wehub/assistant-ui--assistant-ui
1
0
Fork 0
Code Issues Pull Requests Actions Deployments Models Agent Notes Packages Projects Releases Wiki Activity
Files
main
assistant-ui--assistant-ui/packages/react-ai-sdk/CHANGELOG.md
T
wehub-resource-sync e30e75b5d4
Changesets / Create Version PR (push) Has been cancelled
Details
Deploy Shadcn Registry / Deploy Production (push) Has been cancelled
Details
Template Metrics / LOC + Bundle Size (push) Has been cancelled
Details
Code Quality / Oxlint + Oxfmt (push) Has been cancelled
Details
Code Quality / Template Sync (push) Has been cancelled
Details
Code Quality / Build Changed Packages (push) Has been cancelled
Details
Code Quality / Test Changed Packages (push) Has been cancelled
Details
Deploy Expo Example / Deploy Production (push) Has been cancelled
Details
Deploy Ink Example / Deploy Production (push) Has been cancelled
Details
Python Tests / pytest (assistant-stream, 3.10) (push) Has been cancelled
Details
Python Tests / pytest (assistant-stream, 3.12) (push) Has been cancelled
Details
Python Tests / pytest (assistant-ui-sync-server-api, 3.10) (push) Has been cancelled
Details
Python Tests / pytest (assistant-ui-sync-server-api, 3.12) (push) Has been cancelled
Details
chore: import upstream snapshot with attribution
2026-07-13 13:40:13 +08:00

56 KiB
Raw Permalink Blame History

@assistant-ui/react-ai-sdk

1.3.40

Patch Changes

1.3.39

Patch Changes

1.3.38

Patch Changes

  • #4551 49175f3 - refactor: delegate useStreamingTiming to the shared core primitive, with no public API change (@okisdev)

  • #4517 cefcf27 - chore: update dependencies (@okisdev)

  • #4515 f5e94b7 - feat: forward onThreadIdChange through the adapter entry hooks (useLangGraphRuntime, useStreamRuntime, useChatRuntime, useAdkRuntime, useOpenCodeRuntime, usePiRuntime). the option already existed on useRemoteThreadListRuntime but every wrapper dropped it, so routing/persistence built on the settled remote thread id never fired from these hooks. only the settled remote id is emitted; the transient __LOCALID_* placeholder is never surfaced. (@okisdev)

  • #4310 0c51b90 - feat: add unstable_injectInteractableContext for AI SDK route handlers (@AVGVSTVS96)

    When using AI SDK's convertToModelMessages, call unstable_injectInteractableContext(messages) first so the model can read current interactable state:

    import { convertToModelMessages } from "ai";
+import { unstable_injectInteractableContext } from "@assistant-ui/react-ai-sdk";

- messages: await convertToModelMessages(messages),
+ messages: await convertToModelMessages(unstable_injectInteractableContext(messages)),

    Existing interactables apps using AI SDK need this call after upgrading.

  • Updated dependencies [ddc40b7, ea52de0, 29c6fdb, d0987a3, cefcf27, 0c51b90, 3a8f685, ec6adf4, 4acd4c0]:

    • @assistant-ui/core@0.2.19
    • assistant-stream@0.3.24
    • assistant-cloud@0.1.34
    • @assistant-ui/store@0.2.19

1.3.37

Patch Changes

1.3.36

Patch Changes

1.3.35

Patch Changes

1.3.34

Patch Changes

  • #4320 0e923b4 - fix: declare assistant-stream as a regular dependency so it is externalized instead of bundled into dist (the bundled copy broke consumers on its transitive secure-json-parse import) (@Yonom)

  • #4315 60ef0e9 - feat: add runtime support for deleting messages (@Yonom)

  • Updated dependencies [60ef0e9, 1b6a0d6, 1b6a0d6, 1b6a0d6]:

    • @assistant-ui/core@0.2.13
    • @assistant-ui/store@0.2.15

1.3.33

Patch Changes

  • #4271 2a84174 - feat: expose joinStrategy on useAISDKRuntime / useChatRuntime (@okisdev)

    the new AI SDK runtime always merged consecutive role: "assistant" UIMessages into a single rendered turn, with no supported way to opt out (the converter accepts joinStrategy but the runtime never forwarded it, and AISDKMessageConverter is not exported). this follows up on #1633, where the same knob shipped on the legacy useVercelUseChatRuntime as unstable_joinStrategy. pass joinStrategy: "none" to keep proactive or history loaded consecutive assistant messages as separate turns.

    core now exports a shared JoinStrategy type so the "concat-content" | "none" union has a single source of truth across the converter and the runtimes.

