@ai-sdk/openai 4.0.0-beta.11 → 4.0.0-beta.13

package/docs/03-openai.mdx

@@ -257,6 +257,11 @@ The following provider options are available:
 - **forceReasoning** _boolean_
   Force treating this model as a reasoning model. This is useful for "stealth" reasoning models (e.g. via a custom baseURL) where the model ID is not recognized by the SDK's allowlist. When enabled, the SDK applies reasoning-model parameter compatibility rules and defaults `systemMessageMode` to `developer` unless overridden.
 
+- **contextManagement** _Array<object>_
+  Enable server-side context management (compaction). When configured, the server automatically compresses conversation context when token usage crosses a specified threshold. Each object in the array should have:
+  - `type`: `'compaction'`
+  - `compactThreshold`: _number_ — the token count at which compaction is triggered
+
 The OpenAI responses provider also returns provider-specific metadata:
 
 For Responses models, you can type this metadata using `OpenaiResponsesProviderMetadata`:
@@ -1511,6 +1516,125 @@ for (const part of result.content) {
   are fields like `filename` that are directly available on the source object.
 </Note>
 
+#### Compaction
+
+The OpenAI Responses API supports server-side context compaction. When enabled, the server automatically compresses conversation context when token usage crosses a configured threshold. This is useful for long-running conversations or agent loops where you want to stay within token limits without manually managing context.
+
+The compaction item returned by the server is opaque and encrypted — it carries forward key prior state and reasoning into the next turn using fewer tokens. The AI SDK handles this automatically: compaction items are returned as text parts with special `providerMetadata`, and when passed back in subsequent requests they are sent as compaction input items.
+
+```ts highlight="7-11"
+import {
+  openai,
+  type OpenAILanguageModelResponsesOptions,
+} from '@ai-sdk/openai';
+import { generateText } from 'ai';
+
+const result = await generateText({
+  model: openai.responses('gpt-5.2'),
+  messages: conversationHistory,
+  providerOptions: {
+    openai: {
+      store: false,
+      contextManagement: [{ type: 'compaction', compactThreshold: 50000 }],
+    } satisfies OpenAILanguageModelResponsesOptions,
+  },
+});
+```
+
+**Configuration:**
+
+- **type** — Must be `'compaction'`
+- **compactThreshold** — The token count at which compaction is triggered. When the rendered input token count crosses this threshold, the server runs a compaction pass before continuing inference.
+
+<Note>
+  Server-side compaction is ZDR-friendly when you set `store: false` on your
+  requests.
+</Note>
+
+##### Detecting Compaction in Streams
+
+When using `streamText`, you can detect compaction by checking the `providerMetadata` on `text-start` and `text-end` events:
+
+```ts
+import {
+  openai,
+  type OpenAILanguageModelResponsesOptions,
+} from '@ai-sdk/openai';
+import { streamText } from 'ai';
+
+const result = streamText({
+  model: openai.responses('gpt-5.2'),
+  messages: conversationHistory,
+  providerOptions: {
+    openai: {
+      store: false,
+      contextManagement: [{ type: 'compaction', compactThreshold: 50000 }],
+    } satisfies OpenAILanguageModelResponsesOptions,
+  },
+});
+
+for await (const part of result.fullStream) {
+  switch (part.type) {
+    case 'text-start': {
+      const isCompaction = part.providerMetadata?.openai?.type === 'compaction';
+      if (isCompaction) {
+        // ... your logic
+      }
+      break;
+    }
+    case 'text-end': {
+      const isCompaction = part.providerMetadata?.openai?.type === 'compaction';
+      if (isCompaction) {
+        // ... your logic
+      }
+      break;
+    }
+    case 'text-delta': {
+      process.stdout.write(part.text);
+      break;
+    }
+  }
+}
+```
+
+##### Compaction in UI Applications
+
+When using `useChat` or other UI hooks, compaction items appear as text parts with `providerMetadata`. You can detect and style them differently in your UI:
+
+```tsx
+{
+  message.parts.map((part, index) => {
+    if (part.type === 'text') {
+      const isCompaction =
+        (part.providerMetadata?.openai as { type?: string } | undefined)
+          ?.type === 'compaction';
+
+      if (isCompaction) {
+        return (
+          <div
+            key={index}
+            className="bg-yellow-100 border-l-4 border-yellow-500 p-2"
+          >
+            <span className="font-bold">[Context Compacted]</span>
+            <p className="text-sm text-yellow-700">
+              The server compressed the conversation context to reduce token
+              usage.
+            </p>
+          </div>
+        );
+      }
+      return <div key={index}>{part.text}</div>;
+    }
+  });
+}
+```
+
+The metadata includes the following fields:
+
+- **type** — Always `'compaction'`
+- **itemId** _string_ — The ID of the compaction item in the Responses API
+- **encryptedContent** _string_ (optional) — The encrypted compaction state. This is automatically sent back to the API when the message is included in subsequent requests.
+
 ### Chat Models
 
