[markbackman]

SkipTagsAggregator keeps text between a start and an end tag in one aggregation, so a sentence boundary inside the tags doesn't split it. It finds the tags with parse_start_end_tags, which counted start and end tags in the text instead of finding where each one is. That missed two cases. This scans the text by position instead.

• Several tag pairs. The count loop returned on the first pair in the list, whatever it found, so only the first pair was recognized. With [("<spell>", "</spell>"), ("<card>", "</card>")], a <card>…</card> holding a period was split at it. Now the earliest start tag of any pair opens, and that pair's end tag closes it.
• A pair whose start and end are the same. With ("**", "**"), the start and end counts were always equal, so the pair never opened. Now the first ** opens it and the next closes it.
• The returned index. It's still where scanning resumes on the next call, which is how SkipTagsAggregator, its only caller, uses it. It's now just past an open tag's start, or past the last tag that closed, rather than the end of the text.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/pipecat/utils/string.py	89.38% <100.00%> (-0.10%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[aconchillo]

Makes sense. LGTM