felixarntz · Oct 23, 2026 · Oct 23, 2026 · Oct 23, 2026 · Oct 23, 2026 · Oct 23, 2026
diff --git a/.changeset/tall-houses-relax.md b/.changeset/tall-houses-relax.md
@@ -0,0 +1,84 @@
+---
+"@ai-sdk/harness-github-copilot": patch
+"@ai-sdk/harness-claude-code": patch
+"@ai-sdk/harness-deepagents": patch
+"@ai-sdk/harness-grok-build": patch
+"@ai-sdk/black-forest-labs": patch
+"@ai-sdk/openai-compatible": patch
+"@ai-sdk/sandbox-just-bash": patch
+"@ai-sdk/harness-opencode": patch
+"@ai-sdk/workflow-harness": patch
+"@ai-sdk/amazon-bedrock": patch
+"@ai-sdk/harness-cursor": patch
+"@ai-sdk/open-responses": patch
+"@ai-sdk/provider-utils": patch
+"@ai-sdk/sandbox-vercel": patch
+"@ai-sdk/tool-web-fetch": patch
+"@ai-sdk/anthropic-aws": patch
+"@ai-sdk/google-vertex": patch
+"@ai-sdk/harness-cline": patch
+"@ai-sdk/harness-codex": patch
+"@ai-sdk/harness-acp": patch
+"@ai-sdk/huggingface": patch
+"@ai-sdk/typesafe-ai": patch
+"@ai-sdk/assemblyai": patch
+"@ai-sdk/elevenlabs": patch
+"@ai-sdk/fish-audio": patch
+"@ai-sdk/harness-fx": patch
+"@ai-sdk/harness-pi": patch
+"@ai-sdk/llamaindex": patch
+"@ai-sdk/moonshotai": patch
+"@ai-sdk/perplexity": patch
+"@ai-sdk/policy-opa": patch
+"@ai-sdk/togetherai": patch
+"@ai-sdk/anthropic": patch
+"@ai-sdk/bytedance": patch
+"@ai-sdk/code-mode": patch
+"@ai-sdk/deepinfra": patch
+"@ai-sdk/fireworks": patch
+"@ai-sdk/langchain": patch
+"@ai-sdk/replicate": patch
+"@ai-sdk/tool-bash": patch
+"@ai-sdk/cartesia": patch
+"@ai-sdk/cerebras": patch
+"@ai-sdk/deepgram": patch
+"@ai-sdk/deepseek": patch
+"@ai-sdk/devtools": patch
+"@ai-sdk/gmicloud": patch
+"@ai-sdk/provider": patch
+"@ai-sdk/quiverai": patch
+"@ai-sdk/workflow": patch
+"@ai-sdk/alibaba": patch
+"@ai-sdk/baseten": patch
+"@ai-sdk/gateway": patch
+"@ai-sdk/harness": patch
+"@ai-sdk/klingai": patch
+"@ai-sdk/minimax": patch
+"@ai-sdk/mistral": patch
+"@ai-sdk/valibot": patch
+"@ai-sdk/cohere": patch
+"@ai-sdk/gladia": patch
+"@ai-sdk/google": patch
+"@ai-sdk/heygen": patch
+"@ai-sdk/openai": patch
+"@ai-sdk/prodia": patch
+"@ai-sdk/svelte": patch
+"@ai-sdk/voyage": patch
+"@ai-sdk/azure": patch
+"@ai-sdk/react": patch
+"@ai-sdk/revai": patch
+"@ai-sdk/topaz": patch
+"@ai-sdk/groq": patch
+"@ai-sdk/hume": patch
+"@ai-sdk/luma": patch
+"@ai-sdk/otel": patch
+"@ai-sdk/fal": patch
+"@ai-sdk/mcp": patch
+"@ai-sdk/rsc": patch
+"@ai-sdk/tui": patch
+"@ai-sdk/vue": patch
+"@ai-sdk/xai": patch
+"@ai-sdk/zai": patch
+---
+
+chore(docs): update `README.md` for consistent format and to avoid duplication
diff --git a/packages/alibaba/README.md b/packages/alibaba/README.md
@@ -6,10 +6,8 @@ The **[Alibaba provider](https://ai-sdk.dev/providers/ai-sdk-providers/alibaba)*
 
 ## Setup
 
-The Alibaba provider is available in the `@ai-sdk/alibaba` module. You can install it with
-
 ```bash
-npm i @ai-sdk/alibaba
+npm i ai @ai-sdk/alibaba
 ```
 
 ## Skill for Coding Agents
@@ -20,15 +18,7 @@ If you use coding agents such as Claude Code or Cursor, we highly recommend addi
 npx skills add vercel/ai
 ```
 
-## Provider Instance
-
-You can import the default provider instance `alibaba` from `@ai-sdk/alibaba`:
-
-```ts
-import { alibaba } from '@ai-sdk/alibaba';
-```
-
-## Language Model Example
+## Usage
 
