feat: v6#6
Merged
callycodes merged 993 commits intomainfrom Jan 8, 2026
Merged
Conversation
…vercel#10999) Structured outputs section was missing example and link.
vercel#10975 broke submodule action. This should fix.
## Background This adds express mode support to Vertex which is a simplified version of authentication compared to the default OAuth flow. ## Summary Express mode uses a different base URL and requires the api key to be added as a search parameter. Since we need to bypass the default OAuth flow I also added checks to the ## Manual Verification Haven't been able to do this yet due to my Google account being old and messed up. Still posting this PR in the meantime.
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## @ai-sdk/google-vertex@4.0.0-beta.107 ### Patch Changes - 024e778: feat(provider/vertext): add express mode support Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…11007) ## Background xai supports file search through openai compatibility but this wasn't documented ## Summary - added file search tool section to xai docs - added example showing how to use openai provider with xai base url for file search ## Checklist - [ ] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [ ] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review)
…ercel#11005) ## Background Anthropic's web_fetch and web_search tools can return `null` for the title field when a page doesn't have a proper title or content retrieval fails. The zod type validation in the sdk was not accounting for this behavior, causing AI_TypeValidationError when Anthropic returned a null title. ## Summary - Updated title field to be nullable (z.string().nullable()) in: - web-fetch-20250910.ts - web-search_20250305.ts - Updated corresponding TypeScript types to title: string | null - Added unit tests for both schemas to verify null title handling ## Manual Verification Unit tests were added ## Checklist <!-- Do not edit this list. Leave items unchecked that don't apply. If you need to track subtasks, create a new "## Tasks" section Please check if the PR fulfills the following requirements: --> - [x] Tests have been added / updated (for bug fixes / features) - [ ] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) Fixes : vercel#10986 --------- Co-authored-by: Aayush Kapoor <aayushkapoor34@gmail.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## @ai-sdk/amazon-bedrock@4.0.0-beta.90 ### Patch Changes - Updated dependencies [05d5b9a] - @ai-sdk/anthropic@3.0.0-beta.80 ## @ai-sdk/anthropic@3.0.0-beta.80 ### Patch Changes - 05d5b9a: fix(anthropic): make title field nullable in web_fetch and web_search tool output schemas ## @ai-sdk/google-vertex@4.0.0-beta.108 ### Patch Changes - Updated dependencies [05d5b9a] - @ai-sdk/anthropic@3.0.0-beta.80 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…#11006) ## Background follow-up to vercel#10931 which adds express mode support to google vertex. this pr adds the corresponding documentation ## Summary - added express mode section to google vertex provider docs ## Checklist - [ ] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [ ] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Related Issues follow-up to vercel#10931
…es (vercel#11026) ## Background `AgentCallParameters` and `AgentStreamParameters` are needed for building custom `Agent` implementations. ## Summary Export `AgentCallParameters` and `AgentStreamParameters` types.
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.140 ### Patch Changes - 7fdd89d: feat(agent): export AgentCallParameters and AgentStreamParameters types ## @ai-sdk/angular@2.0.0-beta.140 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/langchain@2.0.0-beta.140 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/llamaindex@2.0.0-beta.140 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/react@3.0.0-beta.142 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/rsc@2.0.0-beta.142 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/svelte@4.0.0-beta.140 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 ## @ai-sdk/vue@3.0.0-beta.140 ### Patch Changes - Updated dependencies [7fdd89d] - ai@6.0.0-beta.140 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…cks (vercel#11029) ## Background Having context available in `onFinish` can help with tearing down resources. ## Summary Pass context into the `onFinish` handlers of `streamText` and `generateText`.
