Skip to content

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

fix(codex): apply global service tiers to combo request bodies

Jan Leonaf96e33184dd
diff --git a/src/lib/providers/codexFastTier.ts b/src/lib/providers/codexFastTier.ts
index de796f2087..9fe973e467 100644
--- a/src/lib/providers/codexFastTier.ts
+++ b/src/lib/providers/codexFastTier.ts
@@ -130,8 +130,8 @@ export interface ApplyCodexGlobalFastServiceTierOptions {
*/
model?: string | null;
/**
- * Outbound request body. Per-request body.service_tier is left untouched if
- * already set.
+ * Outbound request body. A valid per-request body.service_tier is left untouched
+ * when already set.
*/
body?: Record<string, unknown> | null;
}
@@ -188,12 +188,11 @@ export function applyCodexGlobalFastServiceTier<T extends JsonRecord | null | un
} as T;
}
- if (resolved.tier === "flex") {
- // Write the wire value directly to the outbound body when possible. The executor
- // also accepts requestDefaults.serviceTier = "flex" for downstream accounting.
- if (body && typeof body === "object" && !Array.isArray(body)) {
- (body as JsonRecord).service_tier = "flex";
- }
+ if (body && typeof body === "object" && !Array.isArray(body)) {
+ // Write the wire value directly to the outbound body when possible, so combo
+ // request previews and logs show the same effective tier that the executor sends.
+ // The executor also accepts requestDefaults.serviceTier for downstream accounting.
+ (body as JsonRecord).service_tier = resolved.tier;
}
// Intentional precedence: body service_tier > global mode > connection defaults.
diff --git a/tests/integration/chat-pipeline.test.ts b/tests/integration/chat-pipeline.test.ts
index 7476240b2c..2773023a11 100644
--- a/tests/integration/chat-pipeline.test.ts
+++ b/tests/integration/chat-pipeline.test.ts
@@ -606,6 +606,47 @@ test("chat pipeline persists Codex responses cache and reasoning tokens to call
assert.equal(callLog.tokens.reasoning, 13);
});
+test("chat pipeline applies global Codex priority service tier inside combos", async () => {
+ await seedConnection("codex", { apiKey: "sk-codex-combo-priority" });
+ await settingsDb.updateSettings({
+ codexServiceTier: { enabled: true, tier: "priority" },
+ });
+ await combosDb.createCombo({
+ name: "codex-priority-combo",
+ strategy: "priority",
+ config: { maxRetries: 0, retryDelayMs: 0 },
+ models: ["codex/gpt-5.5"],
+ });
+ const fetchCalls = [];
+
+ globalThis.fetch = async (url, init: RequestInit = {}) => {
+ fetchCalls.push({
+ url: String(url),
+ headers: toPlainHeaders(init.headers),
+ body: init.body ? JSON.parse(String(init.body)) : null,
+ });
+ return buildOpenAIResponsesSSE({ text: "combo priority ok", model: "gpt-5.5" });
+ };
+
+ const response = await handleChat(
+ buildRequest({
+ body: {
+ model: "codex-priority-combo",
+ stream: false,
+ messages: [{ role: "user", content: "Use Codex combo priority" }],
+ },
+ })
+ );
+
+ const json = (await response.json()) as any;
+ assert.equal(response.status, 200);
+ assert.equal(fetchCalls.length, 1);
+ assert.match(fetchCalls[0].url, /\/responses$/);
+ assert.equal(fetchCalls[0].headers.Authorization, "Bearer sk-codex-combo-priority");
+ assert.equal(fetchCalls[0].body.service_tier, "priority");
+ assert.equal(json.choices[0].message.content, "combo priority ok");
+});
+
test("chat pipeline applies Codex CLI fingerprint to OAuth responses requests", async () => {
setCliCompatProviders(["codex"]);
await seedConnection("codex", {
@@ -950,12 +991,7 @@ test("chat pipeline sends Gemini CLI OAuth requests with native Cloud Code trans
assert.equal(generateCall.body.requestId, undefined);
assert.equal(generateCall.body.user_prompt_id, generateCall.body.request.session_id);
const keys = Object.keys(generateCall.body).slice(0, 4);
- assert.deepEqual(keys.sort(), [
- "model",
- "project",
- "request",
- "user_prompt_id",
- ]);
+ assert.deepEqual(keys.sort(), ["model", "project", "request", "user_prompt_id"]);
assert.equal(generateCall.body.request.sessionId, undefined);
assert.match(generateCall.body.request.session_id, /^[0-9a-f-]{36}$/i);
assert.equal(generateCall.body.request.contents.at(-1).parts[0].text, "Hello Gemini CLI");
diff --git a/tests/unit/codex-fast-tier.test.ts b/tests/unit/codex-fast-tier.test.ts
index 4aea837ccb..f411574345 100644
--- a/tests/unit/codex-fast-tier.test.ts
+++ b/tests/unit/codex-fast-tier.test.ts
@@ -106,15 +106,17 @@ test("Codex global service tier injects selected mode and can override connectio
});
test("Codex global service tier matches provider-prefixed combo model ids", () => {
+ const body: Record<string, unknown> = {};
assert.deepEqual(
applyCodexGlobalFastServiceTier(
"codex",
{ providerSpecificData: {} },
{ codexServiceTier: { enabled: true, tier: "priority" } },
- { model: "codex/gpt-5.5" }
+ { model: "codex/gpt-5.5", body }
),
{ providerSpecificData: { requestDefaults: { serviceTier: "priority" } } }
);
+ assert.equal(body.service_tier, "priority");
const unsupported = { providerSpecificData: {} };
assert.equal(
@@ -155,7 +157,7 @@ test("Codex global service tier only short-circuits on valid body service_tier",
assert.deepEqual(injected, {
providerSpecificData: { requestDefaults: { serviceTier: "priority" } },
});
- assert.equal(invalidBody.service_tier, "invalid");
+ assert.equal(invalidBody.service_tier, "priority");
const validBody: Record<string, unknown> = { service_tier: " Flex " };
const unchanged = { providerSpecificData: {} };