🤖 feat: default OpenAI service_tier to priority

src/common/orpc/schemas/providerOptions.ts

@@ -10,6 +10,10 @@ export const MuxProviderOptionsSchema = z.object({
     .optional(),
   openai: z
     .object({
+      serviceTier: z.enum(["auto", "default", "flex", "priority"]).optional().meta({
+        description:
+          "OpenAI service tier: priority (low-latency), flex (50% cheaper, higher latency), auto/default (standard)",
+      }),
       disableAutoTruncation: z
         .boolean()
         .optional()

src/common/utils/ai/providerOptions.ts

@@ -217,11 +217,12 @@ export function buildProviderOptions(
       disableAutoTruncation,
     });
 
+    const serviceTier = muxProviderOptions?.openai?.serviceTier ?? "priority";
+
     const options: ProviderOptions = {
       openai: {
         parallelToolCalls: true, // Always enable concurrent tool execution
-        // TODO: allow this to be configured
-        serviceTier: "auto", // Use "auto" to automatically select the best service tier
+        serviceTier,
         // Automatically truncate conversation to fit context window, unless disabled for testing
         truncation: disableAutoTruncation ? "disabled" : "auto",
         // Conditionally add reasoning configuration

src/node/services/aiService.ts

@@ -498,6 +498,16 @@ export class AIService extends EventEmitter {
             provider: providerName,
           });
         }
+
+        // Extract serviceTier from config to pass through to buildProviderOptions
+        const configServiceTier = providerConfig.serviceTier as string | undefined;
+        if (configServiceTier && muxProviderOptions) {
+          muxProviderOptions.openai = {
+            ...muxProviderOptions.openai,
+            serviceTier: configServiceTier as "auto" | "default" | "flex" | "priority",
+          };
+        }
+
         const baseFetch = getProviderFetch(providerConfig);
 
         // Wrap fetch to force truncation: "auto" for OpenAI Responses API calls.