Document the AI sync conflict classifier confidence threshold

[venkatapgummadi]

Background

ASCEND Day-1 Action Sheet — Page 6
ConflictClassifier accepts a confidence_threshold parameter (default 0.85) but the
documentation doesn't explain when to tune it.

What needs to be done

Add a section to docs/api-reference.md covering: what the threshold means, how to choose a
value, and how to retrain the classifier on organization-specific data.

Files to look at

• ai-sync/ascend_sync/conflict_classifier.py
• docs/api-reference.md

Definition of done

• New "Tuning the conflict classifier" section added
• Three example values explained (0.7, 0.85, 0.95)
• Cross-link from README and FAQ
  Estimated effort: 1–2 hours · Skills needed: Technical writing, basic ML literacy