From eb5bfbfb11c598826b0d2da3255347a244004e0e Mon Sep 17 00:00:00 2001
From: roboomp <omp@can.ac>
Date: Sat, 1 Aug 2026 03:45:49 +0000
Subject: [PATCH] fix(ai): honored disabled Codex cache retention

Routed Codex prompt cache identity through the shared retention-aware resolver while preserving transport session identity.

Covered explicit and environment-derived opt-outs, option precedence, direct body construction, and transport headers.

Fixes #7219
---
 packages/ai/CHANGELOG.md                      |  4 ++
 .../src/providers/openai-codex-responses.ts   |  5 ++-
 packages/ai/test/openai-codex-stream.test.ts  | 43 +++++++++++++++++++
 3 files changed, 50 insertions(+), 2 deletions(-)

diff --git a/packages/ai/CHANGELOG.md b/packages/ai/CHANGELOG.md
index 0dc86d6e0..a43b95666 100644
--- a/packages/ai/CHANGELOG.md
+++ b/packages/ai/CHANGELOG.md
@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fixed OpenAI Codex Responses ignoring disabled cache retention when deriving `prompt_cache_key`, while preserving transport session identity ([#7219](https://github.com/can1357/oh-my-pi/issues/7219)).
+
 ## [17.2.2] - 2026-07-31
 
 ### Added
diff --git a/packages/ai/src/providers/openai-codex-responses.ts b/packages/ai/src/providers/openai-codex-responses.ts
index 20751703c..a58409222 100644
--- a/packages/ai/src/providers/openai-codex-responses.ts
+++ b/packages/ai/src/providers/openai-codex-responses.ts
@@ -111,6 +111,7 @@ import {
 	finalizePendingResponsesToolCalls,
 	finalizeReasoningThinking,
 	finalizeToolCallArgumentsDone,
+	getOpenAIPromptCacheKey,
 	hasExecutableIncompleteResponsesToolCalls,
 	isOpenAIResponsesProgressEvent,
 	mapOpenAIResponsesStopReason,
@@ -1373,7 +1374,7 @@ async function buildCodexRequestContext(
 	const accountId = getCodexAccountId(apiKey);
 	const baseUrl = model.baseUrl || CODEX_BASE_URL;
 	const url = resolveCodexResponsesUrl(baseUrl);
-	const promptCacheKey = normalizeOpenAIPromptCacheKey(options?.promptCacheKey ?? options?.sessionId);
+	const promptCacheKey = getOpenAIPromptCacheKey(options);
 	const transportSessionId = normalizeOpenAIPromptCacheKey(options?.sessionId);
 	const codexClientVersion = CODEX_CLIENT_VERSION;
 	const transformedBody = await buildTransformedCodexRequestBody(model, context, options, promptCacheKey);
@@ -1463,7 +1464,7 @@ export async function buildTransformedCodexRequestBody(
 	model: Model<"openai-codex-responses">,
 	context: Context,
 	options: OpenAICodexResponsesOptions | undefined,
-	promptCacheKey = normalizeOpenAIPromptCacheKey(options?.promptCacheKey ?? options?.sessionId),
+	promptCacheKey = getOpenAIPromptCacheKey(options),
 ): Promise<RequestBody> {
 	const params: RequestBody = {
 		model: model.requestModelId ?? model.id,
diff --git a/packages/ai/test/openai-codex-stream.test.ts b/packages/ai/test/openai-codex-stream.test.ts
index 11fb57a38..b0a372ef1 100644
--- a/packages/ai/test/openai-codex-stream.test.ts
+++ b/packages/ai/test/openai-codex-stream.test.ts
@@ -2,6 +2,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import { scheduler } from "node:timers/promises";
 import { streamSimple } from "@oh-my-pi/pi-ai";
 import {
+	buildTransformedCodexRequestBody,
 	getOpenAICodexTransportDetails,
 	getOpenAICodexWebSocketDebugStats,
 	prewarmOpenAICodexResponses,
@@ -19,6 +20,7 @@ import type {
 import { __resetProxyCache } from "@oh-my-pi/pi-ai/utils/proxy";
 import { buildModel } from "@oh-my-pi/pi-catalog/build";
 import * as piUtils from "@oh-my-pi/pi-utils";
+import { withEnv } from "./helpers";
 
 const { getAgentDir, setAgentDir, TempDir } = piUtils;
 
@@ -2369,6 +2371,47 @@ describe("openai-codex streaming", () => {
 		expect(capturedHeaders?.get("session_id")).toBe(sessionId);
 		expect(capturedHeaders?.get("x-client-request-id")).toBe(sessionId);
 		expect(capturedBody?.prompt_cache_key).toBe(promptCacheKey);
+
+		await streamOpenAICodexResponses(model, createCodexTestContext(), {
+			fetch: fetchMock as FetchImpl,
+			apiKey: token,
+			sessionId,
+			promptCacheKey,
+			cacheRetention: "none",
+		}).result();
+
+		expect(capturedHeaders?.get("conversation_id")).toBe(sessionId);
+		expect(capturedHeaders?.get("session_id")).toBe(sessionId);
+		expect(capturedHeaders?.get("x-client-request-id")).toBe(sessionId);
+		expect(capturedBody?.prompt_cache_key).toBeUndefined();
+	});
+	it("applies cache retention resolution to direct Codex request body construction", async () => {
+		const model = createCodexTestModel();
+		const context = createCodexTestContext();
+		const disabledPromptKey = await buildTransformedCodexRequestBody(model, context, {
+			promptCacheKey: "disabled-cache",
+			cacheRetention: "none",
+		});
+		const disabledSession = await buildTransformedCodexRequestBody(model, context, {
+			sessionId: "disabled-session",
+			cacheRetention: "none",
+		});
+
+		expect(disabledPromptKey.prompt_cache_key).toBeUndefined();
+		expect(disabledSession.prompt_cache_key).toBeUndefined();
+
+		await withEnv({ PI_CACHE_RETENTION: "none" }, async () => {
+			const disabledEnvironment = await buildTransformedCodexRequestBody(model, context, {
+				sessionId: "environment-session",
+			});
+			const explicitShort = await buildTransformedCodexRequestBody(model, context, {
+				promptCacheKey: "explicit-cache",
+				cacheRetention: "short",
+			});
+
+			expect(disabledEnvironment.prompt_cache_key).toBeUndefined();
+			expect(explicitShort.prompt_cache_key).toBe("explicit-cache");
+		});
 	});
 
 	it("omits unsupported sampling keys (temperature/top_p/top_k/min_p/penalties) from the Codex Responses body", async () => {