feat: add request / response body logging to Claude batch API

[tasoo-oos]

PR Checklist

• Required Checks
  • Have you added type definitions?
  • Have you tested your changes?
  • Have you checked that it won't break any existing features?
• If your PR uses models¹, check the following:
  • Have you checked if it works normally in all models?
  • Have you checked if it works normally in all web, local, and node-hosted versions? If it doesn't, have you blocked it in those versions?
• If your PR is highly AI generated², check the following:
  • Have you understood what the code does?
  • Have you cleaned up any unnecessary or redundant code?
  • Is it not a huge change?
    • We currently do not accept highly AI generated PRs that are large changes.

Note

This PR is a continuation of the pre-existing PR called #1263.
I seperated this PR from that because this PR introduces seperate feature based on the new batch API implementation.

Summary

Unlike Other streaming
Add request / response body logging logic to the claude batch API.

Related Issues

None

Changes

Before changes

Example images

• For per-message debug logging
  • Batch API does not leave any log
• Request Logs viewer UI in advanced settings
  • Are filled with polling requests, which is useless for end users
  • No response body logged at all, only "Streamed Fetch"

After changes

Example images

• Batch API does leave a request / response log
• Now there's two log (initial batch API call, final batch result / errored API call) per one batch request
  • No redundant polling request logs saved
• Now request uses header format defined in anthropic.ts. Refer to Example images for more detail.

Impact

No breaking changes

• fetchNative now has optional function parameter
• No more polling request visible on the native UI
  • visible only with f12 developers panel
  • Also, it was not so helpful for devs anyway

Additional Notes

I renewed the API key so don't worry.
Currently testing.....

Footnotes

1. Modifies the behavior of prompting, requesting, or handling responses from AI models. ↩

2. Over 80% of the code is AI generated. ↩

[cubicj]

The batch logging change looks good overall.

One unrelated change: redacted_thinking now exposes content.data only in the batch path, while the other Claude paths still render {{redacted_thinking}}. It also opens a backtick without closing it.

Could you check if that was intentional?

[tasoo-oos]

Done! I rolled it back to {{redacted_thinking}}.

Changed redacted thinking text was intentional back then, except for the typo. But come to think of it, there would be almost no practical reason to add informational message inside <thinking> block.

That was clumsy change on my side, thanks for the feedback!

[cubicj]

Verified the updated batch path. redacted_thinking now matches the regular Claude path again. Check, tests, and build pass on a trial merge with current origin/main.

Approving — thanks for the contribution!