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.141 ### Patch Changes - b1405bf: feat(ai): send context into streamText / generateText onFinish callbacks ## @ai-sdk/angular@2.0.0-beta.141 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/langchain@2.0.0-beta.141 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/llamaindex@2.0.0-beta.141 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/react@3.0.0-beta.143 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/rsc@2.0.0-beta.143 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/svelte@4.0.0-beta.141 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 ## @ai-sdk/vue@3.0.0-beta.141 ### Patch Changes - Updated dependencies [b1405bf] - ai@6.0.0-beta.141 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
Add a community provider from MiniMax use Minimax-M2 --------- Co-authored-by: qinjiu <qinjiu@minimaxi.com> Co-authored-by: nicoalbanese <gcalbanese96@gmail.com>
…ercel#10171) ## Background `experimental_context` is only available in tool execution but not in the prepareStep callback. This prevents developers from using context data (like user session info or auth state) when making step-specific decisions in multi-step workflows. ## Summary Added experimental_context as an optional parameter to the prepareStep callback. ## Manual Verification Tested with both generateText and streamText. Co-authored-by: Gregor Martynus <39992+gr2m@users.noreply.github.com>
## Background Context needed to be supported in `ToolLoopAgent` `onFinish` as well. ## Summary Add `experimental_context` to `ToolLoopAgent.onFinish` callback definition. ## Related Issues Follow up from vercel#11029
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.142 ### Patch Changes - 7169511: feat(agent): support context in onFinish callback - bbdcb81: Add experimental_context parameter to prepareStep callback ## @ai-sdk/angular@2.0.0-beta.142 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/langchain@2.0.0-beta.142 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/llamaindex@2.0.0-beta.142 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/react@3.0.0-beta.144 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/rsc@2.0.0-beta.144 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/svelte@4.0.0-beta.142 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 ## @ai-sdk/vue@3.0.0-beta.142 ### Patch Changes - Updated dependencies [7169511] - Updated dependencies [bbdcb81] - ai@6.0.0-beta.142 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…#11040) ## Background Similar to modifying messages, model, etc. it would be useful to be able to modify the experimental context for each step. ## Summary * improve jsdoc of `PrepareStepFunction` and `PrepareStepResult` * allow modification of `experimental_context` in `prepareStep` ## Related Issues Builds on vercel#10171 Resolves vercel#10482
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.143 ### Patch Changes - 81e29ab: feat(ai): allow modifying experimental context in prepareStep - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/gateway@2.0.0-beta.75 ## @ai-sdk/amazon-bedrock@4.0.0-beta.91 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/anthropic@3.0.0-beta.81 ## @ai-sdk/angular@2.0.0-beta.143 ### Patch Changes - Updated dependencies [81e29ab] - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - ai@6.0.0-beta.143 ## @ai-sdk/anthropic@3.0.0-beta.81 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/assemblyai@2.0.0-beta.48 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/azure@3.0.0-beta.92 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai@3.0.0-beta.90 ## @ai-sdk/baseten@1.0.0-beta.49 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/black-forest-labs@1.0.0-beta.17 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/cerebras@2.0.0-beta.51 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/cohere@3.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/deepgram@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/deepinfra@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/deepseek@2.0.0-beta.49 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/elevenlabs@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/fal@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/fireworks@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/gateway@2.0.0-beta.75 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/gladia@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/google@3.0.0-beta.69 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/google-vertex@4.0.0-beta.109 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/anthropic@3.0.0-beta.81 - @ai-sdk/google@3.0.0-beta.69 ## @ai-sdk/groq@3.0.0-beta.49 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/huggingface@1.0.0-beta.43 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/hume@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/langchain@2.0.0-beta.143 ### Patch