  • #4266 906fae9 - chore: deprecate generativeTools in favor of AISDKToolkit (@Yonom)

    generativeTools({ toolkit, frontendTools }) is deprecated. Use new AISDKToolkit({ toolkit }).tools({ frontend }) instead: it is a strict superset that also opens MCP server connections, so it replaces generativeTools for every toolkit. The frontendTools option is named frontend on .tools(), and .tools() is async. generativeTools keeps working and will be removed in a future version.

  • #4306 15878d8 - chore: update dependencies (@Yonom)

  • #4285 d4629c4 - fix(react-ai-sdk): make the package metro-bundleable in react native / expo by serving a client-only entry under the react-native export condition (@okisdev)

    importing @assistant-ui/react-ai-sdk in an expo / react native app failed metro bundling because the root entry re-exports the server-only generativeTools / AISDKToolkit, which pull in @ai-sdk/mcp/mcp-stdio and node's child_process. react native now resolves a client-only entry that omits the server toolkit, so metro never reaches the node-only code. the node / server entry is unchanged and still exports the full api.

  • #4304 a18bdad - fix(react-ai-sdk): make the package root bundle on edge, cloudflare workers, deno, and browsers, not just react native. the node-only stdio MCP transport is now resolved through a package.json "imports" condition (#mcp-stdio), so importing @assistant-ui/react-ai-sdk at root carries the full server toolkit on every target with http/sse MCP working on edge, while stdio MCP throws a clear runtime error only where a subprocess cannot be spawned (browser, react native, edge, workers) instead of breaking the build via @ai-sdk/mcp/mcp-stdio -> node:child_process. node, bun, and deno keep the real stdio transport. (@okisdev)

  • Updated dependencies [2a84174, a0a0769, 19c5b5f, dbdfb15, ca191dc, 15878d8, 44ff4bf, 01cf957, 26a365b]:

    • @assistant-ui/core@0.2.11
    • assistant-cloud@0.1.32
    • @assistant-ui/store@0.2.14

1.3.32

Patch Changes

  • #4196 b3655bc - fix: strip stale approval object when cancelling a tool pending approval so the next request passes validateUIMessages (@ShobhitPatra)

  • #4192 4f2e077 - fix: support data message parts in toCreateMessage instead of throwing (@ShobhitPatra)

  • #4216 d440594 - fix: convert generative tool toModelOutput to AI SDK content shape (@AVGVSTVS96)

  • #4212 5fe118d - feat: add MCP server support to generative toolkits (@Yonom)

  • #4213 dcd5897 - feat: add provider-executed tool support to generative toolkits (@Yonom)

  • #4199 d9b3119 - feat: a defineToolkit entry may now be an already-formed ToolDefinition (carrying its own type), not only an inline definition whose type the compiler infers. This is what lets a factory like new JSONGenerativeUI({ library }).present() be used directly as a tool. (@Yonom)

    Renames the authoring types to match defineToolkit: ToolkitDeclaration → ToolkitDefinition, and adds ToolkitDefinitionEntry (the union of an inline tool definition and a pre-formed ToolDefinition). The per-tool inline type is now an internal ToolkitDefinitionInput and is no longer exported.

  • Updated dependencies [4145caa, 78ff336, 5fe118d, dcd5897, 0558db2, 69540af, d9b3119, ae54c55, 7640b31]:

    • @assistant-ui/core@0.2.10
    • assistant-cloud@0.1.31

1.3.31

Patch Changes

  • #4172 1315789 - feat: add generativeTools({ toolkit, frontendTools }) — builds an AI SDK ToolSet from an assistant-ui toolkit (server-side execute) merged with the frontend-uploaded tools, for use in streamText/generateText. Pairs with the "use generative" compiler so a backend tool's execute runs on the server while its render ships to the client. (@Yonom)

  • #4151 299d448 - chore: drop stale biome-ignore pragmas now that the repo lints with oxlint (@okisdev)

  • #4169 8518a29 - fix: don't abort the whole history load when a stored row fails to convert. toExportedMessageRepository now drops a row whose content can't be decoded to a valid message (e.g. a hand-seeded { "foo": "bar" } with no role), along with any descendants that referenced it, and discards a headId that points at a dropped row. previously a single bad row emitted an undefined message (or left a dangling parentId / headId) that threw Cannot read properties of undefined (reading 'id') or Parent message not found during thread.import, taking down the entire thread. the rest of the thread now loads, matching how the cloud adapter filters out rows it can't decode rather than throwing. (@okisdev)

  • #4136 4429aa3 - centralize thread-level shared options forwarding across runtime wrapper hooks. follow-up to #4135. (@okisdev)

    new public exports from @assistant-ui/core (re-exported from @assistant-ui/react):

