Repository navigation
chore(deps): update ai sdk core and providers - #2364
renovate[bot] wants to merge 1 commit into
Conversation
|
Important Review skippedAuto reviews are limited based on label configuration. 🏷️ Required labels (at least one): 1
Please check the settings in the CodeRabbit UI or the
|
Review statusThis PR was opened by an automated account. A human maintainer must verify the change intent, provenance, and validation before merging. Current step: Awaiting fresh human maintainer or CODEOWNER approval. Review-state labels are managed by this workflow; do not edit them manually. |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
2a56f68 to
d7fa1bf
Compare
d7fa1bf to
16f3ef8
Compare
16f3ef8 to
8b5dcb6
Compare
8b5dcb6 to
d5f1049
Compare
d5f1049 to
a9536ab
Compare
a9536ab to
bb9a7c3
Compare
bb9a7c3 to
ef0d098
Compare
ef0d098 to
60acb62
Compare
60acb62 to
c5f29b3
Compare
c5f29b3 to
7a4b2d0
Compare
7a4b2d0 to
a4c42cf
Compare
This PR contains the following updates:
2.0.90→2.0.934.0.64→4.0.667.0.162→7.0.173Release Notes
vercel/ai (@ai-sdk/openai-compatible@2)
v2.0.93Compare Source
Patch Changes
88bb257]v2.0.92Compare Source
Patch Changes
f0dee4b]v2.0.91Compare Source
Patch Changes
14c196a: Avoid empty structured-output answers from Baseten's DeepSeek V4 Pro 0813 when reasoning is enabled. Fall back to JSON prompt instructions while keeping reasoning and SDK output validation, and report the fallback through request transformation warnings.98eb47b: fix(openai,openai-compatible): surface refusal logprobs and openai-compatible refusals@ai-sdk/openai: Chat Completions responses that contain a refusal can returnlogprobs.refusalnext tologprobs.content. The provider previously read onlylogprobs.content, soproviderMetadata.openai.logprobswas empty or missing for refusals. Refusal logprobs are now included in emission order (content logprobs followed by refusal logprobs) in bothgenerateTextandstreamText.Streaming logprobs are now collected across all chunks in
streamText: previously only the logprobs of the last streamed chunk were reported. This applies to content logprobs as well, soproviderMetadata.openai.logprobsnow contains the logprobs of every streamed token instead of only the last chunk's tokens. Explicitly empty logprobs (logprobs: { content: [], refusal: null }) are reported as an empty array, as before.A refusal that arrives in the same streamed chunk as the
finish_reasonis now taken into account: such a response is reported with acontent-filterfinish reason instead ofstop.@ai-sdk/openai-compatible: Chat Completions responses that carrymessage.refusal(non-streaming) ordelta.refusal(streaming) were ignored, so providers such as OpenAI served throughcreateOpenAICompatiblelost the refusal text and reported an emptystopresponse. Refusals are now returned as text content after any content text, and a non-empty refusal with astopfinish reason is reported as acontent-filterfinish reason. This also applies when the refusal and the finish reason arrive in the same streamed chunk. Non-string refusals (for examplerefusal: falsefrom servers that use booleans) are ignored instead of failing the response.vercel/ai (ai)
v7.0.173Compare Source
Patch Changes
3c09e00: Prevent in-place schema validation or input refinement from changing approved tool arguments during approval replay.f43ee39: fix(workflow): log the default step limit warning whenWorkflowAgent.generate()stops at the default limit.7ce7ffc: fix(workflow): log provider warnings inWorkflowAgentWorkflowAgent.generate()andWorkflowAgent.stream()now log the warnings of every model call through the AI SDK warning logger, likeToolLoopAgent. TheAI_SDK_LOG_WARNINGSglobal controls the output.f6809e3: fix(workflow): align theprepareStepresults ofWorkflowAgentwithToolLoopAgent. Returnedinstructions(or the deprecatedsystem) replace the instructions of the step and carry forward to later steps, returnedproviderOptionsare merged with the agent options, and returnedheadersare merged over the headers of the call.2559173: feat(workflow): supporttoolApprovalinWorkflowAgentWorkflowAgentaccepts atoolApprovalconfiguration in the constructor and pergenerate()orstream()call. The configuration takes precedence over a tool'sneedsApprovalsetting, and its functions run in workflow context, outside of the durable model steps. Calls that the configuration approves run in the same step, and calls that it denies receive anexecution-deniedtool result without pausing the workflow.f1de9cb: feat(workflow): support thetimeoutobject inWorkflowAgentgenerate()andstream()accepttimeout: { totalMs, stepMs, firstChunkMs, chunkMs }in addition to a number.stepMslimits each model call. Instream(),firstChunkMsandchunkMslimit the wait for model output. Invalid timeout values are rejected with anInvalidArgumentError. A timeout now also ends a model step when the provider stream ignores the abort signal.Updated dependencies [
9baa2d0]v7.0.172Compare Source
Patch Changes
79d2004: Require an explicitexperimental_downloadhook for remote file URLs inexperimental_extractandexperimental_decideinstead of silently downloading through global fetch. Callers can usecreateDownload({ fetch })with a DNS-protected transport for untrusted URLs; inline file data continues to work without a hook.6d50947]v7.0.171Compare Source
Patch Changes