Changes - Updated dependencies [81e29ab] - ai@6.0.0-beta.143 ## @ai-sdk/llamaindex@2.0.0-beta.143 ### Patch Changes - Updated dependencies [81e29ab] - ai@6.0.0-beta.143 ## @ai-sdk/lmnt@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/luma@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/mcp@1.0.0-beta.32 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/mistral@3.0.0-beta.48 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/openai@3.0.0-beta.90 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/openai-compatible@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/perplexity@3.0.0-beta.48 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/provider-utils@4.0.0-beta.46 ### Patch Changes - 81e29ab: chore: update docs ## @ai-sdk/react@3.0.0-beta.145 ### Patch Changes - Updated dependencies [81e29ab] - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - ai@6.0.0-beta.143 ## @ai-sdk/replicate@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/revai@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/rsc@2.0.0-beta.145 ### Patch Changes - Updated dependencies [81e29ab] - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - ai@6.0.0-beta.143 ## @ai-sdk/svelte@4.0.0-beta.143 ### Patch Changes - Updated dependencies [81e29ab] - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - ai@6.0.0-beta.143 ## @ai-sdk/togetherai@2.0.0-beta.48 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/valibot@2.0.0-beta.46 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 ## @ai-sdk/vercel@2.0.0-beta.47 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 ## @ai-sdk/vue@3.0.0-beta.143 ### Patch Changes - Updated dependencies [81e29ab] - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - ai@6.0.0-beta.143 ## @ai-sdk/xai@3.0.0-beta.54 ### Patch Changes - Updated dependencies [81e29ab] - @ai-sdk/provider-utils@4.0.0-beta.46 - @ai-sdk/openai-compatible@2.0.0-beta.47 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
Following on from vercel#11002
…ncy range (vercel#11062) already addressed in v5 via vercel#10836
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## @ai-sdk/react@3.0.0-beta.146 ### Patch Changes - c388279: force update of vulnerable react versions through peer dependency range Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
## Background Improving ai gateway error message. ## Summary Improving ai gateway error message. ## Checklist - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Future Work Shortlink for v5 will be vercel.link/unauthenticated-ai-gateway. Once v6 stable is release, update links to v6: vercel.link/unauthenticated-ai-gateway v5: vercel.link/unauthenticated-ai-gateway-v5 and update the URLs behind them. --------- Co-authored-by: Gregor Martynus <39992+gr2m@users.noreply.github.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.144 ### Patch Changes - ab6f01a: Improve ai gateway error message when api key is not present ## @ai-sdk/angular@2.0.0-beta.144 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/langchain@2.0.0-beta.144 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/llamaindex@2.0.0-beta.144 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/react@3.0.0-beta.147 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/rsc@2.0.0-beta.146 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/svelte@4.0.0-beta.144 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 ## @ai-sdk/vue@3.0.0-beta.144 ### Patch Changes - Updated dependencies [ab6f01a] - ai@6.0.0-beta.144 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…11011) ## Background AI Gateway is adding support for users to pass BYOK with individual requests via provider options. ## Summary Added new provider option to pass BYOK credentials for providers as desired, and an example script to validate the new option. ## Manual Verification Ran new example with anthropic key against gateway service supporting the feature and observed successful response with provider metadata showing the request-scoped key was used. ## Checklist - [x] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review)
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ `main` is currently in **pre mode** so this branch has prereleases rather than normal releases. If you want to exit prereleases, run `changeset pre exit` on `main`.⚠️ ⚠️ ⚠️ ⚠️ ⚠️ ⚠️ # Releases ## ai@6.0.0-beta.145 ### Patch Changes - Updated dependencies [34ee8d0] - @ai-sdk/gateway@2.0.0-beta.76 ## @ai-sdk/angular@2.0.0-beta.145 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/gateway@2.0.0-beta.76 ### Patch Changes - 34ee8d0: feat (provider/gateway): add support for request-scoped byok ## @ai-sdk/langchain@2.0.0-beta.145 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/llamaindex@2.0.0-beta.145 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/react@3.0.0-beta.148 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/rsc@2.0.0-beta.147 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/svelte@4.0.0-beta.145 ### Patch Changes - ai@6.0.0-beta.145 ## @ai-sdk/vue@3.0.0-beta.145 ### Patch Changes - ai@6.0.0-beta.145 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