 You can create models that call the [OpenAI chat API](https://platform.openai.com/docs/api-reference/chat) using the `.chat()` factory method.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/openai",
-  "version": "4.0.0-beta.11",
+  "version": "4.0.0-beta.13",
   "license": "Apache-2.0",
   "sideEffects": false,
   "main": "./dist/index.js",
@@ -36,8 +36,8 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/provider": "4.0.0-beta.3",
-    "@ai-sdk/provider-utils": "5.0.0-beta.5"
+    "@ai-sdk/provider": "4.0.0-beta.4",
+    "@ai-sdk/provider-utils": "5.0.0-beta.6"
   },
   "devDependencies": {
     "@types/node": "20.17.24",
@@ -71,9 +71,7 @@
     "build": "pnpm clean && tsup --tsconfig tsconfig.build.json",
     "build:watch": "pnpm clean && tsup --watch",
     "clean": "del-cli dist docs *.tsbuildinfo",
-    "lint": "eslint \"./**/*.ts*\"",
     "type-check": "tsc --build",
-    "prettier-check": "prettier --check \"./**/*.ts*\"",
     "test": "pnpm test:node && pnpm test:edge",
     "test:update": "pnpm test:node -u",
     "test:watch": "vitest --config vitest.node.config.js",

package/src/chat/openai-chat-language-model.ts

@@ -17,6 +17,7 @@ import {
   createEventSourceResponseHandler,
   createJsonResponseHandler,
   generateId,
+  isCustomReasoning,
   isParsableJson,
   parseProviderOptions,
   postJsonToApi,
@@ -81,6 +82,7 @@ export class OpenAIChatLanguageModel implements LanguageModelV4 {
     seed,
     tools,
     toolChoice,
+    reasoning,
     providerOptions,
   }: LanguageModelV4CallOptions) {
     const warnings: SharedV4Warning[] = [];
@@ -94,6 +96,12 @@ export class OpenAIChatLanguageModel implements LanguageModelV4 {
       })) ?? {};
 
     const modelCapabilities = getOpenAILanguageModelCapabilities(this.modelId);
+
+    // AI SDK reasoning values map directly to the OpenAI reasoning values.
+    const resolvedReasoningEffort =
+      openaiOptions.reasoningEffort ??
+      (isCustomReasoning(reasoning) ? reasoning : undefined);
+
     const isReasoningModel =
       openaiOptions.forceReasoning ?? modelCapabilities.isReasoningModel;
 
@@ -168,7 +176,7 @@ export class OpenAIChatLanguageModel implements LanguageModelV4 {
       store: openaiOptions.store,
       metadata: openaiOptions.metadata,
       prediction: openaiOptions.prediction,
-      reasoning_effort: openaiOptions.reasoningEffort,
+      reasoning_effort: resolvedReasoningEffort,
       service_tier: openaiOptions.serviceTier,
       prompt_cache_key: openaiOptions.promptCacheKey,
       prompt_cache_retention: openaiOptions.promptCacheRetention,
@@ -184,7 +192,7 @@ export class OpenAIChatLanguageModel implements LanguageModelV4 {
       // when reasoning effort is none, gpt-5.1 models allow temperature, topP, logprobs
       //  https://platform.openai.com/docs/guides/latest-model#gpt-5-1-parameter-compatibility
       if (
-        openaiOptions.reasoningEffort !== 'none' ||
+        resolvedReasoningEffort !== 'none' ||
         !modelCapabilities.supportsNonReasoningParameters
       ) {
         if (baseArgs.temperature != null) {

package/src/index.ts

@@ -15,6 +15,7 @@ export type { OpenAIEmbeddingModelOptions } from './embedding/openai-embedding-o
 export type { OpenAISpeechModelOptions } from './speech/openai-speech-options';
 export type { OpenAITranscriptionModelOptions } from './transcription/openai-transcription-options';
 export type {
+  OpenaiResponsesCompactionProviderMetadata,
   OpenaiResponsesProviderMetadata,
   OpenaiResponsesReasoningProviderMetadata,
   OpenaiResponsesTextProviderMetadata,

