ai 7.0.28 → 7.0.29

package/CHANGELOG.md

@@ -1,5 +1,13 @@
 # ai
 
+## 7.0.29
+
+### Patch Changes
+
+- Updated dependencies [7069785]
+- Updated dependencies [4bf9ac2]
+  - @ai-sdk/gateway@4.0.21
+
 ## 7.0.28
 
 ### Patch Changes

package/dist/index.js

@@ -1089,7 +1089,7 @@ import {
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "7.0.28" : "0.0.0-test";
+var VERSION = true ? "7.0.29" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({

package/dist/internal/index.js

@@ -85,7 +85,7 @@ import {
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "7.0.28" : "0.0.0-test";
+var VERSION = true ? "7.0.29" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({

package/docs/02-foundations/02-providers-and-models.mdx

@@ -14,10 +14,13 @@ To solve these challenges, AI SDK Core offers a standardized approach to interac
 Here is an overview of the AI SDK Provider Architecture:
 
 <MDXImage
+  alt="Diagram showing the AI SDK Core API, provider specifications, and provider implementations"
   srcLight="/images/ai-sdk-diagram.png"
   srcDark="/images/ai-sdk-diagram-dark.png"
-  width={800}
-  height={800}
+  width={1694}
+  height={1206}
+  widthDark={2541}
+  heightDark={1884}
 />
 
 ## AI SDK Providers

package/docs/02-foundations/04-tools.mdx

@@ -37,7 +37,7 @@ Tools with an `execute` function are run automatically when these calls are gene
 The output of the tool calls are returned using tool result objects.
 
 You can automatically pass tool results back to the LLM
-using [multi-step calls](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls) with `streamText` and `generateText`.
+using [multi-step calls](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls-using-stopwhen) with `streamText` and `generateText`.
 
 ## Types of Tools
 

package/docs/02-getting-started/02-nextjs-app-router.mdx

@@ -417,7 +417,7 @@ By setting `stopWhen: isStepCount(5)`, you're allowing the model to use up to 5
 
 Update your `app/api/chat/route.ts` file to add a new tool to convert the temperature from Fahrenheit to Celsius:
 
-```tsx filename="app/api/chat/route.ts" highlight="33-48"
+```tsx filename="app/api/chat/route.ts" highlight="34-47"
 import {
   streamText,
   UIMessage,

package/docs/02-getting-started/03-nextjs-pages-router.mdx

@@ -402,7 +402,7 @@ By setting `stopWhen: isStepCount(5)`, you're allowing the model to use up to 5
 
 Update your `app/api/chat/route.ts` file to add a new tool to convert the temperature from Fahrenheit to Celsius:
 
-```tsx filename="app/api/chat/route.ts" highlight="33-48"
+```tsx filename="app/api/chat/route.ts" highlight="34-47"
 import {
   streamText,
   UIMessage,

package/docs/02-getting-started/06-nodejs.mdx

@@ -250,7 +250,7 @@ Try asking something like "What's the weather in New York?" and see how the agen
 
 Notice the blank "assistant" response? This is because instead of generating a text response, the agent generated a tool call. You can access the tool call and subsequent tool result in the `toolCall` and `toolResult` keys of the result object.
 
-```typescript highlight="46-47"
+```typescript highlight="50-51"
 import { ModelMessage, streamText, tool } from 'ai';
 __PROVIDER_IMPORT__;
 import 'dotenv/config';

package/docs/03-agents/08-terminal-ui.mdx

@@ -5,10 +5,10 @@ description: Run AI SDK agents in an interactive terminal UI.
 
 # Terminal UI
 
-The `@ai-sdk/tui` package lets you run a `ToolLoopAgent` in an interactive
-terminal interface. It is useful for local development, demos, and internal
-tools where a terminal experience is enough and you do not want to build a
-custom UI.
+The `@ai-sdk/tui` package lets you run a local `ToolLoopAgent` or connect to a
+remote agent through a `ChatTransport` in an interactive terminal interface.
+It is useful for local development, demos, and internal tools where a terminal
+experience is enough and you do not want to build a custom UI.
 
 The terminal UI handles prompt input, streamed assistant responses, markdown
 rendering, tool cards, reasoning sections, scrolling, and tool approval prompts.
@@ -55,6 +55,27 @@ await runAgentTUI({
 
 `runAgentTUI` runs until the user exits with `Esc` or `Ctrl+C`.
 