Update wording
Adds documentation for tool execution approval in AI SDK 6. Future: - Update human in the loop recipe - Add recipe for node
## Description This PR adds documentation for the [ai-sdk-provider-codex-cli](https://github.com/ben-vargas/ai-sdk-provider-codex-cli) community provider, which enables using OpenAI's GPT-5 series models through the Codex CLI. ## What does this PR do? - Adds documentation for the Codex CLI provider at `content/providers/03-community-providers/101-codex-cli.mdx` - Updates the community providers list in `content/docs/02-foundations/02-providers-and-models.mdx` ## Key Features The Codex CLI provider offers access to OpenAI's latest models: - **GPT-5 Model Families** - Support for gpt-5.2-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini, and legacy models - **Reasoning Configuration** - Multiple effort levels from 'none' to 'xhigh' - **Sandbox & Approval Modes** - Fine-grained control over tool execution - **Subscription Support** - Use ChatGPT Plus/Pro subscription or API key authentication - **Version Compatibility** - Works with AI SDK v6 and v5 ## Notes This provider is particularly useful for developers who have a ChatGPT Plus/Pro subscription and want to use OpenAI's GPT-5 models with the Vercel AI SDK without managing API keys separately. ## Testing - [x] Documentation follows the existing community provider format - [x] All installation methods (npm, pnpm, yarn, bun) are documented - [x] Model capabilities table accurately reflects the provider's features - [x] Links to the provider repository are working
## Description This PR updates documentation for the [ai-sdk-provider-gemini-cli](https://github.com/ben-vargas/ai-sdk-provider-gemini-cli) community provider to reflect AI SDK v6 compatibility and improvements. ## What does this PR do? - Adds version compatibility table showing v6, v5, and v4 support with correct NPM tags - Updates installation commands with correct version specifiers - Clarifies authentication options (6 auth types, with oauth-personal as optional default) - Moves verbose/logger settings to model settings section where they belong - Adds documentation for Gemini 3 model support (gemini-3-pro-preview, gemini-3-flash-preview) - Documents thinkingLevel configuration for Gemini 3 models ## Testing - [x] Documentation follows the existing community provider format - [x] All installation methods are documented with correct versions - [x] Model capabilities table accurately reflects the provider's features - [x] Authentication methods are clearly documented with examples - [x] Links to the provider repository are working
…#11458) ## Background Anthropic introduced a breaking change in their API requiring that sampling parameters use only `temperature` OR `top_p`, not both. This aligns with their Claude 4 migration guidelines. Previously, the AI SDK Anthropic provider would send both parameters to the API when both were specified, which would result in an API error with the new Anthropic API requirements. See: https://platform.claude.com/docs/en/about-claude/models/migrating-to-claude-4 ## Summary This PR implements mutual exclusivity between `temperature` and `topP` parameters in the Anthropic provider: **Changes:** - Added validation logic that runs when thinking mode is NOT enabled (thinking mode has its own parameter validation) - When both `temperature` and `topP` are provided: - `temperature` takes priority - `topP` is set to `undefined` and excluded from the API request - A warning is added: "topP is not supported when temperature is set. topP is ignored." - Updated existing tests to reflect the new behavior - Added comprehensive test suite for temperature/topP mutual exclusivity **Files Modified:** - `packages/anthropic/src/anthropic-messages-language-model.ts` (lines 436-446) - `packages/anthropic/src/anthropic-messages-language-model.test.ts` ## Manual Verification 1. **Test Suite Verification**: All 235 tests pass, including: - Existing tests updated to remove conflicting parameters - New tests specifically for temperature/topP mutual exclusivity scenarios - Tests verify that thinking mode validation takes precedence (warnings appear in correct order) 2. **Expected Behavior**: - When only `temperature` is provided → sends temperature to API ✅ - When only `topP` is provided → sends topP to API ✅ - When both are provided → sends only temperature, warns about topP being ignored ✅ - When thinking mode is enabled with both → thinking validation handles all parameters ✅ ## Checklist - [x] Tests have been added / updated (for bug fixes / features) - [ ] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Related Issues Resolves the Anthropic API breaking change for Claude 4 parameter requirements.
