[eli-r-ph]

💡 Motivation and Context

Capture v1 now keeps the first event for each uuid in a request and drops the later copies, instead of rejecting the whole batch (PostHog/posthog#122305). It reports the dropped copies as duplicates_dropped: n on that uuid's result. Without this change the SDK counts those copies as delivered with the kept copy's verdict, and resends them when that verdict is retry.

• The first event with a uuid takes the result and details as before. The next n copies are reported in CaptureError.drops as (uuid, "duplicate_event_uuid") and never resent.
• verdict_summary() counts each drop by its own reason, so a dropped copy no longer shows the kept copy's verdict.
• Responses without duplicates_dropped behave as before, and a value that is not a positive integer is ignored.
• The 8.0 migration guide and a changeset describe the new behavior.

💚 How did you test it?

• Parse tests for duplicates_dropped, with string, bool, negative, zero, float and null values ignored.
• One parameterized send test over the kept copy's verdict (ok, warning, drop, retry then ok, and a copy past the count) checks drops, resent uuids, event_results and verdict_summary().
• A loss-message case with a duplicate drop next to a quota drop.
• The new tests fail with the duplicate handling or the verdict_summary() fix reverted.
• pytest posthog/test, ruff check, ruff format --check, mypy and make public_api_check pass locally.

📝 Checklist

• I reviewed the submitted code.
• I added tests to verify the changes.
• I updated the docs if needed.
• No breaking change or entry added to the changelog.

If releasing new changes

• Ran sampo add to generate a changeset file

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Written with Claude Code (Claude Opus 5.5). Copies past n keep the existing behavior and take the kept copy's verdict. The private parse result carries the counts, so the public CaptureEventResult is unchanged.

🤖 Generated with Claude Code

[marandaneto]

The code looks good; no qualifying findings. Verdict: correct.