package/src/responses/convert-to-openai-responses-input.ts

@@ -23,15 +23,16 @@ import {
 } from '../tool/local-shell';
 import { shellInputSchema, shellOutputSchema } from '../tool/shell';
 import {
-  toolSearchInputSchema,
-  toolSearchOutputSchema,
-} from '../tool/tool-search';
-import {
+  OpenAIResponsesCompactionItem,
   OpenAIResponsesCustomToolCallOutput,
   OpenAIResponsesFunctionCallOutput,
   OpenAIResponsesInput,
   OpenAIResponsesReasoning,
 } from './openai-responses-api';
+import {
+  toolSearchInputSchema,
+  toolSearchOutputSchema,
+} from '../tool/tool-search';
 
 /**
  * Check if a string is a file ID based on the given prefixes
@@ -543,6 +544,36 @@ export async function convertToOpenAIResponsesInput({
               }
               break;
             }
+
+            case 'custom': {
+              if (part.kind === 'openai-compaction') {
+                const providerOpts =
+                  part.providerOptions?.[providerOptionsName];
+                const id = providerOpts?.itemId as string | undefined;
+
+                if (hasConversation && id != null) {
+                  break;
+                }
+
+                if (store && id != null) {
+                  input.push({ type: 'item_reference', id });
+                  break;
+                }
+
+                const encryptedContent = providerOpts?.encryptedContent as
+                  | string
+                  | undefined;
+
+                if (id != null) {
+                  input.push({
+                    type: 'compaction',
+                    id,
+                    encrypted_content: encryptedContent!,
+                  } satisfies OpenAIResponsesCompactionItem);
+                }
+              }
+              break;
+            }
           }
         }
 

package/src/responses/openai-responses-api.ts

@@ -34,7 +34,8 @@ export type OpenAIResponsesInputItem =
   | OpenAIResponsesToolSearchCall
   | OpenAIResponsesToolSearchOutput
   | OpenAIResponsesReasoning
-  | OpenAIResponsesItemReference;
+  | OpenAIResponsesItemReference
+  | OpenAIResponsesCompactionItem;
 
 export type OpenAIResponsesIncludeValue =
   | 'web_search_call.action.sources'
@@ -235,6 +236,12 @@ export type OpenAIResponsesItemReference = {
   id: string;
 };
 
+export type OpenAIResponsesCompactionItem = {
+  type: 'compaction';
+  id: string;
+  encrypted_content: string;
+};
+
 /**
  * A filter used to compare a specified attribute key to a given value using a defined comparison operation.
  */
@@ -611,6 +618,11 @@ export const openaiResponsesChunkSchema = lazySchema(() =>
               commands: z.array(z.string()),
             }),
           }),
+          z.object({
+            type: z.literal('compaction'),
+            id: z.string(),
+            encrypted_content: z.string().nullish(),
+          }),
           z.object({
             type: z.literal('shell_call_output'),
             id: z.string(),
@@ -850,6 +862,11 @@ export const openaiResponsesChunkSchema = lazySchema(() =>
               commands: z.array(z.string()),
             }),
           }),
+          z.object({
+            type: z.literal('compaction'),
+            id: z.string(),
+            encrypted_content: z.string(),
+          }),
           z.object({
             type: z.literal('shell_call_output'),
             id: z.string(),
@@ -1289,6 +1306,11 @@ export const openaiResponsesResponseSchema = lazySchema(() =>
                 commands: z.array(z.string()),
               }),
             }),
+            z.object({
+              type: z.literal('compaction'),
+              id: z.string(),
+              encrypted_content: z.string(),
+            }),
             z.object({
               type: z.literal('shell_call_output'),
               id: z.string(),

package/src/responses/openai-responses-language-model.ts

@@ -21,6 +21,7 @@ import {
   createToolNameMapping,
   generateId,
   InferSchema,
+  isCustomReasoning,
   parseProviderOptions,
   ParseResult,
   postJsonToApi,
@@ -67,6 +68,7 @@ import {
 } from './openai-responses-options';
 import { prepareResponsesTools } from './openai-responses-prepare-tools';
 import {
+  ResponsesCompactionProviderMetadata,
   ResponsesProviderMetadata,
   ResponsesReasoningProviderMetadata,
   ResponsesSourceDocumentProviderMetadata,
@@ -129,6 +131,7 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
     frequencyPenalty,
     seed,
     prompt,
+    reasoning,
     providerOptions,
     tools,
     toolChoice,
@@ -174,6 +177,10 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
       });
     }
 