+## Connecting to a Remote Agent
+
+Pass a `ChatTransport` instead of an `agent` to connect the terminal UI to a
+remote AI SDK UI message endpoint:
+
+```ts
+import { runAgentTUI } from '@ai-sdk/tui';
+import { DefaultChatTransport } from 'ai';
+
+await runAgentTUI({
+  title: 'Remote Agent',
+  transport: new DefaultChatTransport({
+    api: 'https://example.com/api/chat',
+  }),
+});
+```
+
+The transport controls the endpoint, authentication, request body, and other
+remote communication behavior. The terminal UI keeps its internal chat id and
+message history private to the transport contract.
+
 ## Sandbox
 
 Pass a sandbox session with the `sandbox` option when your agent tools need an
@@ -134,10 +155,10 @@ await runAgentTUI({ title: 'Weather Agent', agent });
 
 ## Compatibility
 
-`runAgentTUI` is designed for agents that can be run directly from terminal user
-input. The agent must not require per-call options and must not use structured
+When using the `agent` option, the agent must be runnable directly from terminal
+user input. It must not require per-call options and must not use structured
 output, because the terminal UI cannot infer those values from a free-form
-prompt.
+prompt. Use a `transport` for remote agents that need custom request handling.
 
 Use `agent.generate()` or `agent.stream()` directly for examples or apps that
 need fixed prompts, call options, structured output, custom result inspection, or

package/docs/03-ai-sdk-core/01-overview.mdx

@@ -17,7 +17,7 @@ For example, here’s how you can generate text with various models using the AI
 ## AI SDK Core Functions
 
 AI SDK Core has various functions designed for [text generation](./generating-text), [structured data generation](./generating-structured-data), and [tool usage](./tools-and-tool-calling).
-These functions take a standardized approach to setting up [prompts](./prompts) and [settings](./settings), making it easier to work with different models.
+These functions take a standardized approach to setting up [prompts](/docs/foundations/prompts) and [settings](./settings), making it easier to work with different models.
 
 - [`generateText`](/docs/ai-sdk-core/generating-text): Generates text and [tool calls](./tools-and-tool-calling).
   This function is ideal for non-interactive use cases such as automation tasks where you need to write text (e.g. drafting email or summarizing web pages) and for agents that use tools.

package/docs/03-ai-sdk-core/05-generating-text.mdx

@@ -29,7 +29,7 @@ const { text } = await generateText({
 });
 ```
 
-You can use more [advanced prompts](./prompts) to generate text with more complex instructions and content:
+You can use more [advanced prompts](/docs/foundations/prompts) to generate text with more complex instructions and content:
 
 ```tsx
 import { generateText } from 'ai';
@@ -97,7 +97,7 @@ When using `generateText`, you can provide an `onEnd` callback that is triggered
 ).
 It contains the text, usage information, finish reason, messages, steps, total usage, and more:
 
-```tsx highlight="6-8"
+```tsx highlight="7-10"
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -262,7 +262,7 @@ Errors become part of the stream and are not thrown to prevent e.g. servers from
 
 To log errors, you can provide an `onError` callback that is triggered when an error occurs.
 
-```tsx highlight="6-8"
+```tsx highlight="7-9"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -308,7 +308,7 @@ It receives all stream part types from `stream`, including:
 - `error`
 - `raw`
 
-```tsx highlight="6-11"
+```tsx highlight="7-12"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -331,7 +331,7 @@ When using `streamText`, you can provide an `onEnd` callback that is triggered w
 ).
 It contains the text, usage information, finish reason, messages, steps, total usage, and more:
 
-```tsx highlight="6-8"
+```tsx highlight="7-10"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 

package/docs/03-ai-sdk-core/10-generating-structured-data.mdx

@@ -114,7 +114,7 @@ You can consume the structured output on the client with the [`useObject`](/docs
 
 To handle errors, provide an `onError` callback:
 
-```tsx highlight="5-7"
+```tsx highlight="6-8"
 import { streamText, Output } from 'ai';
 
 const result = streamText({
@@ -318,7 +318,7 @@ const { output } = await generateText({
 
 You can add `.describe("...")` to individual schema properties to give the model hints about what each property is for. This helps improve the quality and accuracy of generated structured data:
 
-```ts highlight="5,9"
+```ts highlight="9,16,19-20"
 import { generateText, Output } from 'ai';
 __PROVIDER_IMPORT__;
 import { z } from 'zod';
@@ -355,7 +355,7 @@ Property descriptions are particularly useful for:
 
 You can optionally specify a `name` and `description` for the output. These are used by some providers for additional LLM guidance, e.g. via tool or schema name.
 
-```ts highlight="6-7"
+```ts highlight="8-9"
 import { generateText, Output } from 'ai';
 __PROVIDER_IMPORT__;
 import { z } from 'zod';

package/docs/03-ai-sdk-core/15-tools-and-tool-calling.mdx

@@ -20,7 +20,7 @@ Function tools and dynamic tools contain several core elements:
 
 The `tools` parameter of `generateText` and `streamText` is an object that has the tool names as keys and the tools as values:
 
-```ts highlight="6-17"
+```ts highlight="7-18"
 import { z } from 'zod';
 import { generateText, tool, isStepCount } from 'ai';
 __PROVIDER_IMPORT__;
@@ -63,7 +63,7 @@ model for each generation step. It receives the matching tool `context` from
 `toolsContext` and the current `experimental_sandbox`, if one was provided. If `prepareStep`
 updates `toolsContext` or `experimental_sandbox`, the next step uses those updated values.
 
