fix(core): update chunk concat logic to match on missing ID fields

[mcblum]

This PR updates the logic so that if tool chunk indices match and one chunk does not have an ID, the tool chunks are then merged.

Fixes #8985

[changeset-bot]

🦋 Changeset detected

Latest commit: e8cd966

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@langchain/core	Patch
@langchain/standard-tests	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
langchainjs-docs	Ready	Preview		Sep 17, 2025 4:51pm

1 Skipped Deployment

Project	Deployment	Preview	Comments	Updated (UTC)
langchainjs-api-refs	Ignored			Sep 17, 2025 4:51pm

[mcblum]

@hntrl ok pushed a couple changes. I think either I'm losing it or we are reaching the logic where one or more IDs are undefined, because the tests still pass. Maybe the named conditions make it clearer?

[flipace]

@mcblum i really have no idea anymore what's going on, but i tested this change locally w anthropic claude and I see chunks with 'id' being undefined so the check needs to also merge if id values are falsy?

(note: we're using portkey ai gateway in between our service and anthropic so maybe this is useful information too. we only have this issue with claude/anthropic atm though)

(sorry for the verbose output)

17:12:25.734 info {
  leftItem: {
    index: 0,
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '' }
  },
  item: { index: 0, function: { arguments: '' } }
}
17:12:25.734 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:12:25.734 info { toMerge: 0 }
17:12:25.735 info {
  leftItem: {
    name: 'text_to_speech',
    args: '',
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: '',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:12:25.735 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: false }
17:12:25.735 info { toMerge: -1 }
17:12:25.863 info {
  leftItem: {
    index: 0,
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '' }
  },
  item: { index: 0, function: { arguments: '{"text": "h' } }
}
17:12:25.863 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:12:25.863 info { toMerge: 0 }
17:12:25.863 info {
  leftItem: {
    name: 'text_to_speech',
    args: '',
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: '{"text": "h',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:12:25.863 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: false }
17:12:25.863 info {
  leftItem: {
    name: undefined,
    args: '',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: '{"text": "h',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:12:25.863 info { indiciesMatch: true, idsMatch: true, eitherItemMissingID: false }
17:12:25.863 info { toMerge: 1 }
17:12:25.864 info {
  leftItem: {
    index: 0,
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '{"text": "h' }
  },
  item: { index: 0, function: { arguments: 'allo welt"}' } }
}
17:12:25.864 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:12:25.864 info { toMerge: 0 }
17:12:25.864 info {
  leftItem: {
    name: 'text_to_speech',
    args: '',
    id: 'toolu_01JZCDc1Y43MHPCsiyyZp9SU',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: 'allo welt"}',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:12:25.864 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: false }
17:12:25.864 info {
  leftItem: {
    name: undefined,
    args: '{"text": "h',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: 'allo welt"}',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:12:25.864 info { indiciesMatch: true, idsMatch: true, eitherItemMissingID: false }
17:12:25.864 info { toMerge: 1 }

this ends up as (note: args are in a separate chunk now):

"tool_call_chunks": [
    {
      "name": "text_to_speech",
      "args": "",
      "id": "toolu_01JZCDc1Y43MHPCsiyyZp9SU",
      "index": 0,
      "type": "tool_call_chunk"
    },
    {
      "args": "{\"text\": \"hallo welt\"}",
      "index": 0,
      "type": "tool_call_chunk"
    }
  ]

also checking for falsy id values

adjusting this pr change slightly to:

const eitherItemMissingID = !("id" in leftItem) || !("id" in item) || !leftItem?.id || !item?.id;

leads to this (expected output):

17:14:17.925 info {
  leftItem: {
    index: 0,
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '' }
  },
  item: { index: 0, function: { arguments: '' } }
}
17:14:17.926 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:17.926 info { toMerge: 0 }
17:14:17.926 info {
  leftItem: {
    name: 'text_to_speech',
    args: '',
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: '',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:14:17.926 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:17.926 info { toMerge: 0 }
17:14:18.300 info {
  leftItem: {
    index: 0,
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '' }
  },
  item: { index: 0, function: { arguments: '{"t' } }
}
17:14:18.300 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.300 info { toMerge: 0 }
17:14:18.301 info {
  leftItem: {
    name: 'text_to_speech',
    args: '',
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: '{"t',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:14:18.301 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.301 info { toMerge: 0 }
17:14:18.301 info {
  leftItem: {
    index: 0,
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '{"t' }
  },
  item: { index: 0, function: { arguments: 'ext": "hallo' } }
}
17:14:18.301 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.301 info { toMerge: 0 }
17:14:18.301 info {
  leftItem: {
    name: 'text_to_speech',
    args: '{"t',
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: 'ext": "hallo',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:14:18.302 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.302 info { toMerge: 0 }
17:14:18.320 info {
  leftItem: {
    index: 0,
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    type: 'function',
    function: { name: 'text_to_speech', arguments: '{"text": "hallo' }
  },
  item: { index: 0, function: { arguments: ' welt"}' } }
}
17:14:18.320 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.320 info { toMerge: 0 }
17:14:18.320 info {
  leftItem: {
    name: 'text_to_speech',
    args: '{"text": "hallo',
    id: 'toolu_019JviDh5ExdUjZszB5KA1VA',
    index: 0,
    type: 'tool_call_chunk'
  },
  item: {
    name: undefined,
    args: ' welt"}',
    id: undefined,
    index: 0,
    type: 'tool_call_chunk'
  }
}
17:14:18.320 info { indiciesMatch: true, idsMatch: false, eitherItemMissingID: true }
17:14:18.320 info { toMerge: 0 }

leads to the expected output:

"tool_call_chunks": [
    {
      "name": "text_to_speech",
      "args": "{\"text\": \"hallo welt\"}",
      "id": "toolu_019JviDh5ExdUjZszB5KA1VA",
      "index": 0,
      "type": "tool_call_chunk"
    }
  ]

[hntrl]

@flipace yeah what you're seeing is a different issue that got reported this morning, but is relevant for this PR. Thanks for the suggestion, have added it!