feat(inference-providers): graduate alibaba to user BYOK

apps/web/src/components/organizations/byok/BYOKKeysManager.tsx

@@ -51,6 +51,7 @@ import * as z from 'zod';
 // clause forces new entries here whenever a provider is added to
 // VercelUserByokInferenceProviderIdSchema.
 const VERCEL_BYOK_PROVIDER_NAMES = {
+  alibaba: 'Alibaba',
   anthropic: 'Anthropic',
   bedrock: 'AWS Bedrock',
   openai: 'OpenAI',

apps/web/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts

@@ -21,6 +21,8 @@ import type { BYOKResult, Provider } from '@/lib/ai-gateway/providers/types';
 import { isStepModel } from '@/lib/ai-gateway/providers/stepfun';
 import { isDeepseekModel } from '@/lib/ai-gateway/providers/deepseek';
 import type { FraudDetectionHeaders } from '@/lib/utils';
+import { isAlibabaDirectModel } from '@/lib/ai-gateway/providers/qwen';
+import { addCacheBreakpoints } from '@/lib/ai-gateway/providers/openrouter/request-helpers';
 
 function applyToolChoiceSetting(
   requestedModel: string,
@@ -123,6 +125,10 @@ export function applyProviderSpecificLogic(
     }
   }
 
+  if (isAlibabaDirectModel(requestedModel)) {
+    addCacheBreakpoints(requestToMutate);
+  }
+
   if (isClaudeModel(requestedModel)) {
     applyAnthropicModelSettings(requestToMutate, extraHeaders);
   }

apps/web/src/lib/ai-gateway/providers/openrouter/inference-provider-id.ts

@@ -74,6 +74,7 @@ export const OpenRouterInferenceProviderIdSchema = z.enum([
 ]);
 
 export const VercelUserByokInferenceProviderIdSchema = z.enum([
+  'alibaba',
   'anthropic',
   'bedrock',
   'fireworks',
@@ -113,6 +114,7 @@ export const UserByokProviderIdSchema = VercelUserByokInferenceProviderIdSchema.
 export type UserByokProviderId = z.infer<typeof UserByokProviderIdSchema>;
 
 export const UserByokTestModels = {
+  [VercelUserByokInferenceProviderIdSchema.enum.alibaba]: 'alibaba/qwen3-coder-plus',
   [VercelUserByokInferenceProviderIdSchema.enum.anthropic]: 'anthropic/claude-haiku-4.5',
   [VercelUserByokInferenceProviderIdSchema.enum.bedrock]: 'anthropic/claude-haiku-4.5',
   [VercelUserByokInferenceProviderIdSchema.enum.fireworks]: 'openai/gpt-oss-20b',
@@ -135,7 +137,6 @@ export const UserByokTestModels = {
 } satisfies Record<UserByokProviderId, string>;
 
 export const VercelNonUserByokInferenceProviderIdSchema = z.enum([
-  'alibaba',
   'arcee-ai',
   'azure',
   'baseten',
@@ -201,7 +202,7 @@ const modelPrefixToVercelInferenceProviderMapping = {
   openai: VercelUserByokInferenceProviderIdSchema.enum.openai,
   minimax: VercelUserByokInferenceProviderIdSchema.enum.minimax,
   mistralai: VercelUserByokInferenceProviderIdSchema.enum.mistral,
-  qwen: VercelNonUserByokInferenceProviderIdSchema.enum.alibaba,
+  qwen: VercelUserByokInferenceProviderIdSchema.enum.alibaba,
   'x-ai': VercelUserByokInferenceProviderIdSchema.enum.xai,
   'z-ai': VercelUserByokInferenceProviderIdSchema.enum.zai,
 } as Record<string, VercelInferenceProviderId | undefined>;

apps/web/src/lib/ai-gateway/providers/provider-definitions.ts

@@ -1,8 +1,5 @@
 import { getEnvVariable } from '@/lib/dotenvx';
-import {
-  addCacheBreakpoints,
-  isReasoningExplicitlyDisabled,
-} from '@/lib/ai-gateway/providers/openrouter/request-helpers';
+import { isReasoningExplicitlyDisabled } from '@/lib/ai-gateway/providers/openrouter/request-helpers';
 import type { Provider } from '@/lib/ai-gateway/providers/types';
 import { applyVercelSettings } from '@/lib/ai-gateway/providers/vercel';
 
@@ -22,7 +19,6 @@ export default {
     supportedChatApis: ['chat_completions' /*, 'responses'*/],
     transformRequest(context) {
       context.request.body.enable_thinking = !isReasoningExplicitlyDisabled(context.request);
-      addCacheBreakpoints(context.request);
     },
   },
   SEED: {