaveLength(1);
+		expect(global.fetch).not.toHaveBeenCalled();
+		expect(getOpenAICodexWebSocketDebugStats("session-auto")).toMatchObject({
+			cachedContextRequests: 1,
+			fullContextRequests: 1,
+		});
+	});
+
 	it("sends only response input deltas in websocket-cached mode", async () => {
 		const token = mockToken();
 		const sentBodies: unknown[] = [];

packages/coding-agent/CHANGELOG.md

@@ -24,6 +24,7 @@
 
 ### Fixed
 
+- Fixed the default transport setting to use `auto`, allowing OpenAI Codex to use cached WebSocket context when available ([#4083](https://github.com/badlogic/pi-mono/issues/4083)).
 - Fixed `pi.registerProvider()` to honor per-model `baseUrl` overrides ([#4063](https://github.com/badlogic/pi-mono/issues/4063)).
 - Fixed self-update detection so `pi` correctly identifies when a newer version is available and applies updates ([#3942](https://github.com/badlogic/pi-mono/issues/3942), [#3980](https://github.com/badlogic/pi-mono/issues/3980), [#3922](https://github.com/badlogic/pi-mono/issues/3922)).
 

packages/coding-agent/src/core/settings-manager.ts

@@ -78,7 +78,7 @@ export interface Settings {
 	defaultProvider?: string;
 	defaultModel?: string;
 	defaultThinkingLevel?: "off" | "minimal" | "low" | "medium" | "high" | "xhigh";
-	transport?: TransportSetting; // default: "sse"
+	transport?: TransportSetting; // default: "auto"
 	steeringMode?: "all" | "one-at-a-time";
 	followUpMode?: "all" | "one-at-a-time";
 	theme?: string;
@@ -656,7 +656,7 @@ export class SettingsManager {
 	}
 
 	getTransport(): TransportSetting {
-		return this.settings.transport ?? "sse";
+		return this.settings.transport ?? "auto";
 	}
 
 	setTransport(transport: TransportSetting): void {