anthropic[patch]: ensure content exists when parsing tool tokens

[snlamm]

Fixes #6045

When anthropic extracts tokens during tool use, it's throwing an error on the message_start event, as it results in a chunk with an empty content array, but the current token parsing assumes that a single content item always exists. This PR fixes the issue.

{
  type: 'message_start',
  message: {
    id: 'msg123',
    type: 'message',
    role: 'assistant',
    model: 'claude-3-5-sonnet-20240620',
    content: [],
    stop_reason: null,
    stop_sequence: null,
    usage: { input_tokens: 1486, output_tokens: 12 }
  }
}

The current error thrown:

error: TypeError: Cannot use 'in' operator to search for 'input' in undefined
        at extractToken

[vercel]

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

Name	Status	Preview	Comments	Updated (UTC)
langchainjs-api-refs	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 23, 2024 8:40pm
langchainjs-docs	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 23, 2024 8:40pm

[snlamm]

cc @bracesproul

[bracesproul]

Thanks for pushing up this fix. I pushed a commit adding a test as well.

[snlamm]

nice!