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## ai@6.0.10 ### Patch Changes - ae26f95: Add missing `.catch()` handler to `executeToolCall` promise in `runToolsTransformation` to prevent potential stream hang when the promise rejects. ## @ai-sdk/amazon-bedrock@4.0.8 ### Patch Changes - Updated dependencies [2231e84] - @ai-sdk/anthropic@3.0.6 ## @ai-sdk/angular@2.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/anthropic@3.0.6 ### Patch Changes - 2231e84: fix(anthropic): implement temperature/topP mutual exclusivity Resolves the Anthropic API breaking change where sampling parameters must use only `temperature` OR `top_p`, not both. When both parameters are provided: - Temperature takes priority and topP is ignored - A warning is added to inform users: "topP is not supported when temperature is set. topP is ignored." - The validation only runs when thinking mode is not enabled (thinking mode has its own parameter validation) See Anthropic migration guide: <https://platform.claude.com/docs/en/about-claude/models/migrating-to-claude-4> ## @ai-sdk/google-vertex@4.0.6 ### Patch Changes - Updated dependencies [2231e84] - @ai-sdk/anthropic@3.0.6 ## @ai-sdk/langchain@2.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/llamaindex@2.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/mistral@3.0.4 ### Patch Changes - 45631ab: Update refrence_ids type to be union of number and string ## @ai-sdk/react@3.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/rsc@2.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/svelte@4.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/vue@3.0.10 ### Patch Changes - Updated dependencies [ae26f95] - ai@6.0.10 ## @ai-sdk/xai@3.0.5 ### Patch Changes - e7f0a6f: fix(xai): add missing stream response schema and provider tool name parsing Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
## Background When using `@ai-sdk/langchain` with LangGraph agents that have tools, multi-turn conversations fail with: ``` 400 An assistant message with 'tool_calls' must be followed by tool messages responding to each 'tool_call_id'. The following tool_call_ids did not have response messages: call_xxxxx ``` The root cause is in `toUIMessageStream`'s handling of LangGraph `values` events. When a `values` event contains the full message history (including AIMessages with `tool_calls` from previous turns), the adapter emits `tool-input-start` and `tool-input-available` events for ALL historical tool calls, but does NOT emit corresponding `tool-output-available` events for them. This creates orphaned tool parts in the UIMessage with state `input-available` (no output). When `toBaseMessages` converts these back for the next request, it creates AIMessages with `tool_calls` but no following ToolMessages, causing the OpenAI API error. ## Summary Added a two-pass approach when processing `values` events in `processLangGraphEvent`: 1. **First pass**: Collect all tool call IDs that have been responded to by ToolMessages in the message history. These are "completed" historical tool calls. 2. **Second pass**: When processing AIMessages with tool calls, skip emitting events for any tool call that exists in the completed set. This ensures that: - Historical tool calls that already have ToolMessage responses are NOT re-emitted - Only new/current tool calls (those without ToolMessage responses yet) are emitted - The client's message history remains consistent without orphaned tool parts ## Manual Verification I double checked this resolution using the provided reproduction case and was able to validate that all is working fine: <img width="1624" height="1056" alt="Screenshot 2025-12-23 at 5 25 32 PM" src="https://github.com/user-attachments/assets/94f08b19-2cb6-4bb7-8b46-8c3b2a3863d0" /> ## Checklist - [x] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Future Work n/a ## Related Issues Fixes vercel#11415
…l#11536) ## Background xAI's Chat API returns `completion_tokens` that include both text and reasoning tokens, matching OpenAI's o1 format. The SDK was treating this value as text-only, causing incorrect calculations that resulted in negative text token counts when reasoning tokens exceeded the misinterpreted value. ## Summary **Updated token calculation in `convert-xai-chat-usage.ts`:** - `outputTokens.total`: `completion_tokens + reasoningTokens` → `completion_tokens` - `outputTokens.text`: `completion_tokens` → `completion_tokens - reasoningTokens` **Example:** ```typescript // xAI API