Add GPT 5 support (#902)

src/ipc/utils/get_model_client.ts

@@ -183,7 +183,7 @@ function getRegularModelClient(
       const provider = createOpenAI({ apiKey });
       return {
         modelClient: {
-          model: provider(model.name),
+          model: provider.responses(model.name),
           builtinProviderId: providerId,
         },
         backupModelClients: [],

src/ipc/utils/thinking_utils.ts

@@ -23,6 +23,11 @@ export function getExtraProviderOptions(
   if (!providerId) {
     return {};
   }
+  if (providerId === "openai") {
+    return {
+      reasoning_effort: "medium",
+    };
+  }
   if (PROVIDERS_THAT_SUPPORT_THINKING.includes(providerId)) {
     const budgetTokens = getThinkingBudgetTokens(settings?.thinkingBudget);
     return {

src/ipc/utils/token_utils.ts

@@ -24,10 +24,16 @@ export async function getContextWindow() {
   return modelOption?.contextWindow || DEFAULT_CONTEXT_WINDOW;
 }
 
-// Most models support at least 8000 output tokens so we use it as a default value.
-const DEFAULT_MAX_TOKENS = 8_000;
-
-export async function getMaxTokens(model: LargeLanguageModel) {
+export async function getMaxTokens(
+  model: LargeLanguageModel,
+): Promise<number | undefined> {
   const modelOption = await findLanguageModel(model);
-  return modelOption?.maxOutputTokens || DEFAULT_MAX_TOKENS;
+  return modelOption?.maxOutputTokens ?? undefined;
+}
+
+export async function getTemperature(
+  model: LargeLanguageModel,
+): Promise<number> {
+  const modelOption = await findLanguageModel(model);
+  return modelOption?.temperature ?? 0;
 }