@ai-sdk/gateway 4.0.2 → 4.0.3

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @ai-sdk/gateway
 
+## 4.0.3
+
+### Patch Changes
+
+- 728eaa0: feat(provider/gateway): add `has` provider option to restrict routing to models with given capabilities (e.g. `implicit-caching`)
+
 ## 4.0.2
 
 ### Patch Changes

package/dist/index.d.ts

@@ -786,6 +786,11 @@ type GatewayProviderOptions = {
     caching?: 'auto';
     /** Filter to providers that do not train on prompt data. */
     disallowPromptTraining?: boolean;
+    /**
+     * Restrict routing to models that have all of the given capabilities.
+     * Currently supports `'implicit-caching'`.
+     */
+    has?: Array<'implicit-caching'>;
     /** Filter to providers that are HIPAA compliant with Vercel AI Gateway. */
     hipaaCompliant?: boolean;
     /** Array of model slugs specifying fallback models to use in order. */

package/dist/index.js

@@ -2356,7 +2356,7 @@ async function getVercelRequestId() {
 }
 
 // src/version.ts
-var VERSION = true ? "4.0.2" : "0.0.0-test";
+var VERSION = true ? "4.0.3" : "0.0.0-test";
 
 // src/gateway-provider.ts
 var AI_GATEWAY_PROTOCOL_VERSION = "0.0.1";

package/docs/00-ai-gateway.mdx

@@ -1045,6 +1045,12 @@ The following gateway provider options are available:
 
   The unique identifier for the entity against which quota is tracked. Used for quota management and enforcement purposes.
 
+- **has** _Array<'implicit-caching'>_
+
+  Restricts routing to provider models that have all of the specified capabilities. Currently supports `'implicit-caching'`, which limits routing to models that perform automatic (implicit) prompt caching. Applies to both BYOK and system credentials, since the capability is a property of the model rather than the credential. If no provider model for the requested model satisfies the capabilities, the request fails. Unsupported values are rejected.
+
+  Example: `has: ['implicit-caching']` will only route to models that support implicit caching.
+
 - **providerTimeouts** _object_
 
   Per-provider timeouts for BYOK credentials in milliseconds. Controls how long to wait for a provider to start responding before falling back to the next available provider.
@@ -1198,6 +1204,25 @@ const { text } = await generateText({
 });
 ```
 
+#### Filtering by Model Capability
+
+Set `has` to restrict routing to provider models that have the specified capabilities. This applies to both BYOK and system credentials, since the capability is a property of the model rather than the credential. Currently `'implicit-caching'` is supported, which limits routing to models that perform automatic prompt caching. If no provider model for the requested model satisfies the capabilities, the request fails.
+
+```ts
+import type { GatewayProviderOptions } from '@ai-sdk/gateway';
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: 'openai/gpt-5.5',
+  prompt: 'Summarize this report...',
+  providerOptions: {
+    gateway: {
+      has: ['implicit-caching'],
+    } satisfies GatewayProviderOptions,
+  },
+});
+```
+
 ### Provider-Specific Options
 
 When using provider-specific options through AI Gateway, use the actual provider name (e.g. `anthropic`, `openai`, not `gateway`) as the key:

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@ai-sdk/gateway",
   "private": false,
-  "version": "4.0.2",
+  "version": "4.0.3",
   "type": "module",
   "license": "Apache-2.0",
   "sideEffects": false,

package/src/gateway-provider-options.ts

@@ -15,6 +15,12 @@ export type GatewayProviderOptions = {
   /** Filter to providers that do not train on prompt data. */
   disallowPromptTraining?: boolean;
 
+  /**
+   * Restrict routing to models that have all of the given capabilities.
+   * Currently supports `'implicit-caching'`.
+   */
+  has?: Array<'implicit-caching'>;
+
   /** Filter to providers that are HIPAA compliant with Vercel AI Gateway. */
   hipaaCompliant?: boolean;