feat(studio): deploy manual designer and configurable AI model

app/lib/studio-ai.ts

@@ -3,7 +3,7 @@ import type {
   GenerationRequest,
 } from "@wrnexus/design-studio";
 
-export const APIMART_MODEL = "gpt-6-luna";
+export const APIMART_MODEL = "gpt-5-mini";
 export const AI_TIMEOUT_MS = 60_000;
 export const AI_OUTPUT_TOKENS = 8192;
 type FailureKind =
@@ -22,7 +22,7 @@ const failureMessages: Record<FailureKind, string> = {
   quota:
     "APIMart reports insufficient credit or quota. Check the provider account balance and limits.",
   model:
-    "APIMart could not access gpt-6-luna. Check that this model is enabled for the server API key.",
+    "APIMart could not access the configured model. Check that this model is enabled for the server API key.",
   rate: "APIMart is rate limiting requests. Wait a minute and generate again.",
   unavailable:
     "APIMart is temporarily unavailable. Try generating again shortly.",
@@ -70,6 +70,7 @@ function failureFor(status: number, code?: unknown) {
 export function apimartProvider(
   apiKey: string,
   transport: typeof fetch = fetch,
+  model = APIMART_MODEL,
 ): GenerationProvider {
   return {
     name: "APIMart",
@@ -87,7 +88,7 @@ export function apimartProvider(
               "content-type": "application/json",
             },
             body: JSON.stringify({
-              model: APIMART_MODEL,
+              model,
               stream: false,
               max_tokens: options.maxTokens ?? AI_OUTPUT_TOKENS,
               messages: [
@@ -137,7 +138,7 @@ export function apimartProvider(
       return {
         value: choice.message.content,
         provider: "APIMart",
-        model: APIMART_MODEL,
+        model,
       };
     },
   };
@@ -166,7 +167,11 @@ async function readBounded(source: Request | Response, limit: number) {
   }
 }
 
-function synchronousStudioAI(apiKey?: string, transport: typeof fetch = fetch) {
+function synchronousStudioAI(
+  apiKey?: string,
+  transport: typeof fetch = fetch,
+  model = APIMART_MODEL,
+) {
   let active = 0,
     windowStart = Date.now(),
     count = 0;
@@ -198,7 +203,7 @@ function synchronousStudioAI(apiKey?: string, transport: typeof fetch = fetch) {
     if (url.pathname.endsWith("/generation") && request.method === "GET")
       return json({
         provider: "APIMart",
-        model: APIMART_MODEL,
+        model,
         templateVersion: 1,
         delivery: "polling",
         maxAttempts: 2,
@@ -236,7 +241,7 @@ function synchronousStudioAI(apiKey?: string, transport: typeof fetch = fetch) {
       let calls = 0;
       let exhausted = false;
       let failure: APIMartFailure | undefined;
-      const upstream = apimartProvider(apiKey, transport);
+      const upstream = apimartProvider(apiKey, transport, model);
       const provider: GenerationProvider = {
         name: upstream.name,
         async generate(prompt, options) {
@@ -266,7 +271,7 @@ function synchronousStudioAI(apiKey?: string, transport: typeof fetch = fetch) {
       const generate = () =>
         generateDesigns(input, {
           provider,
-          model: APIMART_MODEL,
+          model,
           signal: deadline,
           timeoutMs: AI_TIMEOUT_MS,
           maxOutputTokens: AI_OUTPUT_TOKENS,
@@ -351,8 +356,12 @@ function synchronousStudioAI(apiKey?: string, transport: typeof fetch = fetch) {
 }
 
 /** Short HTTP requests keep the deployment proxy out of the provider wait. */
-export function studioAI(apiKey?: string, transport: typeof fetch = fetch) {
-  const run = synchronousStudioAI(apiKey, transport);
+export function studioAI(
+  apiKey?: string,
+  transport: typeof fetch = fetch,
+  model = APIMART_MODEL,
+) {
+  const run = synchronousStudioAI(apiKey, transport, model);
   type Job = {
     status: "pending" | "complete" | "failed" | "cancelled";
     controller: AbortController;