-```ts highlight="6-14,28-33"
+```ts highlight="5-16,32-35"
 import { generateText, tool } from 'ai';
 import { z } from 'zod';
 
@@ -413,7 +413,7 @@ In the following example, there are two steps:
    1. The tool result is sent to the model.
    1. The model generates a response considering the tool result.
 
-```ts highlight="18-19"
+```ts highlight="19-20"
 import { z } from 'zod';
 import { generateText, tool, isStepCount } from 'ai';
 __PROVIDER_IMPORT__;
@@ -447,7 +447,7 @@ It contains all the text, tool calls, tool results, per-step `performance`, and
 
 #### Example: Extract tool results from all steps
 
-```ts highlight="3,9-10"
+```ts highlight="4,10-11"
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -772,7 +772,7 @@ It supports the following settings:
 - `none`: the model must not call tools
 - `{ type: 'tool', toolName: string (typed) }`: the model must call the specified tool
 
-```ts highlight="18"
+```ts highlight="19"
 import { z } from 'zod';
 import { generateText, tool } from 'ai';
 __PROVIDER_IMPORT__;
@@ -877,7 +877,7 @@ const result = await generateText({
 The abort signals from `generateText` and `streamText` are forwarded to the tool execution.
 You can access them in the second parameter of the `execute` function and e.g. abort long-running computations or forward them to fetch calls inside tools.
 
-```ts highlight="6,11,14"
+```ts highlight="7,12,15"
 import { z } from 'zod';
 import { generateText, tool } from 'ai';
 __PROVIDER_IMPORT__;
@@ -1047,7 +1047,7 @@ The following tool input lifecycle hooks are available:
 
 ### Example
 
-```ts highlight="15-23"
+```ts highlight="16-24"
 import { streamText, tool } from 'ai';
 __PROVIDER_IMPORT__;
 import { z } from 'zod';
@@ -1334,7 +1334,7 @@ the AI SDK provides the `activeTools` property.
 It is an array of tool names that are currently active.
 By default, the value is `undefined` and all tools are active.
 
-```ts highlight="7"
+```ts highlight="8"
 import { openai } from '@ai-sdk/openai';
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
@@ -1357,7 +1357,7 @@ debugging. The list can be partial: tools listed in `toolOrder` are sent first
 in that order, and any remaining tools are sent afterwards in alphabetical
 order. Tool names are typed from your `tools` object.
 
-```ts highlight="9"
+```ts highlight="7"
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -1397,7 +1397,7 @@ that converts the tool result into a content part.
 
 Here is an example for converting a screenshot into a content part:
 
-```ts highlight="22-27"
+```ts highlight="23-38"
 const result = await generateText({
   model: __MODEL__,
   tools: {

package/docs/03-ai-sdk-core/25-settings.mdx

@@ -7,7 +7,7 @@ description: Learn how to configure the AI SDK.
 
 Large language models (LLMs) typically provide settings to augment their output.
 
-All AI SDK functions support the following common settings in addition to the model, the [prompt](./prompts), and additional provider-specific settings:
+All AI SDK functions support the following common settings in addition to the model, the [prompt](/docs/foundations/prompts), and additional provider-specific settings:
 
 ```ts highlight="3-6"
 const result = await generateText({

package/docs/03-ai-sdk-core/30-embeddings.mdx

@@ -109,7 +109,7 @@ Google's `gemini-embedding-2` model (also available as `gemini-embedding-2-previ
 
 The `embedMany` function now supports parallel processing with configurable `maxParallelCalls` to optimize performance:
 
-```ts highlight={"3"}
+```ts highlight={"4"}
 import { embedMany } from 'ai';
 
 const { embeddings, usage } = await embedMany({

package/docs/03-ai-sdk-core/31-reranking.mdx

@@ -101,7 +101,7 @@ Each item in the `ranking` array contains:
 
 Use `topN` to limit the number of results returned. This is useful for retrieving only the most relevant documents:
 
-```ts highlight={"7"}
+```ts highlight={"8"}
 import { cohere } from '@ai-sdk/cohere';
 import { rerank } from 'ai';
 
@@ -139,7 +139,7 @@ The `rerank` function accepts an optional `maxRetries` parameter of type `number
 that you can use to set the maximum number of retries for the reranking process.
 It defaults to `2` retries (3 attempts in total). You can set it to `0` to disable retries.
 
-```ts highlight={"7"}
+```ts highlight={"8"}
 import { cohere } from '@ai-sdk/cohere';
 import { rerank } from 'ai';
 
@@ -157,7 +157,7 @@ The `rerank` function accepts an optional `abortSignal` parameter of
 type [`AbortSignal`](https://developer.mozilla.org/en-US/docs/Web/API/AbortSignal)
 that you can use to abort the reranking process or set a timeout.
 
-```ts highlight={"7"}
+```ts highlight={"8"}
 import { cohere } from '@ai-sdk/cohere';
 import { rerank } from 'ai';
 
@@ -174,7 +174,7 @@ const { ranking } = await rerank({
 The `rerank` function accepts an optional `headers` parameter of type `Record<string, string>`
 that you can use to add custom headers to the reranking request.
 
-```ts highlight={"7"}
+```ts highlight={"8"}
 import { cohere } from '@ai-sdk/cohere';
 import { rerank } from 'ai';
 

package/docs/03-ai-sdk-core/35-image-generation.mdx

@@ -119,7 +119,7 @@ You can pass such settings to the `generateImage` function
 using the `providerOptions` parameter. The options for the provider
 (`openai` in the example below) become request body properties.
 
-```tsx highlight={"12"}
+```tsx highlight={"8-13"}
 import { generateImage } from 'ai';
 import { openai, type OpenAIImageModelGenerationOptions } from '@ai-sdk/openai';
 

package/docs/03-ai-sdk-core/36-transcription.mdx

@@ -98,7 +98,11 @@ The `audio` stream must contain raw audio chunks. `Uint8Array` chunks are raw by
   the provider.
 </Note>
 
-OpenAI streaming transcription uses `openai.transcription('gpt-realtime-whisper')`. xAI uses the same `xai.transcription()` model for request/response and streaming transcription; `experimental_streamTranscribe` uses xAI's WebSocket STT transport under the hood.
+OpenAI streaming transcription uses `openai.transcription('gpt-realtime-whisper')`.
+Cartesia uses `cartesia.transcription('ink-2')` for streaming-only Ink 2
+transcription. xAI uses the same `xai.transcription()` model for request/response
+and streaming transcription; `experimental_streamTranscribe` selects the
+provider's WebSocket STT transport.
 