    • ExternalStoreSharedOptions, a typed Pick over ExternalStoreAdapter covering the four thread-level optional fields every wrapper forwards: isDisabled, isSendDisabled, unstable_capabilities, suggestions.
    • pickExternalStoreSharedOptions(options), plucks those four fields from a wider options object. the body uses satisfies Required<...> so adding a key to the type without copying it in the function is a compile error rather than a silent missing-field bug.
    • useExternalStoreSharedOptions(options) (from @assistant-ui/core/react), a memoized variant for wrappers that wrap their store in useMemo. lets the wrapper list a single stable shared reference as a dep instead of enumerating the four fields. same satisfies guard internally so the destructure stays in sync with the type.

    internal: every runtime wrapper hook (useChatRuntime, useAISDKRuntime, useLangGraphRuntime, useA2ARuntime, useAgUiRuntime, useAdkRuntime, useStreamRuntime, useOpenCodeRuntime) now uses these helpers instead of inlining the conditional spreads added in #4135. each wrapper sheds 20 to 40 lines of duplicated declarations and conditional spreads; future additions to the shared option set propagate through a single edit in pickExternalStoreSharedOptions instead of touching every wrapper. no user-facing behavior change.

  • #4162 eef724e - fix: drop phantom sibling messages when an external store swaps an optimistic message id mid-run (#4037). (@Yonom)

    Messages can now be flagged metadata.isOptimistic. Optimistic messages are treated as ephemeral: they only ever live on the current head branch (the repository evicts off-branch optimistic messages whenever the head moves) and they are never written to persisted state (export() omits them). The AI SDK v6 adapter flags the streaming assistant message as optimistic, so when its client-generated id is replaced by a server-provided one mid-run, the stale placeholder no longer lingers as a phantom branch (e.g. BranchPicker showing 2/2 on a turn the user never branched). Unlike the reverted blanket id-diff (#4040), only explicitly-optimistic messages are affected, so legitimate onEdit / onReload / switchToBranch branches are preserved.

  • #4175 2dec3ae - chore: update dependencies (@Yonom)

  • #4135 e7c2396 - align the runtime wrapper hooks so every distribution forwards the same set of optional adapter-level fields to useExternalStoreRuntime. closes #4134. (@okisdev)

    useChatRuntime and useAISDKRuntime (which already accepted suggestions) gain three new options:

    • isDisabled, disables the composer input entirely.
    • isSendDisabled, keeps the input usable but makes send() a no-op (paired with composer.canSend).
    • unstable_capabilities, per-thread capability overrides (currently { copy?: boolean }).

    useLangGraphRuntime, useA2ARuntime, useAgUiRuntime, useAdkRuntime, useStreamRuntime, useOpenCodeRuntime gain all four (the three above plus suggestions).

    adapter-level additions, where missing:

    • useChatRuntime / useAISDKRuntime already accepted dictation and voice through the ExternalStoreAdapter adapter shape; this just confirms the typing.
    • useLangGraphRuntime, useA2ARuntime, useAgUiRuntime, useAdkRuntime, useStreamRuntime, useOpenCodeRuntime now accept dictation and voice in their adapters object and forward them through.
    • useOpenCodeRuntime gains an adapters option for the first time (attachments / speech / dictation / voice / feedback).

    every new field is optional and defaults to the prior behavior, so existing call sites need no changes.

  • #4133 c4d3eea - forward per-tool providerOptions from useAssistantTool through toToolsJSONSchema and frontendTools into the AI SDK request body, and emit tool entries in alphabetical order so identical tool sets produce byte-identical request bodies for stable prompt caching. react-ag-ui inherits the sort via toAgUiTools, so identical tool sets reach the AG-UI runtime in a stable order regardless of mount order. (@okisdev)

    this lets you opt into provider-specific tool features (e.g. Anthropic's defer_loading, Anthropic Tool Search Tool) without any provider-aware code in assistant-ui:

    useAssistantTool({
  toolName: "get_weather",
  parameters: schema,
  providerOptions: { anthropic: { deferLoading: true } },
  execute: async ({ city }) => fetchWeather(city),
});

    the value is passed through verbatim; the AI SDK provider (@ai-sdk/anthropic, @ai-sdk/openai, ...) interprets it.

