fix: guard Anthropic thinking budgetTokens for unknown output limits

[comzip]

Issue for this PR

Fixes #22253

Type of change

• Bug fix

What does this PR do?

Custom provider models without a limit field get limit.output = 0. The budgetTokens formula Math.floor(output / 2 - 1) returns -1 in this case, which causes an error when Anthropic's API validates the thinking config. Extracted highBudget() to centralize the formula and guard against invalid results (Anthropic requires budgetTokens >= 1024). Both variants() and the kimi-k2.5 path now share this helper, so the guard lives in one place instead of being duplicated.

How did you verify your code works?

Added two unit tests in transform.test.ts:

• output: 0 → returns {}
• output: 2049 → returns {} (budgetTokens would be below Anthropic's 1024 minimum)

Screenshots / recordings

N/A

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search results, I found one potentially related PR:

Related PR:

• fix: handle invalid maxOutputTokens value (0 or undefined) #22016 - "fix: handle invalid maxOutputTokens value (0 or undefined)"
  • fix: handle invalid maxOutputTokens value (0 or undefined) #22016
  • This PR appears to address a similar issue with invalid maxOutputTokens values, which is closely related to the budgetTokens calculation problem in PR fix: guard Anthropic thinking budgetTokens for unknown output limits #22581 where output=0 causes negative budget values.

The other results (#12544, #12342) are older PRs about Anthropic thinking features but not directly addressing this specific bug fix.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.