 ```ts
 import { xai } from '@ai-sdk/xai';
@@ -150,7 +154,7 @@ const transcript = await transcribe({
 When `audio` is a URL, the SDK downloads the file with a default **2 GiB** size limit.
 You can customize this using `createDownload`:
 
-```ts highlight="1,8"
+```ts highlight="1,7"
 import { transcribe, createDownload } from 'ai';
 import { openai } from '@ai-sdk/openai';
 
@@ -163,7 +167,7 @@ const transcript = await transcribe({
 
 You can also provide a fully custom download function:
 
-```ts highlight="6-12"
+```ts highlight="7-13"
 import { transcribe } from 'ai';
 import { openai } from '@ai-sdk/openai';
 
@@ -206,7 +210,7 @@ that you can use to abort the transcription process or set a timeout.
 
 This is particularly useful when combined with URL downloads to prevent long-running requests:
 
-```ts highlight="8"
+```ts highlight="7"
 import { openai } from '@ai-sdk/openai';
 import { transcribe } from 'ai';
 
@@ -316,5 +320,6 @@ try {
 | [Google Vertex](/providers/ai-sdk-providers/google-vertex#transcription-models) | `telephony`              |
 | [xAI](/providers/ai-sdk-providers/xai#transcription-models)                     | `default`                |
 | [Cartesia](/providers/ai-sdk-providers/cartesia#transcription-models)           | `ink-whisper`            |
+| [Cartesia](/providers/ai-sdk-providers/cartesia#streaming-transcription-models) | `ink-2`                  |
 
 Above are a small subset of the transcription models supported by the AI SDK providers. For more, see the respective provider documentation.

package/docs/03-ai-sdk-core/37-speech.mdx

@@ -151,6 +151,7 @@ try {
 | [OpenAI](/providers/ai-sdk-providers/openai#speech-models)               | `tts-1`                             |
 | [OpenAI](/providers/ai-sdk-providers/openai#speech-models)               | `tts-1-hd`                          |
 | [OpenAI](/providers/ai-sdk-providers/openai#speech-models)               | `gpt-4o-mini-tts`                   |
+| [Mistral](/providers/ai-sdk-providers/mistral#speech-models)             | `voxtral-mini-tts-2603`             |
 | [ElevenLabs](/providers/ai-sdk-providers/elevenlabs#speech-models)       | `eleven_v3`                         |
 | [ElevenLabs](/providers/ai-sdk-providers/elevenlabs#speech-models)       | `eleven_multilingual_v2`            |
 | [ElevenLabs](/providers/ai-sdk-providers/elevenlabs#speech-models)       | `eleven_flash_v2_5`                 |

package/docs/03-ai-sdk-core/38-video-generation.mdx

@@ -146,7 +146,7 @@ const { videos } = await generateVideo({
 
 Some video models support generating videos from an input image. You can provide an image using the prompt object:
 
-```tsx highlight={"7-10"}
+```tsx highlight={"6-9"}
 import { experimental_generateVideo as generateVideo } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -177,7 +177,7 @@ Some video models support first-last-frame generation, where you provide the
 starting and/or ending frames of the video. Use the `frameImages` option to pass
 role-tagged images in a provider-agnostic way:
 