  • Updated dependencies [1315789, 299d448, 4429aa3, e76611f, 76f7d16, eef724e, 2dec3ae, fcb6baf, 331f2f7]:

    • @assistant-ui/core@0.2.8
    • @assistant-ui/store@0.2.13
    • assistant-cloud@0.1.30

1.3.30

Patch Changes

  • #4125 e639a11 - chore: drop tracker-behaviour explainer comments left behind in satellite runtimes (@Yonom)

1.3.29

Patch Changes

1.3.28

Patch Changes

  • #4107 32ae846 - feat: surface AI SDK v6 tool approvals as a first-class respondToApproval prop on tool components. tool-call parts in the approval-requested state now carry part.approval = { id, isAutomatic? }; tool components call respondToApproval({ approved, reason? }) to ack the gate without threading chatHelpers through application context. also fixes a transient requires-action flicker for the approval-responded state and tightens the external-message converter so interrupt vs pending tool calls are distinguished by an actual interrupt/approval field rather than by result === undefined. (@okisdev)

  • Updated dependencies [372d4f0, 32ae846]:

    • @assistant-ui/core@0.2.6

1.3.27

Patch Changes

1.3.26

Patch Changes

  • #4030 798a5ce - fix(react-ai-sdk): resolve MCP app metadata from tool output _meta["ui/resourceUri"] as a fallback when it isn't present in callProviderMetadata.mcp.app. MCP-UI tools (e.g. xmcp) surface the UI pointer in the call result, so the renderer previously never picked it up. (@Yonom)

1.3.25

Patch Changes

  • #4024 19d4d94 - feat: add native MCP Apps renderer — McpAppRenderer composes into Tools to render MCP UI resources inline in chat over a JSON-RPC postMessage bridge on SafeContentFrame. Adds an mcp field to ToolCallMessagePart and forwards callProviderMetadata.mcp.app through the AI SDK message converter. (@Yonom)

  • Updated dependencies [19d4d94]:

    • @assistant-ui/core@0.2.2

1.3.24

Patch Changes

  • #4003 717bed2 - feat: expose suggestions on useAISDKRuntime and useChatRuntime (@okisdev)

    both hooks now accept an optional suggestions: readonly ThreadSuggestion[] option that is forwarded to the underlying useExternalStoreRuntime. this lets AI SDK callers drive follow up suggestions from application state, tool results, or backend responses without dropping down to the raw external store runtime.

  • #4001 283c250 - feat(react-ai-sdk): expose onResume on useAISDKRuntime and useChatRuntime (@okisdev)

    AISDKRuntimeAdapter and UseChatRuntimeOptions now accept onResume, which is forwarded to the underlying useExternalStoreRuntime adapter. runtime.thread.resumeRun(config) previously threw "Runtime does not support resuming runs." because the inner adapter literal omitted the field; consumers had to monkey-patch runtime.thread.__internal_threadBinding.getState().resumeRun to bridge their own replay channels (e.g. SSE reconnect endpoints keyed by turn id). This is a thin pass-through; the existing transport-level resume on AssistantChatTransport (auto-fired by useChatRuntime on mount) is unchanged and complementary.

  • #3979 9ecda1d - feat(react-ai-sdk): native resumable stream client integration (@okisdev)

    AssistantChatTransport accepts a resumable: { storage, resumeApi, isFinishEvent? } option that captures the stream id from the response header, watches the SSE body for the AI SDK finish marker so the stored id is cleared on natural completion (cancellation leaves it intact for the next mount), and redirects chat.resumeStream() reconnects to resumeApi. createResumableSessionStorage is the default sessionStorage-backed ResumableClientStorage. useChatRuntime auto-fires chat.resumeStream() once on mount when storage already has an id, so adopters drop the manual useEffect.

  • #4008 fa4510a - feat: support multi-modal tool results via toModelOutput (@okisdev)

    frontend tools can now project their execution output into multi-modal model content (text + image / pdf / arbitrary file parts), aligning with the AI SDK v6 toModelOutput callback. previously, tool results were always serialized as a single JSON value, so a "read pdf" style tool had no way to send the PDF back to a multi-modal model.

    • assistant-stream exports a new ToolModelContentPart type ({ type: "text", text } | { type: "file", data, mediaType, filename? }) and a ToolModelOutputFunction<TArgs, TResult> callback type. Tool.toModelOutput is wired through unstable_runPendingTools and ToolExecutionStream, attaching the resulting modelContent to the tool-call part on the assistant message.
    • @assistant-ui/core re-exports ToolModelContentPart and adds an optional modelContent?: readonly ToolModelContentPart[] field on ToolCallMessagePart. existing tools and renderers are unchanged.
    • @assistant-ui/react-ai-sdk's frontendTools(...) helper now also registers a toModelOutput on each forwarded tool. it transparently unwraps an envelope that useAISDKRuntime writes when a frontend-executed tool produced modelContent, turning it into AI SDK's { type: "content", value: [...] } output. plain (non-envelope) outputs fall back to the existing { type: "text" | "json", value } shape, so behavior for tools without toModelOutput is unchanged.

