fix: subsequent tool call args are empty for gpt-5

[uinstinct]

Description

Previously, the subsequent tool calls for gpt-5 models always produced empty tool arguments. This PR fixes that

Diagnosis:

GPT-5 models tend to be less verbose between tool calls - meaning they do tool calls without explanations in between.

streamUpdate was transforming the second tool call message twice here and here for the same first message after the tool call.
This added an extra character (for example, "{\ at the beginning of "{\"{\"command\":\"pwd\",\"waitForCompletion\":true") for which parsing the arguments failed.

resolves CON-4997

AI Code Review

• Team members only: AI review runs automatically when PR is opened or marked ready for review
• Team members can also trigger a review by commenting @continue-review

Checklist

• [] I've read the contributing guide
• [] The relevant docs, if any, have been updated or created
• [] The relevant tests, if any, have been updated or created

Screen recording or screenshot

before.mp4

after.mp4

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

---

Summary by cubic

Fixed empty tool arguments on subsequent GPT-5 tool calls by removing duplicate processing of the first message after a tool call. This prevents a stray leading character that broke JSON parsing and ensures tools receive correct args across consecutive calls (CON-4997).

^{Written for commit de5062f. Summary will update automatically on new commits.}

[continue]

Keep this PR in a mergeable state →

Learn more

---

All Green is an AI agent that automatically:

✅ Addresses code review comments

✅ Fixes failing CI checks

✅ Resolves merge conflicts

[continue]

Reviewed this PR for documentation updates.

No documentation changes needed because:

• This is an internal bug fix in the Redux state management layer
• No user-facing configuration or API changes
• No new features or capabilities introduced
• Users don't need to take any action

The fix is well-documented in the PR description and commit message, which is appropriate for an internal bug fix like this.

[cubic-dev-ai]

No issues found across 1 file

[RomneyDa]

@uinstinct nice digging to find the root cause here. The handleToolCallsInMessage which is removed here is the only path which is reached if there is no thinking for streamed tool calls. It seems like removing this would break tool calls for many other models/situations (not fully sure). Could we instead update handleToolCallsInMessage so that it doesn't add the extraneous character which is causing the parsing failures?

[uinstinct]

@RomneyDa Actually the reason for removing handleToolCallsInMessage in that if condition was to keep that if condition explicit for creating a new message. It eventually reaches the tool call condition which gets the tool call parsing and args.

I had tested them on other models and all seem to work fine. The only model which replicated the tool call behaviour sometimes (i.e. tool call without any message content) was kimi k2, which was also working fine.

kimi k2 subsequent tool usage

[RomneyDa]

@uinstinct thanks for the explanation!

[sestinj]

🎉 This PR is included in version 1.8.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[sestinj]

🎉 This PR is included in version 1.37.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[sestinj]

🎉 This PR is included in version 1.34.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[sestinj]

🎉 This PR is included in version 1.7.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