response { completion_tokens: 95, // includes both text (6) and reasoning (89) completion_tokens_details: { reasoning_tokens: 89 } } // Before: text = 6 - 89 = -83 ❌ // After: text = 95 - 89 = 6 ✅ ``` **Updated test cases:** - Updated all tests in `convert-xai-chat-usage.test.ts` to reflect correct xAI API format where `completion_tokens` includes reasoning tokens - Fixed 3 inline snapshots in `xai-chat-language-model.test.ts` for reasoning model tests to match the corrected token calculation **Added changeset** for the `@ai-sdk/xai` package to ensure proper versioning. ## Manual Verification Verified with reported scenario: - Input: `completion_tokens: 95`, `reasoning_tokens: 89` - Output: `text: 6`, `reasoning: 89`, `total: 95` (no negative values) - All 126 tests passing (both node and edge environments) ## Checklist - [x] Tests have been added / updated (for bug fixes / features) - [ ] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Related Issues - Fixes vercel#11487 --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: Lars Grammel <lars.grammel@gmail.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## @ai-sdk/langchain@2.0.11 ### Patch Changes - 61ff058: fix(langchain): do not re-emit executed tool calls Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## @ai-sdk/xai@3.0.6 ### Patch Changes - 77012ef: fix(xai): correct token counting for reasoning models to prevent negative text_output_tokens Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## ai@6.0.11 ### Patch Changes - Updated dependencies [8ec1984] - @ai-sdk/gateway@3.0.8 ## @ai-sdk/angular@2.0.11 ### Patch Changes - ai@6.0.11 ## @ai-sdk/gateway@3.0.8 ### Patch Changes - 8ec1984: fix(gateway): bump `@vercel/oidc` to 3.1.0 ## @ai-sdk/langchain@2.0.12 ### Patch Changes - ai@6.0.11 ## @ai-sdk/llamaindex@2.0.11 ### Patch Changes - ai@6.0.11 ## @ai-sdk/react@3.0.11 ### Patch Changes - ai@6.0.11 ## @ai-sdk/rsc@2.0.11 ### Patch Changes - ai@6.0.11 ## @ai-sdk/svelte@4.0.11 ### Patch Changes - ai@6.0.11 ## @ai-sdk/vue@3.0.11 ### Patch Changes - ai@6.0.11 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
<!-- Welcome to contributing to AI SDK! We're excited to see your changes. We suggest you read the following contributing guide we've created before submitting: https://github.com/vercel/ai/blob/main/CONTRIBUTING.md --> ## Background Prodia doesn't have an OpenAI-compatible API, so they need a provider package ## Summary Created it ## Manual Verification Unit tests and example scripts ## Checklist <!-- Do not edit this list. Leave items unchecked that don't apply. If you need to track subtasks, create a new "## Tasks" section Please check if the PR fulfills the following requirements: --> - [x] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [X] I have reviewed this pull request (self-review) ## Future Work <!-- Feel free to mention things not covered by this PR that can be done in future PRs. Remove the section if it's not needed. --> ## Related Issues <!-- List related issues here, e.g. "Fixes vercel#1234". Remove the section if it's not needed. --> --------- Co-authored-by: Gregor Martynus <39992+gr2m@users.noreply.github.com> Co-authored-by: vercel[bot] <35613825+vercel[bot]@users.noreply.github.com>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## @ai-sdk/prodia@1.0.0 ### Major Changes - bb3d30e: feat(provider/prodia): Create Prodia provider package Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…#11549) ## Summary - Remove `gateway()` wrapper from all tool code examples in the tools registry - Update imports to remove `gateway` from `'ai'` package imports - Update ctx-zip and perplexity-search examples to use `openai/gpt-5.2` ## Test plan - [ ] Verify code examples render correctly on the tools registry page 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
## Background Attempting to use function tools with the `xai` provider `responses` api fails with invalid json parse error on xai server side. Sample error: ``` Failed to deserialize the JSON body into the target type: tools[0]: missing field `name` at line 1 column 496 ``` ## Summary Revised format of function calls to use a flat format matching existing openai responses api implementation. The official xai docs do not seem to define the format to this level of detail: https://docs.x.ai/docs/api-reference#create-new-response ## Manual Verification Added example script to explicitly test tool calls with the responses API. Fails before the change and passes after.