    route handlers that adopt toModelOutput also need to pass tools to convertToModelMessages (this is the AI SDK's documented pattern):

    const aiSDKTools = { ...frontendTools(tools ?? {}) };
streamText({
  messages: await convertToModelMessages(messages, { tools: aiSDKTools }),
  tools: aiSDKTools,
});

    templates and existing examples are unchanged. they keep the simpler convertToModelMessages(messages) form because none of the tools they ship with use toModelOutput. the new tools guide page documents how to opt in.

    reserved key. when a frontend tool defines toModelOutput, its result is persisted in the AI SDK chat as { __aui_modelContent: ToolModelContentPart[], value: <your result> }. tools must not return objects whose top-level key is literally __aui_modelContent, or convertMessage will misread the result. the prefix is namespaced for this reason.

    read/write compatibility for persisted threads. the envelope is recognized by @assistant-ui/react-ai-sdk from this version onward. if you persist UI messages and read them from multiple environments, upgrade every reader before any writer starts producing toModelOutput; otherwise older readers will treat the envelope object as the result and break the affected tool render functions.

  • #4004 d3cf8cd - fix(react-ai-sdk): preserve tool args when AI SDK briefly emits null input (@okisdev)

    the AI SDK occasionally emits snapshots of an in-flight tool call where input is null/undefined, which previously collapsed argsText to "{}" mid-stream and tripped the runtime's append-only invariant (warning + tool args stream restart, losing accumulated state). convertMessage now caches the last good input per (message.id, toolCallId) via a new toolLastInputCache on AISDKMessageConverterMetadata and falls back to it when a later snapshot drops input.

  • #3634 9c3d24d - Support AI SDK source-document parts by preserving them as assistant-ui (@sicko7947) document source message parts across conversion and cloud serialization, including the legacy React cloud encoder.

  • Updated dependencies [35d0146, fa4510a, c9dd16c, dea8bc7, 9c3d24d]:

    • @assistant-ui/core@0.2.1

1.3.23

Patch Changes

  • Updated dependencies [040d469]:
    • @assistant-ui/core@0.2.0

1.3.22

Patch Changes

  • #3942 fbd7ce3 - fix: aui.thread().append({ startRun: false }) no longer triggers a run on AI SDK runtime (@Yonom)

    useAISDKRuntime's onNew and onEdit always called chatHelpers.sendMessage, ignoring message.startRun. The hook now appends the message via chatHelpers.setMessages (with a generated id when needed) and returns early when startRun: false, so the message lands in the chat without kicking off a model call.

  • #3962 b090acb - chore: update dependencies (@Yonom)

  • Updated dependencies [7098bab, b090acb, 5fdf17e]:

    • @assistant-ui/core@0.1.18
    • @assistant-ui/store@0.2.10

1.3.21

Patch Changes

1.3.20

Patch Changes

  • #3876 ce865bc - chore: update dependencies (@Yonom)

  • #3841 435cfa0 - fix: request body id in useChatRuntime is now the real thread id instead of the literal "DEFAULT_THREAD_ID". AssistantChatTransport was resolving remoteId from the inner ExternalStoreThreadListRuntimeCore (which only echoes its default id); it now uses the outer RemoteThreadListRuntimeCore that actually calls the adapter. (@okisdev)

  • #3858 da0f598 - fix: useAISDKRuntime now throws when the supplied ThreadHistoryAdapter omits withFormat, instead of silently dropping all history load/append/update calls. The optional-call chain historyAdapter.withFormat?.(…).load() previously short-circuited to undefined. The withFormat-wrapped adapter is now memoized, and the persist effect short-circuits when no adapter is supplied (avoiding a redundant thread subscription). ThreadHistoryAdapter.withFormat gains a JSDoc note clarifying that it is required on the AI SDK path. (@okisdev)

  • Updated dependencies [c7a274e, ce865bc, ca8f526, c56f98f, 974d15e, 4b19d42, da0f598, d53ff4f, 20f8404, 17958c9]:

    • @assistant-ui/core@0.1.15
    • assistant-cloud@0.1.27
    • @assistant-ui/store@0.2.8

1.3.19

Patch Changes

  • c988db8: chore: update dependencies
  • a5bce86: fix: preserve latest thread token usage during pending turns
  • Updated dependencies [f20b9ca]
  • Updated dependencies [c988db8]
    • @assistant-ui/core@0.1.14
    • assistant-cloud@0.1.26
    • @assistant-ui/store@0.2.7

1.3.18

Patch Changes

  • 376bb00: chore: update dependencies
  • Updated dependencies [42bc640]
  • Updated dependencies [376bb00]
  • Updated dependencies [87e7761]
    • @assistant-ui/core@0.1.13
    • assistant-cloud@0.1.25