-```tsx highlight={"5-16"}
+```tsx highlight={"4-13"}
 const { video } = await generateVideo({
   model: __VIDEO_MODEL__,
   prompt: 'The cat walks across the scene and transforms into a dog by the end',
@@ -200,7 +200,7 @@ Some video models support reference-to-video generation, where you provide one o
 more reference images or videos that the model incorporates into the generated video. Use the `inputReferences` option to pass
 these inputs in a provider-agnostic way:
 
-```tsx highlight={"5-8"}
+```tsx highlight={"4-7"}
 const { video } = await generateVideo({
   model: __VIDEO_MODEL__,
   prompt: 'The two characters meet in a bustling market',
@@ -213,7 +213,7 @@ const { video } = await generateVideo({
 
 For URL-based video references, use the object form with an explicit `mediaType`:
 
-```tsx highlight={"5-10"}
+```tsx highlight={"4-9"}
 const { video } = await generateVideo({
   model: __VIDEO_MODEL__,
   prompt: 'Match the motion in the reference clip',
@@ -295,7 +295,7 @@ Video generation is an asynchronous process that can take several minutes to com
 
 You can configure the polling timeout using provider-specific options. Each provider exports a type for its options that you can use with `satisfies` for type safety:
 
-```tsx highlight={"10-12"}
+```tsx highlight={"8-12"}
 import { experimental_generateVideo as generateVideo } from 'ai';
 import { fal, type FalVideoModelOptions } from '@ai-sdk/fal';
 

package/docs/03-ai-sdk-core/50-error-handling.mdx

@@ -9,7 +9,7 @@ description: Learn how to handle errors in the AI SDK Core
 
 Regular errors are thrown and can be handled using the `try/catch` block.
 
-```ts highlight="3,8-10"
+```ts highlight="4,9-11"
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -31,7 +31,7 @@ When errors occur during streams that do not support error chunks,
 the error is thrown as a regular error.
 You can handle these errors using the `try/catch` block.
 
-```ts highlight="3,12-14"
+```ts highlight="4,13-15"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -56,7 +56,7 @@ You can handle those parts similar to other parts.
 It is recommended to also add a try-catch block for errors that
 happen outside of the streaming.
 
-```ts highlight="13-21"
+```ts highlight="14-22"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -99,7 +99,7 @@ When streams are aborted (e.g., via chat stop button), you may want to perform c
 
 The `onAbort` callback is called when a stream is aborted via `AbortSignal`, but `onEnd` is not called. This ensures you can still update your UI state appropriately.
 
-```ts highlight="5-9"
+```ts highlight="6-10"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -127,7 +127,7 @@ The `onAbort` callback receives:
 
 You can also handle abort events directly in the stream:
 
-```ts highlight="10-13"
+```ts highlight="11-14"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 

package/docs/03-ai-sdk-core/55-testing.mdx

@@ -181,7 +181,7 @@ const result = streamText({
 
 ### Simulate UI Message Stream Responses
 
-You can also simulate [UI Message Stream](/docs/ai-sdk-ui/stream-protocol#ui-message-stream) responses for testing,
+You can also simulate [UI Message Stream](/docs/ai-sdk-ui/stream-protocol#ui-message-stream-example) responses for testing,
 debugging, or demonstration purposes.
 
 Here is a Next example:

package/docs/03-ai-sdk-core/65-lifecycle-callbacks.mdx

@@ -55,7 +55,7 @@ Because callbacks run as part of the lifecycle, keep them fast or enqueue expens
 Use `onStart` and `onEnd` to record one application log for the beginning and end of a call.
 The `callId` is available across lifecycle events, so you can correlate logs from the same request.
 
-```tsx highlight="8-20"
+```tsx highlight="8-23"
 import { generateText } from 'ai';
 __PROVIDER_IMPORT__;
 
@@ -124,7 +124,7 @@ Use step events when you want timing that includes SDK-managed work such as loca
 When you use tools with `generateText` or `streamText`, a single user request can involve multiple model calls.
 Each model call is a step. The model may call a tool in one step, receive the tool result, and then produce a final answer in the next step.
 
-```tsx highlight="10-29"
+```tsx highlight="17-31"
 import { generateText, isStepCount, tool } from 'ai';
 import { z } from 'zod';
 __PROVIDER_IMPORT__;
@@ -171,7 +171,7 @@ This helps answer questions such as:
 Tool execution callbacks run around the tool's `execute` function.
 Use them to record tool usage, latency, successful results, and tool errors.
 
-```tsx highlight="17-35"
+```tsx highlight="19-36"
 import { generateText, tool } from 'ai';
 import { z } from 'zod';
 __PROVIDER_IMPORT__;
@@ -218,7 +218,7 @@ const result = await generateText({
 Embedding and reranking callbacks are simpler: they expose `onStart` and `onEnd` around the operation.
 This is useful for retrieval pipelines where you want to understand how often you are embedding, how many values you embed, and how reranking changes result sets.
 