+    const resolvedReasoningEffort =
+      openaiOptions?.reasoningEffort ??
+      (isCustomReasoning(reasoning) ? reasoning : undefined);
+
     const isReasoningModel =
       openaiOptions?.forceReasoning ?? modelCapabilities.isReasoningModel;
 
@@ -337,13 +344,21 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
       top_logprobs: topLogprobs,
       truncation: openaiOptions?.truncation,
 
+      // context management (server-side compaction):
+      ...(openaiOptions?.contextManagement && {
+        context_management: openaiOptions.contextManagement.map(cm => ({
+          type: cm.type,
+          compact_threshold: cm.compactThreshold,
+        })),
+      }),
+
       // model-specific settings:
       ...(isReasoningModel &&
-        (openaiOptions?.reasoningEffort != null ||
+        (resolvedReasoningEffort != null ||
           openaiOptions?.reasoningSummary != null) && {
           reasoning: {
-            ...(openaiOptions?.reasoningEffort != null && {
-              effort: openaiOptions.reasoningEffort,
+            ...(resolvedReasoningEffort != null && {
+              effort: resolvedReasoningEffort,
             }),
             ...(openaiOptions?.reasoningSummary != null && {
               summary: openaiOptions.reasoningSummary,
@@ -359,7 +374,7 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
       //  https://platform.openai.com/docs/guides/latest-model#gpt-5-1-parameter-compatibility
       if (
         !(
-          openaiOptions?.reasoningEffort === 'none' &&
+          resolvedReasoningEffort === 'none' &&
           modelCapabilities.supportsNonReasoningParameters
         )
       ) {
@@ -977,6 +992,21 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
 
           break;
         }
+
+        case 'compaction': {
+          content.push({
+            type: 'custom',
+            kind: 'openai-compaction',
+            providerMetadata: {
+              [providerOptionsName]: {
+                type: 'compaction',
+                itemId: part.id,
+                encryptedContent: part.encrypted_content,
+              } satisfies ResponsesCompactionProviderMetadata,
+            },
+          });
+          break;
+        }
       }
     }
 
@@ -1782,6 +1812,18 @@ export class OpenAIResponsesLanguageModel implements LanguageModelV4 {
                 }
 
                 delete activeReasoning[value.item.id];
+              } else if (value.item.type === 'compaction') {
+                controller.enqueue({
+                  type: 'custom',
+                  kind: 'openai-compaction',
+                  providerMetadata: {
+                    [providerOptionsName]: {
+                      type: 'compaction',
+                      itemId: value.item.id,
+                      encryptedContent: value.item.encrypted_content,
+                    } satisfies ResponsesCompactionProviderMetadata,
+                  },
+                });
               }
             } else if (isResponseFunctionCallArgumentsDeltaChunk(value)) {
               const toolCall = ongoingToolCalls[value.output_index];

package/src/responses/openai-responses-options.ts

@@ -300,6 +300,18 @@ export const openaiLanguageModelResponsesOptionsSchema = lazySchema(() =>
        * and defaults `systemMessageMode` to `developer` unless overridden.
        */
       forceReasoning: z.boolean().optional(),
+
+      /**
+       * Enable server-side context management (compaction).
+       */
+      contextManagement: z
+        .array(
+          z.object({
+            type: z.literal('compaction'),
+            compactThreshold: z.number(),
+          }),
+        )
+        .nullish(),
     }),
   ),
 );

package/src/responses/openai-responses-provider-metadata.ts

@@ -30,6 +30,16 @@ export type OpenaiResponsesProviderMetadata = {
   openai: ResponsesProviderMetadata;
 };
 
+export type ResponsesCompactionProviderMetadata = {
+  type: 'compaction';
+  itemId: string;
+  encryptedContent?: string;
+};
+
+export type OpenaiResponsesCompactionProviderMetadata = {
+  openai: ResponsesCompactionProviderMetadata;
+};
+
 export type ResponsesTextProviderMetadata = {
   itemId: string;
   phase?: 'commentary' | 'final_answer' | null;