1.3.17

Patch Changes

  • bdce66f: chore: update dependencies
  • 209ae81: chore: remove aui-source export condition from package.json exports
  • Updated dependencies [6554892]
  • Updated dependencies [9103282]
  • Updated dependencies [876f75d]
  • Updated dependencies [bdce66f]
  • Updated dependencies [4abb898]
  • Updated dependencies [209ae81]
  • Updated dependencies [2dd0c9f]
  • Updated dependencies [af70d7f]
    • @assistant-ui/core@0.1.10
    • assistant-cloud@0.1.24
    • @assistant-ui/store@0.2.6

1.3.16

Patch Changes

  • 781f28d: feat: accept all file types and validate against adapter's accept constraint
  • 0924711: fix(react-ai-sdk): convert assistant file parts to FileMessagePart instead of dropping them
  • 52403c3: chore: update dependencies
  • Updated dependencies [781f28d]
  • Updated dependencies [3227e71]
  • Updated dependencies [0f55ce8]
  • Updated dependencies [83a15f7]
  • Updated dependencies [52403c3]
  • Updated dependencies [ffa3a0f]
    • @assistant-ui/core@0.1.9
    • assistant-cloud@0.1.23
    • @assistant-ui/store@0.2.5

1.3.15

Patch Changes

  • 01a59da: fix(react-ai-sdk): preserve runConfig.custom metadata after tool call resume in human-in-the-loop tools
  • 736344c: chore: update dependencies
  • c71cb58: chore: update dependencies
  • Updated dependencies [1406aed]
  • Updated dependencies [9480f30]
  • Updated dependencies [28a987a]
  • Updated dependencies [736344c]
  • Updated dependencies [ff3be2a]
  • Updated dependencies [70b19f3]
  • Updated dependencies [c71cb58]
    • @assistant-ui/core@0.1.8
    • @assistant-ui/store@0.2.4

1.3.14

Patch Changes

  • 349f3c7: chore: update deps
  • 642bcda: Add quote.tsx registry components and injectQuoteContext helper
  • Updated dependencies [1ed9867]
  • Updated dependencies [427ffaa]
  • Updated dependencies [349f3c7]
  • Updated dependencies [02614aa]
  • Updated dependencies [6cc4122]
  • Updated dependencies [642bcda]
    • @assistant-ui/core@0.1.6
    • assistant-cloud@0.1.22
    • @assistant-ui/store@0.2.3

1.3.13

Patch Changes

  • 8ed9d6f: Refactor React Native component API: move shared runtime logic (remote thread list, external store, cloud adapters, message converter, tool invocations) into @assistant-ui/core for reuse across React and React Native
  • Updated dependencies [5ae74fe]
  • Updated dependencies [8ed9d6f]
  • Updated dependencies [01bee2b]
    • @assistant-ui/core@0.1.3

1.3.12

Patch Changes

  • 57e26d2: chore: update dependencies

1.3.11

Patch Changes

  • a845911: chore: update dependencies
  • Updated dependencies [07dcce0]
  • Updated dependencies [a845911]
  • Updated dependencies [bc40eaf]
  • Updated dependencies [be23d74]
  • Updated dependencies [1eb059c]

1.3.10

Patch Changes

  • 51de636: feat(react-ai-sdk): add thread token-usage extraction helpers and hook
  • 7ad20d1: fix(react-ai-sdk): stabilize tool args key order to prevent duplicate toolCallId crash in tapResources
  • Updated dependencies [17cf9a8]

1.3.9

Patch Changes

  • 36ef3a2: chore: update dependencies
  • Updated dependencies [36ef3a2]
  • Updated dependencies [6692226]
  • Updated dependencies [c31c0fa]
  • Updated dependencies [1672be8]
  • Updated dependencies [28f39fe]
  • Updated dependencies [3a1cb66]
  • Updated dependencies [14769af]
  • Updated dependencies [7c360ce]
  • Updated dependencies [a638f05]
  • Updated dependencies [8a78cd2]