## Summary add JSON parsing note to AGENTS.md
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## @ai-sdk/xai@3.0.7 ### Patch Changes - 23179aa: fix (provider/xai): use correct format for function tools Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
## Background The ability to run middleware on ImageModels to support semantic caching or safety guardrails without requiring adhoc implementations is desirable. Extend wrapProvider to accept newly created ImageModelMiddleware for consistent implementation as LanguageModelMiddleware ## Summary - Added ImageModelV3Middleware to @ai-sdk/provider (packages/provider/src/image-model-middleware/v3/image-model-v3-middleware.ts) - Added wrapImageModel to ai (packages/ai/src/middleware/wrap-image-model.ts) and ImageModelMiddleware alias (packages/ai/src/types/image-model-middleware.ts). - Extended middleware plumbing to support images in wrapProvider (packages/ai/src/middleware/wrap-provider.ts) and createProviderRegistry (packages/ai/src/registry/provider-registry.ts). - Tests: new wrapImageModel suite (packages/ai/src/middleware/wrap-image-model.test.ts) + updated registry/provider wrapper tests. - Docs: added Image Middleware section (content/docs/03-ai-sdk-core/35-image-generation.mdx), new reference page (content/docs/07-reference/01-ai-sdk-core/61-wrap-image-model.mdx) ## Manual Verification Manually built and implemented migrated our existing middleware approach to use new middleware implementation ## Checklist - [x] Tests have been added / updated (for bug fixes / features) - [x] Documentation has been added / updated (for bug fixes / features) - [x] A _patch_ changeset for relevant packages has been added (for bug fixes / features - run `pnpm changeset` in the project root) - [x] I have reviewed this pull request (self-review) ## Future Work Create a `defaultImageSettingsMiddleware` to be used for reference implementation. Extend wrapProvider to accept wrapping embedding models too ## Related Issues Resolves vercel#11546
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## ai@6.0.12 ### Patch Changes - d937c8f: Add Image model middleware support via `wrapImageModel` and `ImageModelV3Middleware`. - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/gateway@3.0.9 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/amazon-bedrock@4.0.9 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/anthropic@3.0.7 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/angular@2.0.12 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/anthropic@3.0.7 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/assemblyai@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/azure@3.0.5 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai@3.0.5 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/baseten@1.0.5 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/black-forest-labs@1.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/cerebras@2.0.5 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/cohere@3.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/deepgram@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/deepinfra@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/deepseek@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/devtools@0.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 ## @ai-sdk/elevenlabs@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/fal@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/fireworks@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/gateway@3.0.9 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/gladia@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/google@3.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/google-vertex@4.0.7 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/anthropic@3.0.7 - @ai-sdk/google@3.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/groq@3.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/huggingface@1.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/hume@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/langchain@2.0.13 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 ## @ai-sdk/llamaindex@2.0.12 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 ## @ai-sdk/lmnt@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/luma@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/mcp@1.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/mistral@3.0.5 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/openai@3.0.5 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/openai-compatible@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/perplexity@3.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/prodia@1.0.1 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/provider@3.0.2 ### Patch Changes - d937c8f: Add Image model middleware support via `wrapImageModel` and `ImageModelV3Middleware`. ## @ai-sdk/provider-utils@4.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 ## @ai-sdk/react@3.0.12 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/replicate@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/revai@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/rsc@2.0.12 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - ai@6.0.12 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/svelte@4.0.12 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/togetherai@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/valibot@2.0.4 ### Patch Changes - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/vercel@2.0.4 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/vue@3.0.12 ### Patch Changes - Updated dependencies [d937c8f] - ai@6.0.12 - @ai-sdk/provider-utils@4.0.4 ## @ai-sdk/xai@3.0.8 ### Patch Changes - Updated dependencies [d937c8f] - @ai-sdk/provider@3.0.2 - @ai-sdk/openai-compatible@2.0.4 - @ai-sdk/provider-utils@4.0.4 Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
…h tool (vercel#10765) ## Background streamText was throwing a validation error when using the x_search tool because some event types returned by xAI were missing from the `xaiResponsesChunkSchema` ## Summary I updated the chunkSchema with the missing types and also added a test in `packages/xai/src/responses/xai-responses-language-model.test.ts` ## Manual Verification I tested it by running the xai-x-search-video-understanding example ## Related Issues fixes vercel#10628 Co-authored-by: josh <josh@afterima.ge>
This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated. # Releases ## @ai-sdk/xai@3.0.9 ### Patch Changes - 173dcfd: added support for streaming custom tool input chunks in xAI Co-authored-by: vercel-ai-sdk[bot] <225926702+vercel-ai-sdk[bot]@users.noreply.github.com>
- Merged 987 commits from vercel/ai upstream - Preserved @zenning/* package renames (ai, openai, provider, react) - Re-applied multi-message stream support - Re-applied file-search optional query documentation fix
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Background
Summary
Manual Verification
Checklist
pnpm changesetin the project root)pnpm prettier-fixin the project root)Future Work
Related Issues