-```tsx highlight="10-23"
+```tsx highlight="10-24,32-37"
 import { embedMany, rerank } from 'ai';
 import { cohere } from '@ai-sdk/cohere';
 
@@ -293,7 +293,7 @@ This distinction matters when the step includes local tool execution: the step d
 Lifecycle callbacks receive the full `runtimeContext` and `toolsContext` values that flow through the call.
 This makes callbacks useful for attaching application context without changing prompts or tool inputs.
 
-```tsx highlight="8-17,23-28"
+```tsx highlight="8-24,26-39"
 import { generateText, tool } from 'ai';
 import { z } from 'zod';
 __PROVIDER_IMPORT__;

package/docs/04-ai-sdk-ui/02-chatbot.mdx

@@ -1092,8 +1092,8 @@ By using `FileList`, you can send multiple files as attachments along with a mes
 <Note>
   Currently, only `image/*` and `text/*` content types get automatically
   converted into [multi-modal content
-  parts](/docs/foundations/prompts#multi-modal-messages). You will need to
-  handle other content types manually.
+  parts](/docs/foundations/prompts#file-parts). You will need to handle other
+  content types manually.
 </Note>
 
 ```tsx filename="app/page.tsx"

package/docs/04-ai-sdk-ui/03-chatbot-message-persistence.mdx

@@ -96,7 +96,7 @@ When processing messages on the server that contain tool calls, custom metadata,
 
 When your messages include tool calls, validate them against your tool definitions:
 
-```tsx filename="app/api/chat/route.ts" highlight="9-27,34-39"
+```tsx filename="app/api/chat/route.ts" highlight="7-8,14-27,35-45"
 import {
   convertToModelMessages,
   createUIMessageStreamResponse,
@@ -286,7 +286,7 @@ export default function Chat({
 
 When loading messages from storage that contain tools, metadata, or custom data
 parts, validate them using `validateUIMessages` before processing (see the
-[validation section](#validating-messages-from-database) above).
+[validation section](#validating-messages-on-the-server) above).
 
 </Note>
 
@@ -416,7 +416,7 @@ export async function POST(req: Request) {
 
 Alternatively, you can use `createUIMessageStream` to control the message ID by writing a start message part:
 
-```tsx filename="app/api/chat/route.ts" highlight="9-19"
+```tsx filename="app/api/chat/route.ts" highlight="14-18,25-27"
 import {
   generateId,
   streamText,

package/docs/04-ai-sdk-ui/20-streaming-data.mdx

@@ -124,7 +124,7 @@ export async function POST(req: Request) {
 <Note>
   You can also send stream data from custom backends, e.g. Python / FastAPI,
   using the [UI Message Stream
-  Protocol](/docs/ai-sdk-ui/stream-protocol#ui-message-stream-protocol).
+  Protocol](/docs/ai-sdk-ui/stream-protocol#data-stream-protocol).
 </Note>
 
 ## Types of Streamable Data

package/docs/05-ai-sdk-rsc/10-migrating-to-ui.mdx

@@ -25,7 +25,7 @@ Due to these limitations, AI SDK RSC is marked as experimental, and we do not re
 
 As a result, we strongly recommend migrating to AI SDK UI, which has undergone extensive development to provide a more stable and production grade experience.
 
-In building [v0](https://v0.dev), we have invested considerable time exploring how to create the best chat experience on the web. AI SDK UI ships with many of these best practices and commonly used patterns like [language model middleware](/docs/ai-sdk-core/middleware), [multi-step tool calls](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls), [attachments](/docs/ai-sdk-ui/chatbot#attachments-experimental), [telemetry](/docs/ai-sdk-core/telemetry), [provider registry](/docs/ai-sdk-core/provider-management#provider-registry), and many more. These features have been considerately designed into a neat abstraction that you can use to reliably integrate AI into your applications.
+In building [v0](https://v0.dev), we have invested considerable time exploring how to create the best chat experience on the web. AI SDK UI ships with many of these best practices and commonly used patterns like [language model middleware](/docs/ai-sdk-core/middleware), [multi-step tool calls](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls-using-stopwhen), [attachments](/docs/ai-sdk-ui/chatbot#attachments), [telemetry](/docs/ai-sdk-core/telemetry), [provider registry](/docs/ai-sdk-core/provider-management#provider-registry), and many more. These features have been considerately designed into a neat abstraction that you can use to reliably integrate AI into your applications.
 
 ## Streaming Chat Completions
 

package/docs/06-advanced/04-caching.mdx

@@ -112,7 +112,7 @@ export const cacheMiddleware: LanguageModelV4Middleware = {
 
 `LanguageModelV4Middleware` has two methods: `wrapGenerate` and `wrapStream`. `wrapGenerate` is called when using [`generateText`](/docs/reference/ai-sdk-core/generate-text), while `wrapStream` is called when using [`streamText`](/docs/reference/ai-sdk-core/stream-text).
 
-For `wrapGenerate`, you can cache the response directly. Instead, for `wrapStream`, you cache an array of the stream parts, which can then be used with [`simulateReadableStream`](/docs/ai-sdk-core/testing#simulate-data-stream-protocol-responses) function to create a simulated `ReadableStream` that returns the cached response. In this way, the cached response is returned chunk-by-chunk as if it were being generated by the model. You can control the initial delay and delay between chunks by adjusting the `initialDelayInMs` and `chunkDelayInMs` parameters of `simulateReadableStream`.
+For `wrapGenerate`, you can cache the response directly. Instead, for `wrapStream`, you cache an array of the stream parts, which can then be used with [`simulateReadableStream`](/docs/ai-sdk-core/testing#simulate-ui-message-stream-responses) function to create a simulated `ReadableStream` that returns the cached response. In this way, the cached response is returned chunk-by-chunk as if it were being generated by the model. You can control the initial delay and delay between chunks by adjusting the `initialDelayInMs` and `chunkDelayInMs` parameters of `simulateReadableStream`.
 
 You can see a full example of caching with Redis in a Next.js application in our [Caching Middleware Recipe](/cookbook/next/caching-middleware).
 

package/docs/07-reference/01-ai-sdk-core/22-dynamic-tool.mdx

@@ -14,7 +14,7 @@ The `dynamicTool` function creates tools where the input and output types are no
 
 Unlike the regular `tool` function, `dynamicTool` accepts and returns `unknown` types, allowing you to work with tools that have runtime-determined schemas.
 
-```ts highlight={"1,4,9,10,11"}
+```ts highlight={"1,4,9,11"}
 import { dynamicTool } from 'ai';
 import { z } from 'zod';
 

package/docs/07-reference/01-ai-sdk-core/33-safe-validate-ui-messages.mdx

@@ -5,7 +5,7 @@ description: API Reference for safeValidateUIMessages
 
 # `safeValidateUIMessages`
 
-`safeValidateUIMessages` is an async function that validates UI messages like [`validateUIMessages`](https://ai-sdk.dev/docs/reference/ai-sdk-core/validate-ui-messages), but instead of throwing it returns an object with a `success` key and either `data` or `error`.
+`safeValidateUIMessages` is an async function that validates UI messages like [`validateUIMessages`](/docs/reference/ai-sdk-core/validate-ui-messages), but instead of throwing it returns an object with a `success` key and either `data` or `error`.
 
 ## Basic Usage
 

package/docs/07-reference/01-ai-sdk-core/index.mdx

@@ -111,11 +111,6 @@ It also contains the following helper functions:
       description: 'Creates a client for connecting to MCP servers.',
       href: '/docs/reference/ai-sdk-core/create-mcp-client',
     },
-    {
-      title: 'validateJSONRPCMessage()',
-      description: 'Validates unknown values as MCP JSON-RPC messages.',
-      href: '/docs/reference/ai-sdk-core/validate-json-rpc-message',
-    },
     {
       title: 'MCP Apps',
       description:

package/docs/07-reference/06-ai-sdk-tui/01-run-agent-tui.mdx

@@ -5,9 +5,9 @@ description: API Reference for the runAgentTUI function.
 
 # `runAgentTUI()`
 
-Runs a `ToolLoopAgent` in an interactive terminal UI. The terminal UI reads user
-prompts, streams assistant responses, renders markdown, displays tool and
-reasoning sections, and handles manual tool approvals.
+Runs a local agent or chat transport in an interactive terminal UI. The
+terminal UI reads user prompts, streams assistant responses, renders markdown,
+displays tool and reasoning sections, and handles manual tool approvals.
 
 `runAgentTUI` runs until the user exits with `Esc` or `Ctrl+C`.
 
@@ -49,9 +49,16 @@ await runAgentTUI({
             {
               name: 'agent',
               type: 'AgentTUIAgent',
-              isRequired: true,
+              isOptional: true,
+              description:
+                'The agent to run. Provide exactly one of `agent` or `transport`. The agent must not require per-call options and must not use structured output.',
+            },
+            {
+              name: 'transport',
+              type: 'ChatTransport<UIMessage>',
+              isOptional: true,
               description:
-                'The agent to run. The agent must not require per-call options and must not use structured output.',
+                'The transport used to communicate with a remote agent. Provide exactly one of `agent` or `transport`.',
             },
             {
               name: 'title',
@@ -168,6 +175,19 @@ await runAgentTUI({
 });
 ```
 