11323f9: Prevent excessive memory and CPU usage when validating malformed UI messages.validateUIMessages,safeValidateUIMessages, and the agent UI stream helpers reported the issues of every part schema for every invalid part, so a small request with many malformed parts could exhaust memory. Each part is now validated against the schemas for its type only, validation stops at the first invalid message or part, and an invalid JSON value in provider or tool metadata is reported as a single issue.2ebc78b: Forward the provider's configured fetch transport when downloading file inputs. Add afetchoption tocreateDownloadso core downloads can also useguarded-fetchfor opt-in DNS validation on Node.js. The function returned bycreateDownloadcan now also be passed asexperimental_downloadtogenerateTextandstreamText, andexperimental_downloadnow also handles generated file URLs returned by the model, so one hook covers all core downloads. Document that the default URL checks do not resolve DNS and recommend the guarded transport or network egress controls for untrusted URLs.Keep Azure image and Vertex tool-result downloads separate from provider authentication wrappers so file hosts never receive Azure bearer tokens or Vertex Express Mode API keys.
Pass the call's abort signal to custom download requests, including prompt assets and generated files, so
experimental_download: createDownload({ fetch })preserves cancellation and timeouts.cb4fa52: add experimental extract api and extraction model specceeee29: Warn when a UI message stream sends a tool error text that differs from the text the model received and then sends reasoning from a later step. By default,toUIMessageStreamsends "An error occurred." to the client for a failed tool call, while the model receives the full error text during a multi-step call. Converting the stored UI messages back withconvertToModelMessagesthen changes the tool result, and models that bind reasoning to the preceding messages (for example Anthropic models with preserved thinking) reject the request. The request still fails after this change; the warning and the updated documentation explain the cause and the available options.1a94af3: Send the error text of failed tool calls to the client as the model received it.toUIMessageStreamandtoUIMessageChunkreplaced the error text of failed tool calls with "An error occurred." by default. The model had already received the full text during a multi-step call, so UI messages that were converted back withconvertToModelMessagescontained a different tool result. Models that bind reasoning to the preceding messages rejected that history (for example Anthropic models with preserved thinking, with HTTP 400 "The block is bound to a different conversation"), and other models continued from an error they never received.UI messages now keep the model's error text for tool execution errors and invalid tool calls. This text reaches the client. To keep details away from the model and the client, throw an error with a safe message from the tool's
executefunction. The error text of an invalid tool call contains the generated input, the validation messages of the input schema, or the names of the available tools;executedoes not run for it. To restore the previous behavior, set the newredactToolErrors: trueoption. A providedonErrorkeeps handling failed tool calls unless you setredactToolErrors: false. Stream errors are still masked by default.Updated dependencies [
640569d]Updated dependencies [
4e43721]Updated dependencies [
2ebc78b]Updated dependencies [
cb4fa52]v7.0.170Compare Source
Patch Changes
91a67c2: Fix excessive CPU usage when streaming long arrays withOutput.arrayorstreamObject. Parse and validate completed elements incrementally instead of repeatedly processing the entire accumulated array on every text chunk.84b1cb0: Cancel pending send-now operations when a later stop or dispose occurs, preventing an abandoned chat's queued messages and history from being sent through a replacement transport.eaeb37e]v7.0.169Compare Source
Patch Changes
4311fa1: Use Zod Mini for internal validation schemas to reduce bundle size while preserving the public message schemas' classic Zod methods, error messages, and compatibility with supported Zod 3 and Zod 4 versions.d9dd2b7: keep the request options when the useChat queue sends a tool approval first, and hold the queue while an async sendAutomaticallyWhen decides4311fa1]c160b74]4311fa1]838dd53]v7.0.168Compare Source
Patch Changes
65aa669: add experimental_queue to useChat to queue messages sent while a response is in progress8585d7e: fix(ai): calltoUIMessageStreamonErroronce per invalid tool call5497ff7]88b9575]v7.0.166Compare Source
Patch Changes
44db225]v7.0.165Compare Source
Patch Changes
883fcde: remove transcription word timings from telemetry when recordOutputs is false6d21e8e: Sanitize provider request URLs before text generation results reach telemetry integrations and tracing subscribers, preventing endpoint credentials from being exposed.6400cf8]f608c05]cab9c1f]c70adf6]v7.0.164Compare Source
Patch Changes
bace553]6737c77]bb4e2d0]v7.0.163Compare Source
Patch Changes
5d18351: fix(ai): detect downloaded audio types from bytes when servers return a generic or non-audio content type such as application/octet-stream8da48db: RespectrecordInputsandrecordOutputswhen delivering telemetry events to integrations and theai:telemetrychannel. Filter nested content and omit raw span results when content recording is disabled, while preserving application results.fd7e8a7: Preserve the order of text and complete function calls in streamed Google responses, including standalone thought signatures on empty text or reasoning parts.Keep provider metadata on empty text parts when converting results to response messages so it can be passed back in subsequent requests.