 ```ts
 import { alibaba } from '@ai-sdk/alibaba';
@@ -40,166 +30,4 @@ const { text } = await generateText({
 });
 ```
 
-## Thinking Mode Example (Qwen Reasoning Models)
-
-Alibaba's Qwen models support thinking/reasoning mode for complex problem-solving:
-
-```ts
-import { alibaba } from '@ai-sdk/alibaba';
-import { generateText } from 'ai';
-
-const { text, reasoningText } = await generateText({
-  model: alibaba('qwen3-max'),
-  providerOptions: {
-    alibaba: {
-      enableThinking: true,
-      thinkingBudget: 2048,
-    },
-  },
-  prompt: 'How many "r"s are in the word "strawberry"?',
-});
-
-console.log('Reasoning:', reasoningText);
-console.log('Answer:', text);
-```
-
-## Preserved Thinking Example (Multi-Turn Reasoning)
-
-For models that support preserved thinking, the AI SDK sends reasoning from
-previous assistant messages back as Alibaba `reasoning_content` by default
-(`preserve_thinking`), so the model can build on its earlier thought process:
-
-```ts
-import { alibaba } from '@ai-sdk/alibaba';
-import { generateText } from 'ai';
-
-const providerOptions = {
-  alibaba: {
-    enableThinking: true,
-    thinkingBudget: 2048,
-  },
-};
-
-const opening = {
-  role: 'user' as const,
-  content: 'Is Kafka or RocketMQ a better fit for transactional messages?',
-};
-
-const first = await generateText({
-  model: alibaba('qwen3.7-max'),
-  messages: [opening],
-  providerOptions,
-});
-
-const second = await generateText({
-  model: alibaba('qwen3.7-max'),
-  messages: [
-    opening,
-    ...first.responseMessages, // append unchanged to keep the reasoning parts
-    { role: 'user', content: 'Which tradeoff mattered most?' },
-  ],
-  providerOptions,
-});
-```
-
-When continuing the conversation, append `responseMessages` unchanged so the
-reasoning parts survive to be serialized as `reasoning_content`. Set the
-`preserveThinking` provider option to `false` to opt out. Keep in mind:
-
-- `preserveThinking` does not enable thinking by itself.
-- It is enabled by default only for models that Alibaba documents as supporting
-  preserved thinking; for other models the option is not sent unless you set it
-  explicitly. See Alibaba's
-  [preserved-thinking documentation](https://docs.qwencloud.com/developer-guides/text-generation/thinking#preserve-thinking-in-multi-turn).
-- Reasoning from the current tool-call round is always sent back with tool
-  results, as Alibaba recommends.
-- Preserved reasoning increases input token usage and billing.
-- Historical reasoning remains separate from visible assistant text; it is never
-  merged into `content`.
-
-## Embedding Model Example
-
-```ts
-import { alibaba, type AlibabaEmbeddingModelOptions } from '@ai-sdk/alibaba';
-import { embed } from 'ai';
-
-const { embedding, usage } = await embed({
-  model: alibaba.embedding('text-embedding-v4'),
-  value: 'sunny day at the beach',
-  providerOptions: {
-    alibaba: {
-      textType: 'document',
-      dimension: 1024,
-      outputType: 'dense',
-    } satisfies AlibabaEmbeddingModelOptions,
-  },
-});
-```
-
-## Tool Calling Example
-
-```ts
-import { alibaba } from '@ai-sdk/alibaba';
-import { generateText, tool } from 'ai';
-import { z } from 'zod';
-
-const { text } = await generateText({
-  model: alibaba('qwen-plus'),
-  tools: {
-    weather: tool({
-      description: 'Get the weather in a location',
-      inputSchema: z.object({
-        location: z.string().describe('The location to get the weather for'),
-      }),
-      execute: async ({ location }) => ({
-        location,
-        temperature: 72 + Math.floor(Math.random() * 21) - 10,
-      }),
-    }),
-  },
-  prompt: 'What is the weather in San Francisco?',
-});
-```
-
-## Explicit Caching Example
-
-Alibaba supports both implicit and explicit prompt caching to reduce costs for repeated prompts.
-
-**Implicit caching** works automatically - the provider caches appropriate content without any configuration. For more control, you can use **explicit caching** by marking specific messages with `cacheControl`:
-
-```ts
-import { alibaba } from '@ai-sdk/alibaba';
-import { generateText } from 'ai';
-
-const longDocument = '... large document content ...';
-
-const { text, usage } = await generateText({
-  model: alibaba('qwen-plus'),
-  messages: [
-    {
-      role: 'user',
-      content: [
-        {
-          type: 'text',
-          text: 'Context: Please analyze this document.',
-        },
-        {
-          type: 'text',
-          text: longDocument,
-          providerOptions: {
-            alibaba: {
-              cacheControl: { type: 'ephemeral' },
-            },
-          },
-        },
-      ],
-    },
-  ],
-});
-```
-
-**Note:** The minimum content length for a cache block is 1,024 tokens.
-
-## Documentation
-
-Please check out the **[Alibaba provider documentation](https://ai-sdk.dev/providers/ai-sdk-providers/alibaba)** for more information.
+See the [Alibaba Provider documentation](https://ai-sdk.dev/providers/ai-sdk-providers/alibaba) for more information.
diff --git a/packages/amazon-bedrock/README.md b/packages/amazon-bedrock/README.md
@@ -7,10 +7,8 @@ contains language model support for the Amazon Bedrock [converse API](https://do
 
 ## Setup
 
-The Amazon Bedrock provider is available in the `@ai-sdk/amazon-bedrock` module. You can install it with
-
 ```bash
-npm i @ai-sdk/amazon-bedrock
+npm i ai @ai-sdk/amazon-bedrock
 ```
 
 ## Skill for Coding Agents
@@ -21,90 +19,7 @@ If you use coding agents such as Claude Code or Cursor, we highly recommend addi
 npx skills add vercel/ai
 ```
 