+## Example with a Chat Transport
+
+```ts
+import { DefaultChatTransport } from 'ai';
+
+await runAgentTUI({
+  title: 'Remote Assistant',
+  transport: new DefaultChatTransport({
+    api: 'https://example.com/api/chat',
+  }),
+});
+```
+
 ## Example with Sandbox
 
 ```ts
@@ -192,8 +212,9 @@ working directory or exposed ports.
 
 ## Compatibility
 
-Use `runAgentTUI` for agents that can run directly from free-form user input.
-Use `agent.generate()` or `agent.stream()` directly when you need fixed prompts,
+Use the `agent` option for agents that can run directly from free-form user
+input. Use the `transport` option to communicate with a remote agent. Use
+`agent.generate()` or `agent.stream()` directly when you need fixed prompts,
 per-call options, structured output, custom result inspection, or custom stream
 processing.
 
@@ -201,4 +222,4 @@ processing.
 
 - [Terminal UI guide](/docs/agents/terminal-ui)
 - [Building Agents](/docs/agents/building-agents)
-- [ToolLoopAgent guide](/docs/agents/overview#tooloopagent-class)
+- [ToolLoopAgent guide](/docs/agents/overview#toolloopagent-class)

package/docs/08-migration-guides/23-migration-guide-7-0.mdx

@@ -1902,7 +1902,7 @@ const model = xai('grok-4.3');
 const chatModel = xai.chat('grok-4.3');
 ```
 
