Suppress cross-channel transcript echo

[ComputelessComputer]

Add phrase-based live transcript echo suppression across mic and speaker channels, including partial and held-tail handling.

---

Note

Medium Risk
Changes real-time transcript output in the listener path; wrong heuristics could drop legitimate simultaneous speech, though tests and distinct-text cases mitigate that.

Overview
Adds cross-channel echo suppression in LiveTranscriptEngine so mic/speaker bleed does not duplicate the live transcript.

After normalization and TranscriptProcessor, a new CrossChannelEchoSuppressor filters new_words and partials by matching normalized tokens on different channels within a time window (≥3-token runs), plus echo-tail chaining for words that continue a recently suppressed duplicate. Kept finals extend rolling history (30s prune); distinct overlapping speech on another channel is left alone.

process and flush both run the suppressor and skip emitting an update when the delta is empty and there is no segment change. Unit tests cover final/partial echo, tail chaining after interrupted runs, and non-echo cross-channel text.

^{Reviewed by Cursor Bugbot for commit eb044c2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[netlify]

✅ Deploy Preview for old-char canceled.

Name	Link
🔨 Latest commit	eb044c2
🔍 Latest deploy log	https://app.netlify.com/projects/old-char/deploys/6a4627620b3de4000846db00

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 19c913f. Configure here.}