1.3.8

Patch Changes

  • aeec3b9: fix: handle AI SDK v6 approval tool states (approval-requested, approval-responded, output-denied)
  • 7836760: fix(assistant-cloud): expand joined messages for AI SDK v6 history export and telemetry reporting
  • 61b54e9: Add message timing metadata: AssistantMessageTiming type, automatic timing tracking in AssistantMessageAccumulator, MessageTiming type, useMessageTiming() hook, and client-side streaming timing for AI SDK runtime.
  • a247fc9: feat(assistant-cloud): allow save complete multi-step message
  • 93910bd: Rename .tsx files to .ts where no JSX syntax is used
  • Updated dependencies [d08a488]
  • Updated dependencies [5bbe8a9]
  • Updated dependencies [5e304ea]
  • Updated dependencies [546c053]
  • Updated dependencies [a7039e3]
  • Updated dependencies [16c10fd]
  • Updated dependencies [98c3d54]
  • Updated dependencies [b181803]
  • Updated dependencies [7836760]
  • Updated dependencies [9276547]
  • Updated dependencies [b65428e]
  • Updated dependencies [af5b085]
  • Updated dependencies [61b54e9]
  • Updated dependencies [a094c45]
  • Updated dependencies [4d7f712]
  • Updated dependencies [ecc29ec]
  • Updated dependencies [6e97999]
  • Updated dependencies [a247fc9]
  • Updated dependencies [f414af9]
  • Updated dependencies [b48912c]
  • Updated dependencies [93910bd]
  • Updated dependencies [58a8472]

1.3.7

Patch Changes

  • afaaf3b: fix: duplicate key toolCallId error in HITL tools (#3197)
  • Updated dependencies [afaaf3b]
  • Updated dependencies [afaaf3b]
  • Updated dependencies [afaaf3b]
  • Updated dependencies [afaaf3b]
  • Updated dependencies [51d24be]
  • Updated dependencies [afaaf3b]

1.3.6

Patch Changes

  • a088518: chore: update dependencies
  • Updated dependencies [a088518]
  • Updated dependencies [d8122cc]

1.3.5

Patch Changes

1.3.4

Patch Changes

  • d45b893: chore: update dependencies
  • Updated dependencies [d45b893]
  • Updated dependencies [fe71bfc]

1.3.3

Patch Changes

  • acbaf07: feat: add framework-agnostic toToolsJSONSchema and toGenericMessages utilities to assistant-stream
  • Updated dependencies [07d1c65]
  • Updated dependencies [b591d72]
  • Updated dependencies [59a338a]
  • Updated dependencies [acbaf07]
  • Updated dependencies [c665612]
  • Updated dependencies [0371d72]
  • Updated dependencies [e8b3f34]

1.3.2

Patch Changes

  • 605d825: chore: update dependencies
  • Updated dependencies [1ea3e28]
  • Updated dependencies [8cbf686]
  • Updated dependencies [a8be364]
  • Updated dependencies [605d825]

1.3.1

Patch Changes

  • 5876c0f: feat(react-ai-sdk): propagate ModelContext config (modelName) through AssistantChatTransport
  • Updated dependencies [8672695]

1.2.1

Patch Changes

  • 083ed83: fix(cloud): serialize image and file parts across all formats
  • Updated dependencies [6eab31e]
  • Updated dependencies [9314b36]
  • Updated dependencies [083ed83]
  • Updated dependencies [6511990]
  • Updated dependencies [a526e63]

1.2.0

Minor Changes

  • 6fd744a: Upgrade to Vercel AI SDK v6
    • Updated peer dependencies: ai@^6.0.0, @ai-sdk/react@^3.0.0
    • Renamed format adapter from aiSDKV5FormatAdapter to aiSDKV6FormatAdapter
    • Updated @ai-sdk/provider to use LanguageModelV2 types

Patch Changes

1.1.21

Patch Changes

  • 5ce45bd: fix(react-ai-sdk): preserve metadata in message conversion
  • Updated dependencies [3a0b9c8]
  • Updated dependencies [b8f62e1]
  • Updated dependencies [8d5c20c]

1.1.20

Patch Changes

  • 57bd207: chore: update dependencies
  • cce009d: chore: use tsc for building packages
  • Updated dependencies [57bd207]
  • Updated dependencies [cce009d]

1.1.19

Patch Changes

  • e8ea57b: chore: update deps
  • Updated dependencies [bae3aa2]
  • Updated dependencies [e8ea57b]

1.1.18

Patch Changes

  • 89aec17: feat: AI SDK frontend tool execution cancellation support fix: AI SDK isRunning status when running frontend tools
  • Updated dependencies [89aec17]
  • Updated dependencies [ee7040f]
  • Updated dependencies [bd27465]
  • Updated dependencies [a3e9549]
  • Updated dependencies [206616b]
  • Updated dependencies [7aa77b5]

1.1.17

Patch Changes

  • 01c31fe: chore: update dependencies
  • Updated dependencies [ba26b22]
  • Updated dependencies [d169e4f]
  • Updated dependencies [da9f8a6]
  • Updated dependencies [01c31fe]

1.1.16

Patch Changes

  • c4142ac: fix(react-ai-sdk): pass runConfig metadata to backend API request

1.1.15

Patch Changes

  • ab8953b: feat(react): add allowNesting option to allow wrapping runtimes with custom thread list adapters
  • Updated dependencies [ab8953b]

