NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
npm · #2684 most downloaded on npm
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.
Last release today
06 Oct 2026
Ships on a steady schedule
a new release about every 8 days
Some releases are documented
notes for 27 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
13 years old
1572 releases · first in 2014
ede5b89 : chore: migrate package builds from tsup to tsdown
c2511c1 : fix: use standards-compliant User-Agent header
One column per quarter.
Updated dependencies [ e3605f6 ]
experimental_streamData: true setting from AIStreamCallbacksAndOptions. You can delete occurrences in your code. The stream data protocol is now used by default.Updated dependencies [ af9597b ]
63d587e: Add Anthropic provider for ai/core functions (no tool calling).
Updated dependencies [ fe07867 ]
Updated dependencies [ ffb0e76 ]
Updated dependencies [ 2973485 ]
ai/spec.Updated dependencies [ 0455398 ]
StreamableValueUpdated dependencies [ 215b25e ]
Updated dependencies [ 11b4c2d ]
Updated dependencies [ 5c0054d ]
33d261a: fix(ai/rsc): Fix .append() behavior
Updated dependencies [ 9942196 ]
Updated dependencies [ 912fb01 ]
Updated dependencies [ a4ba394 ]
Updated dependencies [ e5a22f0 ]
Updated dependencies [ 6bcc0f8 ]
7962862: fix useActions type inference
useActions type inferenceUpdated dependencies [ d3bcad9 ]
Updated dependencies [ 90192f1 ]
Updated dependencies [ 3e125ba ]
### Major Changes - 51054a9: add ai/rsc
a6b5764: Add support for Mistral's JavaScript SDK
141f0ce: Fix: onFinal callback is invoked with text from onToolCall when onToolCall returns string
b717dad: Adding Inkeep as a stream provider
2c8ffdb: cohere-stream: support AsyncIterable
8542ae7: react/use-assistant: add onError handler
7851fa0: StreamData: add annotations and appendMessageAnnotation support
annotations and appendMessageAnnotation support9b89c4d: react/use-assistant: Expose setInput
ac503e0: ai/solid: add chat request options to useChat
5a9ae2e: ai/prompt: add experimental_buildOpenAIMessages to validate and cast AI SDK messages to OpenAI messages
experimental_buildOpenAIMessages to validate and cast AI SDK messages to OpenAI messages07a679c: Add data message support to useAssistant & assistantResponse.
api functions are no longer used as a cache key in useChat0fd1205: ai/vue: Add complex response parsing and StreamData support to useCompletion
df1ad33: ai/vue: Add complex response parsing and StreamData support to useChat
generateId to use-chat params to allow overriding message ID generation1e61c69: chore: specify the minimum react version to 18
69ca8f5: ai/react: add experimental_useAssistant hook and experimental_AssistantResponse
5a04321: add StreamData support to StreamingReactResponse, add client-side data API to react/use-chat
4529831: ai/react: Do not store initialMessages in useState
2c8d4bd: Support openai@4.16.0 and later
424d5ee: experimental_StreamData: fix trailing newline parsing bug in decoder
699552d: add experimental_StreamingReactResponse
0bd27f6: react/use-chat: allow client-side handling of function call without following response
5ed581d: Use interface instead of type for Message to allow declaration merging
function_call and name fields in subsequent requestse569688: Fix for #637, resync interfaces
c5d1857: fix: return complete response in onFinish when onCompletion isn't passed
6229d6b: openai: fix OpenAIStream types with openai@4.11+
a4a997f: all providers: reset error message on (re)submission
cb181b4: ai/vue: wrap body with unref to support reactivity
2470658: ai/react: fix: handle partial chunks in react getStreamedResponse when using experimental_StreamData
8a2cbaf: vue/use-completion: fix: don't send network request for loading state"
writer from LangChainStream3fc2b32: ai/vue: fix: make body parameter reactive
26bf998: ai/react: make reload/complete/append functions stable via useCallback
2f97630: react/use-chat: fix aborting clientside function calls too early
e5bf68d: react/use-chat: fix experimental functions returning proper function messages
e5bf68d: react/use-chat: fix experimental functions returning proper function messages
Closes #478
7b389a7: fix: improve safety for type check in openai-stream
867a3f9: Fix client-side function calling (#467, #469)
867a3f9: Fix client-side function calling (#467, #469)
add Completion type from the openai SDK to openai-stream (#472)
84e0cc8: Add experimental_StreamData and new opt-in wire protocol to enable streaming additional data. See https://github.com/vercel/ai/pull/425.
84e0cc8: Add experimental_StreamData and new opt-in wire protocol to enable streaming additional data. See https://github.com/vercel/ai/pull/425.
Changes onCompletion back to run every completion, including recursive function calls. Adds an onFinish callback that runs once everything has streamed.
If you're using experimental function handlers on the server and caching via onCompletion,
you may want to adjust your caching code to account for recursive calls so the same key isn't used.
let depth = 0
const stream = OpenAIStream(response, {
async onCompletion(completion) {
depth++
await kv.set(key + '_' + depth, completion)
await kv.expire(key + '_' + depth, 60 * 60)
}
})
04084a8: openai-stream: fix experimental_onFunctionCall types for OpenAI SDK v4
dca1ed9: Update packages and examples to use OpenAI SDK v4
Your coding agent can read these notes before it upgrades. Set up the MCP server →