fix: handle tool validation errors in stopWhen and willContinue

[elpiarthera]

Summary

• hasSuccessfulToolCall(toolName) — new stop condition that only matches tool calls which produced a tool-result (not tool-error). Use instead of the AI SDK's hasToolCall when agents should retry on argument validation failures.
• willContinue fix — counts tool-error content parts as completed outputs. In AI SDK v6, failed tool calls produce tool-error instead of tool-result, causing willContinue to incorrectly treat them as pending and stop prematurely.

Context

When a tool call fails Zod argument validation in AI SDK v6, the result is a tool-error content part (not tool-result). Two problems occurred:

1. willContinue checked step.toolCalls.length > step.toolResults.length — since tool-error isn't in toolResults, it looked like a pending result, returning false (stop)
2. The AI SDK's hasToolCall matches any tool call regardless of success, so stopWhen: [hasToolCall("myTool")] would stop even on validation failures

Test plan

• npm run build passes
• npm run typecheck passes
• npm run lint passes (0 errors)
• npm test passes (262/262)
• Impact analysis: willContinue has a single caller (streamText.ts:143), change is backwards-compatible

Fixes #172

Orchestrator: Zeta — VantageOS Team Dev | 2026-04-06

Summary by CodeRabbit

• New Features

  • Added functionality to check whether a specific tool call completed successfully.

• Bug Fixes

  • Enhanced tool call completion detection to properly recognize and handle both successful results and error responses in tool interactions.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 00b72f2f-26e1-46e6-9fd8-a5d05ca3ef37

📥 Commits

Reviewing files that changed from the base of the PR and between 92164db and e6238e1.

📒 Files selected for processing (2)

• src/client/index.ts
• src/client/utils.ts

---

📝 Walkthrough

Walkthrough

Added a new hasSuccessfulToolCall stop condition helper and updated tool completion detection to recognize both tool-error and tool-result content as completed tool execution, enabling proper distinction between attempted and successful tool calls.

Changes

Cohort / File(s)	Summary
Export Configuration src/client/index.ts	Added import and re-export of the new hasSuccessfulToolCall helper function to the public API.
Tool Execution Detection src/client/utils.ts	Added hasSuccessfulToolCall(toolName) function that returns true only when the last step contains a successful tool result matching the tool name. Updated willContinue() to count both tool-result and tool-error content parts as completed outputs, and changed control-flow comparison to check toolCalls.length against completedOutputs.

Possibly related issues

• Agent stopWhen stops even when tool call fails Zod validation #172 — Directly addresses the issue where stopWhen conditions trigger even when tool calls fail validation; the new hasSuccessfulToolCall distinguishes successful tool calls from failed ones by checking toolResults instead of merely detecting tool call attempts.
• willContinue doesn't count tool-error as completed output (AI SDK v6) #240 — Related modification to willContinue logic in src/client/utils.ts regarding tool-error content handling.

Poem

🐰 Hops of joy through tool-call land,
Now success and failure we understand!
No more stopping when validation fails,
Only true completions ring success bells! 🔔

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main fix: handling tool validation errors in stopWhen and willContinue functions.
Linked Issues check	✅ Passed	The PR fully addresses issue #172 by adding hasSuccessfulToolCall and fixing willContinue to properly handle tool-error parts.
Out of Scope Changes check	✅ Passed	All changes in src/client/index.ts and src/client/utils.ts are directly related to fixing the tool validation error handling.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}