chore(deps): update dependency xsschema to ^0.5.1

[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
xsschema (source)	^0.4.2 → ^0.5.1

---

Release Notes

moeru-ai/xsai (xsschema)

v0.5.1

Compare Source

   🐞 Bug Fixes

• responses:
  • Support reasoning_text stream events  -  by @​kwaa, @​luoling8192 and @​nekomeowww (f7210)
  • Finish and clean up response streams  -  by @​kwaa, @​luoling8192 and @​nekomeowww (08cf9)
  • Replay output items unchanged  -  by @​kwaa (f7e52)
  • Convert chat tool output parts for Responses  -  by @​kwaa, @​luoling8192 and @​nekomeowww (9e8ca)
  • Await async event and step callbacks  -  by @​kwaa, @​luoling8192 and @​nekomeowww (e4608)
  • Pass native tools through to the provider  -  by @​kwaa, @​luoling8192 and @​nekomeowww (25f1d)
  • Validate tool parts and widen request options  -  by @​kwaa (b16d2)

    View changes on GitHub

v0.5.0: "mirai"

Compare Source

   🧳 Migration Guides

Version 0.5.0 includes breaking API changes. Upgrade all xsAI packages to 0.5.0 together and avoid mixing 0.4.x and 0.5.x packages.

Replace maxSteps with stopWhen

Replace:

generateText({
  maxSteps: 2,
})

with:

import { stepCountAtLeast } from '@xsai/shared-chat'

generateText({
  stopWhen: stepCountAtLeast(2),
})

The default remains one step. CompletionStep.stepType has been removed; use finishReason, toolCalls, and steps instead.

stopWhen is evaluated before tool execution. With the default stepCountAtLeast(1), generateText() returns pending toolCalls for manual execution and toolResults may be empty. If your 0.4.4 code relied on automatic tool execution, set stopWhen: stepCountAtLeast(2) or use a custom condition that allows another step. If you need manual approval, keep the default behavior, execute or approve the returned toolCalls, append the resulting tool messages to messages, and call generateText() again. Top-level tool results describe the final step; results from earlier steps are available in steps[n].toolResults.

Update stream event handling

streamText().eventStream now provides normalized xsAI events. fullStream now provides raw provider chat-completion chunks.

Common event changes:

0.4.4	0.5.0
text-delta with text	text.delta with delta
reasoning-delta with text	reasoning.delta with delta
tool-call-streaming-start	tool-call.start
tool-call-delta with argsTextDelta	tool-call.delta with delta
tool-call	tool-call.done
tool-result	tool-result.done
finish	step.done

Update onEvent handlers accordingly. The same changes apply to streamObject.

Update usage fields

Normalized usage fields now use camelCase:

// 0.4.4
usage.prompt_tokens
usage.completion_tokens
usage.total_tokens

// 0.5.0
usage.inputTokens
usage.outputTokens
usage.totalTokens

generateText() and generateObject() now also return totalUsage, which includes usage from all steps and tool-call follow-ups.

Update tool integrations

If you use executeTool() directly, its return field has changed from message to result.

Version 0.5.0 also adds tool input validation, preToolCall, and postToolCall. defineTool() is the recommended synchronous API for schemas that support Standard JSON Schema.

Tool parsing and execution failures are generally returned as tool results with isError: true.

Update error handling

Errors are now represented by dedicated classes such as APICallError, InvalidResponseError, JSONParseError, InvalidToolCallError, InvalidToolInputError, and ToolExecutionError.

If your code checks XSAIError fields directly, update it to use the specialized error classes or the error code property.

Removed and deprecated APIs

toAsyncIterator has been removed from @xsai/utils-stream.

extractReasoning() and extractReasoningStream() are deprecated. Use generateText().reasoningText and streamText().reasoningTextStream instead.

For speech streaming, use the new @xsai/stream-speech package and its streamSpeech() API.

   🚨 Breaking Changes

• Remove step type  -  by @​kwaa in #​295 (a56ae)
• New responses-style usage  -  by @​kwaa in #​297 (539b2)
• Reuse stop condition  -  by @​kwaa in #​298 (6e481)
• Rename prepare step options, reuse  -  by @​kwaa in #​300 (8fa74)
• New event stream  -  by @​kwaa in #​311 (0f1a3)
• Handle tool call error  -  by @​kwaa in #​319 (cccf4)
• Check should stop before execute tool  -  by @​kwaa in #​320 (256d8)
• providers: Sync upstream  -  by @​kwaa (dcc39)
• stream-text: Rename event stream, add new full stream  -  by @​kwaa in #​303 (565f7)
• utils-reasoning: Deprecate  -  by @​kwaa in #​290 (63c35)
• utils-stream: Remove to async iterator  -  by @​kwaa in #​284 (0ace2)

   🚀 Features

• Responses  -  by @​kwaa in #​263 (945c3)
• Stop when  -  by @​kwaa in #​271 (32a10)
• Prepare step  -  by @​kwaa in #​275 (b545b)
• Typed error  -  by @​kwaa in #​277 (e015d)
• PreToolCall, postToolCall  -  by @​exoticknight in #​309 (905b9)
• Stream speech, update speech options  -  by @​kwaa in #​310 (4d3e8)
• generate-text:
  • Support total usage  -  by @​kwaa in #​304 (18f28)
• providers:
  • Sync upstream  -  by @​kwaa in #​281 (688fc)
  • Add Baidu Qianfan provider  -  by @​jimmyzhuu in #​287 (3e1a3)
  • Sync upstream  -  by @​PierreLeGuen in #​308 (168ea)
• responses:
  • Camel case, use chat tool, typed error  -  by @​kwaa in #​286 (c49ff)
  • Reasoning text stream  -  by @​kwaa in #​293 (3ee9d)
  • Hooks  -  by @​kwaa in #​301 (9850f)
  • Return input  -  by @​kwaa in #​302 (52873)
  • Mark as public  -  by @​kwaa (3b2f1)
• tool:
  • Define tool  -  by @​kwaa in #​312 (e233b)
  • Validate schema  -  by @​kwaa in #​314 (47a68)

   🐞 Bug Fixes

• Lock workspace version  -  by @​kwaa (44c35)
• Correctly clean request body  -  by @​kwaa in #​305 (083ff)
• providers: Workers ai models type  -  by @​kwaa (08683)
• stream: Remove controller  -  by @​kwaa in #​316 (2cba9)
• stream-text: Handle callback error  -  by @​kwaa (984eb)
• telemetry: Align  -  by @​kwaa in #​280 (eb664)

   🏎 Performance

• New shared-stream package  -  by @​kwaa in #​283 (b2790)
• Reduce size  -  by @​kwaa in #​306 (f9a52)

    View changes on GitHub

v0.4.4

Compare Source

   🐞 Bug Fixes

• shared-chat:
  • Handle empty tool call arguments in executeTool  -  by @​lietblue in #​266 (e23b9)
  • Clean execute only  -  by @​Wyatex in #​270 (02092)

    View changes on GitHub

v0.4.3

Compare Source

   🏎 Performance

• Use Promise.all for tool calling  -  by @​dsh0416 in #​264 (9241e)

    View changes on GitHub

---

Configuration

📅 Schedule: (in timezone UTC)

• Branch creation
  • "on sunday before 6:00am"
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Enabled.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[netlify]

✅ Deploy Preview for neuri ready!

Name	Link
🔨 Latest commit	f3cb3ff
🔍 Latest deploy log	https://app.netlify.com/projects/neuri/deploys/6abde1d8517d81000737c51f
😎 Deploy Preview	https://deploy-preview-175--neuri.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.