-# Migration Skill
+## Migration Skill
 
 Use the command below to add the migration skill and guide your agent
 

package/docs/08-migration-guides/29-migration-guide-4-0.mdx

@@ -817,7 +817,7 @@ The following options have been removed from the `useChat` hook:
 - `maxAutomaticRoundtrips`
 - `maxToolRoundtrips`
 
-Please use the [`maxSteps`](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls) option instead.
+Please use the [`maxSteps`](/docs/ai-sdk-core/tools-and-tool-calling#multi-step-calls-using-stopwhen) option instead.
 The value of `maxSteps` is equal to roundtrips + 1.
 
 ```ts filename="AI SDK 3.4"

package/docs/08-migration-guides/37-migration-guide-3-3.mdx

@@ -20,7 +20,7 @@ introducing significant improvements and new features across the AI SDK and its
 
 ### Open Telemetry Support
 
-- Added experimental [OpenTelemetry support](/docs/ai-sdk-core/telemetry#telemetry) for all [AI SDK Core functions](/docs/ai-sdk-core/overview#ai-sdk-core-functions), enabling better observability and tracing capabilities.
+- Added experimental [OpenTelemetry support](/docs/ai-sdk-core/telemetry) for all [AI SDK Core functions](/docs/ai-sdk-core/overview#ai-sdk-core-functions), enabling better observability and tracing capabilities.
 
 ### AI SDK UI Improvements
 

package/docs/08-migration-guides/39-migration-guide-3-1.mdx

@@ -129,7 +129,7 @@ async function submitMessage(userInput = 'What is the weather in SF?') {
 }
 ```
 
-With the new [`streamUI`](/docs/reference/ai-sdk-rsc/stream-ui) function, you can now use any compatible AI SDK provider. In this example, you import the AI SDK OpenAI provider. Then, you pass it to the [`model`](/docs/reference/ai-sdk-rsc/stream-ui#model) key of the new [`streamUI`](/docs/reference/ai-sdk-rsc/stream-ui) function. Finally, you declare a tool and return a React Server Component, defined in the [`generate`](/docs/reference/ai-sdk-rsc/stream-ui#tools-generate) key of the tool.
+With the new [`streamUI`](/docs/reference/ai-sdk-rsc/stream-ui) function, you can now use any compatible AI SDK provider. In this example, you import the AI SDK OpenAI provider. Then, you pass it to the [`model`](/docs/reference/ai-sdk-rsc/stream-ui#model) key of the new [`streamUI`](/docs/reference/ai-sdk-rsc/stream-ui) function. Finally, you declare a tool and return a React Server Component, defined in the [`generate`](/docs/reference/ai-sdk-rsc/stream-ui#tools.tool.generate) key of the tool.
 
 ```tsx
 import { streamUI } from '@ai-sdk/rsc';

package/docs/09-troubleshooting/14-tool-calling-with-structured-outputs.mdx

@@ -45,4 +45,4 @@ const result = await generateText({
 });
 ```
 
-For more information about using structured outputs with `generateText` and `streamText` see [Generating Structured Data](/docs/ai-sdk-core/generating-structured-data#structured-outputs-with-generatetext-and-streamtext).
+For more information about using structured outputs with `generateText` and `streamText` see [Generating Structured Data](/docs/ai-sdk-core/generating-structured-data#generating-structured-outputs).

package/docs/09-troubleshooting/15-stream-text-not-working.mdx

@@ -19,7 +19,7 @@ Errors become part of the stream and are not thrown to prevent e.g. servers from
 
 To log errors, you can provide an `onError` callback that is triggered when an error occurs.
 
-```tsx highlight="6-8"
+```tsx highlight="7-9"
 import { streamText } from 'ai';
 __PROVIDER_IMPORT__;
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai",
-  "version": "7.0.28",
+  "version": "7.0.29",
   "type": "module",
   "description": "AI SDK by Vercel - build apps like ChatGPT, Claude, Gemini, and more with a single interface for any model using the Vercel AI Gateway or go direct to OpenAI, Anthropic, Google, or any other model provider.",
   "license": "Apache-2.0",
@@ -42,7 +42,7 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/gateway": "4.0.20",
+    "@ai-sdk/gateway": "4.0.21",
     "@ai-sdk/provider": "4.0.3",
     "@ai-sdk/provider-utils": "5.0.10"
   },