1.1.14

Patch Changes

  • ec662cd: chore: update dependencies
  • cdb5ea5: mark tool call as complete once user sends new message when tool calling
  • 5dd925e: feat(ai-sdk): allow updates to headers/body
  • Updated dependencies [ec662cd]

1.1.13

Patch Changes

1.1.12

Patch Changes

  • faed815: feat: AI SDK error toolOutput support
  • 2c33091: chore: update deps
  • Updated dependencies [2c33091]

1.1.11

Patch Changes

  • 0bcbb58: feat: custom toCreateMessage callback fix: use AI SDK's idGenerator function for new messages
  • b408005: feat(react-ai-sdk): Integrate AI SDK v5 data parts in message content
  • Updated dependencies [b408005]
  • Updated dependencies [7a6d9ca]
  • Updated dependencies [70d5966]
  • Updated dependencies [3754bdd]
  • Updated dependencies [0a4bdc1]

1.1.10

Patch Changes

  • 34d1c78: fix(react-ai-sdk): correctly initialize history loading state
  • Updated dependencies [66a13a0]
  • Updated dependencies [4e3877e]
  • Updated dependencies [eef682b]

1.1.9

Patch Changes

  • 81b581f: feat: display AI SDK errors
  • 6d2c134: feat: useChatRuntime should use the assistant ui thread id remote id as the threadid by default
  • 2fc7e99: chore: update deps
  • Updated dependencies [3ab9484]
  • Updated dependencies [7a88ead]
  • Updated dependencies [81b581f]
  • Updated dependencies [2fc7e99]

1.1.8

Patch Changes

  • 953db24: chore: update deps
  • Updated dependencies [953db24]
  • Updated dependencies

1.1.7

Patch Changes

1.1.6

Patch Changes

1.1.5

Patch Changes

  • e6a46e4: chore: update deps
  • Updated dependencies [e6a46e4]

1.1.4

Patch Changes

  • e81784b: feat: Tool Call interrupt() resume() API
  • Updated dependencies [e8d6d7b]
  • Updated dependencies [e81784b]

1.1.3

Patch Changes

  • e46e4d3: fix: guard threadItem access for useAISDKRuntime
  • Updated dependencies [c0f5003]
    • assistant-stream@0.2.31

1.1.2

Patch Changes

  • 8812f86: chore: update deps
  • Updated dependencies [8812f86]
    • assistant-stream@0.2.30
    • assistant-cloud@0.1.2

1.1.1

Patch Changes

  • 68ef242: feat(ui): load external history only when thread has remote id
  • Updated dependencies [2c6198a]

1.1.0

Patch Changes

  • 39ac2f3: feat: AI SDK v5 import support
  • Updated dependencies [39ac2f3]
  • Updated dependencies [5437dbe]

1.0.7

Patch Changes

  • d7d9058: fix: cloud chat history not loading in some configurations

1.0.6

Patch Changes

  • 860bf42: feat: useAISDKRuntime cloud history support (without useChatRuntime)
  • Updated dependencies [3498c99]

1.0.5

Patch Changes

  • 90fc83b: fixes attachment naming
  • e64b20c: fix: persistence only saving the first two messages

1.0.4

Patch Changes

  • 9235fe1: update dep array in external history adapter

1.0.3

Patch Changes

  • ceedf45: feat: pass run-config to ai-sdk metadata to let user decide what to do after
  • 5504836: pass callsettings in extra body object to AI chat transport
  • Updated dependencies [a80dcff]

1.0.2

Patch Changes

  • 672db5a: feat: frontend function calling support
  • 12e0a77: chore: update deps
  • Updated dependencies [12e0a77]

1.0.1

Patch Changes

1.0.0

Patch Changes

0.11.5

Patch Changes

0.11.4

Patch Changes

  • 979ee67: feat: forward system and tools to the backend for useChatRuntime
  • 979ee67: feat: assistant cloud support for AI SDK v5
  • 979ee67: feat: add AssistantChatTransport
  • Updated dependencies [979ee67]

0.11.3

Patch Changes

  • 2ef6cae: feat: Add useChatRuntime as new recommended entry point for AI-SDK V5
  • Updated dependencies [f32b6a4]

0.11.1

Patch Changes

  • 20ffa06: fix: Don't omit attachments from AISDKRuntimeAdapter type
  • Updated dependencies [ed78407]
  • Updated dependencies [77ce337]
  • Updated dependencies [f59959e]

0.11.0

Patch Changes

  • 0f063e0: chore: update dependencies
  • Updated dependencies [0f063e0]
  • Updated dependencies [5d8b074]
Reference in New Issue View Git Blame Copy Permalink