fix(converters): strip Claude Code 2.1.x billing attribution from sys…

[liwenquantop-dot]

Problem

Claude Code 2.1.x prepends a per-request system text block of the form:

x-anthropic-billing-header: cc_version=2.1.153.d02; cc_entrypoint=sdk-cli; cch=<5hex>;

The cch token is a fresh random hex on every request. Because extract_system_prompt() concatenates all system text blocks into a single string before forwarding to Kiro, this random prefix poisons the prompt seen by upstream and invalidates any prefix-keyed prompt cache. (The attribution header is meant to be an HTTP header but ships in the request body for claude -p / SDK mode.)

Solution

Add _strip_billing_attribution() and call it from extract_system_prompt() for both string and list system formats. Pure-attribution blocks are dropped; mixed blocks have the leading attribution line removed. Behavior is gated by STRIP_BILLING_HEADER (default true) so it can be disabled for A/B comparison.

Empirical validation (local)

Replay benchmark using real captured Claude Code body (212KB, 3 system blocks), N=30 interleaved, randomized cch per request:

Metric	Baseline (no strip)	Patched (strip)	Δ
TTFB mean	4813ms	4238ms	-574ms (-11.9%)
TTFB median	4506ms	3952ms	-554ms (-12.3%)
TTFB p95	6263ms	4958ms	-1305ms (-20.8%)

Byte-level check via DEBUG_MODE=all: incoming body contains cch=<hex> and x-anthropic-billing-header; outgoing Kiro payload has zero matches for either string.

Alternative

CLAUDE_CODE_ATTRIBUTION_HEADER=0 set in the client environment achieves the same effect on the Claude Code side. The gateway-level fix is complementary and protects users / clients that don't (or can't) set the env var.

Files

• kiro/config.py: add STRIP_BILLING_HEADER env (default true)
• kiro/converters_anthropic.py: add _strip_billing_attribution(), apply in extract_system_prompt()

[cla-bot]

Thanks for the PR! 🎉

Before merge, we need a one-time CLA confirmation.
It confirms that you have the right to contribute this code and allow the project to use it.

Full CLA text:
https://github.com/jwadow/kiro-gateway/blob/main/CLA.md

Please reply once with:

I have read the CLA and I accept its terms

You need to write once, all further messages from me can be ignored.

[liwenquantop-dot]

I have read the CLA and I accept its terms

[cla-bot]

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: liwenquan.
This is most likely caused by a git client misconfiguration; please make sure to:

1. check if your git client is configured with an email to sign commits git config --list | grep email
2. If not, set it up using git config --global user.email email@example.com
3. Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

[cla-bot]

Thanks for the PR! 🎉

Before merge, we need a one-time CLA confirmation.
It confirms that you have the right to contribute this code and allow the project to use it.

Full CLA text:
https://github.com/jwadow/kiro-gateway/blob/main/CLA.md

Please reply once with:

I have read the CLA and I accept its terms

You need to write once, all further messages from me can be ignored.

[liwenquantop-dot]

@jwadow Please add liwenquantop-dot to the .clabot contributors list.
I have read the CLA and accept its terms.