-## Provider Instance
-
-You can import the default provider instance `bedrock` from `@ai-sdk/amazon-bedrock`:
-
-```ts
-import { bedrock } from '@ai-sdk/amazon-bedrock';
-```
-
-## Authentication
-
-The Amazon Bedrock provider supports two authentication methods with automatic fallback:
-
-### API Key Authentication (Recommended)
-
-API key authentication provides a simpler setup process compared to traditional AWS SigV4 authentication. You can authenticate using either environment variables or direct configuration.
-
-#### Using Environment Variable
-
-Set the `AWS_BEARER_TOKEN_BEDROCK` environment variable with your API key:
-
-```bash
-export AWS_BEARER_TOKEN_BEDROCK=your-api-key-here
-```
-
-```ts
-import { bedrock } from '@ai-sdk/amazon-bedrock';
-import { generateText } from 'ai';
-
-const { text } = await generateText({
-  model: bedrock('us.anthropic.claude-haiku-5-5'),
-  prompt: 'Write a vegetarian lasagna recipe for 4 people.',
-  // API key is automatically loaded from AWS_BEARER_TOKEN_BEDROCK
-});
-```
-
-#### Using Direct Configuration
-
-You can also pass the API key directly in the provider configuration:
-
-```ts
-import { bedrock } from '@ai-sdk/amazon-bedrock';
-import { generateText } from 'ai';
-
-const bedrockWithApiKey = bedrock.withSettings({
-  apiKey: process.env.AWS_BEARER_TOKEN_BEDROCK, // or your API key directly
-  region: 'us-east-1', // Optional: specify region
-});
-
-const { text } = await generateText({
-  model: bedrockWithApiKey('us.anthropic.claude-haiku-5-5'),
-  prompt: 'Write a vegetarian lasagna recipe for 4 people.',
-});
-```
-
-### SigV4 Authentication (Fallback)
-
-If no API key is provided, the provider automatically falls back to AWS SigV4 authentication using standard AWS credentials:
-
-```ts
-import { bedrock } from '@ai-sdk/amazon-bedrock';
-import { generateText } from 'ai';
-
-// Uses AWS credentials from environment variables or AWS credential chain
-const { text } = await generateText({
-  model: bedrock('us.anthropic.claude-haiku-5-5'),
-  prompt: 'Write a vegetarian lasagna recipe for 4 people.',
-});
-```
-
-This method requires standard AWS environment variables:
-
-- `AWS_ACCESS_KEY_ID`
-- `AWS_SECRET_ACCESS_KEY`
-- `AWS_SESSION_TOKEN` (optional, for temporary credentials)
-
-### Authentication Precedence
-
-The provider uses the following authentication precedence:
-
-1. **API key from direct configuration** (`apiKey` in `withSettings()`)
-2. **API key from environment variable** (`AWS_BEARER_TOKEN_BEDROCK`)
-3. **SigV4 authentication** (AWS credential chain fallback)
-
-## Example
+## Usage
 
 ```ts
 import { bedrock } from '@ai-sdk/amazon-bedrock';
@@ -116,6 +31,4 @@ const { text } = await generateText({
 });
 ```
 
-## Documentation
-
-Please check out the **[Amazon Bedrock provider documentation](https://ai-sdk.dev/providers/ai-sdk-providers/amazon-bedrock)** for more information.
+See the [Amazon Bedrock Provider documentation](https://ai-sdk.dev/providers/ai-sdk-providers/amazon-bedrock) for more information.