Omit foreign-provider empty signature text when replaying history through Anthropic or Bedrock, preserving Bedrock’s existing native reasoning-history handling.
36802d4: Support native URL transcription with an optional provider capability, keeping existing file-generation contracts unchanged. Groq sends HTTP(S) audio URLs through itsurlparameter without downloading and re-uploading them, avoiding the attachment upload limit. Supplying a customdownloadfunction retains local download behavior.For telemetry integrations,
TranscriptionEndEvent.audio.byteLengthandaudio.mediaTypecan now beundefinedwhen a provider fetches the audio URL natively. Check for missing values before using them as a number or string. Downloaded and inline audio continue to report their metadata; streaming transcription metadata is unchanged.Strip media fragments before sending URLs to Groq. If Groq rejects a redirect, fall back to the existing secure local downloader and file upload rather than failing previously working URL inputs. Other provider errors are preserved.
933dae3: fix (ai): download deprecatedfile-urlandimage-urltool result contentTool result content parts that use the deprecated
file-urlandimage-urltypes were converted to
fileparts with{ type: 'url' }data, but they werenever passed to the download function. When the model does not support the URL,
the provider received a bare URL instead of the file contents, even though the
replacement
filepart with{ type: 'url' }data was downloaded and inlined.Downloaded legacy URL parts now use the same media type normalization as file
parts, resolving generic types from the response header or image bytes.
Legacy base64 data URLs also normalize to inline data with the media type from
their data URL, matching replacement file parts.
e0143b0: feat(ai): support provider-owned realtime transportsAdd an optional
createTransportinterface andproviderconnection capability.934100c: Report the actual OpenAI Chat HTTP endpoint (also used by Azure Chat and Bedrock Mantle Chat) through optional request metadata and model-call end events. Record its host and port on OpenTelemetry chat spans without exposing URL credentials, paths, queries, or fragments. Expose only the sanitized HTTP origin in text and object generation results and step/finish callbacks. Providers that do not report an endpoint leave these attributes unset.c389374: fix(ai): infer union output types when passing a union of schemas toOutput.objectandOutput.arrayPreviously, passing a schema whose type is a union of schemas (e.g. a schema selected from a map at runtime) failed to type-check because the output type was inferred from the first union member only. The output type is now inferred as the union of the schemas' output types.
892af4d: feat (ai): re-export realtime server-connection and usage typesAdds
Experimental_RealtimeServerConnection,Experimental_RealtimeSessionIntent, andExperimental_RealtimeUsagealiases for the new experimentalRealtimeModelV4server-side spec types.1a1da8d: stream steering messages to the ui and keep them in converted messages41465fd: Expose optional transcription word timings separately from segments and accept shared timestampGranularities. Explicit provider timing settings take precedence. OpenAI now sends timestamp_granularities only with verbose_json and warns when a requested timing is unavailable or excluded.9754014: Add typed streaming heartbeat parts. Anthropic emits them only withincludePings: true(disabled by default), without requiring raw chunks. Core forwards pings through stream retries without storing them as generated content or sending them to UI message streams. Consumers with exhaustive stream-part switches must handle the newpingunion variant when upgrading, even when the opt-in is disabled. Typed pings do not flush partial words during stream smoothing.2e44144: Represent videos delivered directly to caller-managed storage with an explicit uploaded outcome. Video generation reportsuploadedVideoswithout downloading or synthesizing video bytes; accessvideoonly whenvideosis nonempty.Updated dependencies [
722d4a9]Updated dependencies [
acea2b5]Updated dependencies [
89aff56]Updated dependencies [
41465fd]Updated dependencies [
2e44144]Updated dependencies [
36802d4]Updated dependencies [
ef2dfb5]Updated dependencies [
e0143b0]Updated dependencies [
934100c]Updated dependencies [
892af4d]Updated dependencies [
892af4d]Updated dependencies [
89aff56]Updated dependencies [
32acf43]Updated dependencies [
41465fd]Updated dependencies [
9754014]Updated dependencies [
4bb73b0]Updated dependencies [
2e44144]Configuration
📅 Schedule: (in timezone America/Los_Angeles)
🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.
♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.
This PR was generated by Mend Renovate. View the repository job log.