doobidoo/mcp-memory-service v8.27.1

v8.27.1 - Critical Timestamp Preservation Hotfix

on GitHub

Critical Timestamp Regression Fix

This is a critical hotfix addressing a timestamp corruption bug that affected v8.25.0-v8.27.0.

Problem

The bidirectional sync and drift detection features incorrectly reset created_at timestamps to the current time during metadata updates. This caused all synced memories to appear "just created", destroying historical timestamp data.

Impact: 2,174+ memories in production databases had timestamps corrupted. All hybrid backend users on v8.25.0-v8.27.0 are affected.

Root Cause

The preserve_timestamps=False parameter in update_memory_metadata() reset both created_at and updated_at timestamps, when it should only update updated_at.

Solution

• Modified update_memory_metadata() to always preserve created_at from source memory during sync
• Hybrid storage now passes all 4 timestamp fields during drift detection
• Cloudflare storage updated for consistent timestamp handling across backends

Files Modified:

• src/mcp_memory_service/storage/sqlite_vec.py (lines 1389-1406)
• src/mcp_memory_service/storage/hybrid.py (lines 625-637, 935-947)
• src/mcp_memory_service/storage/cloudflare.py (lines 833-864)

Tests Added:

• tests/test_timestamp_preservation.py - 7 comprehensive regression tests

Recovery Tools

If your database was affected (hybrid backend, v8.25.0-v8.27.0), use these recovery tools:

1. Validate Timestamp Integrity

python scripts/validation/validate_timestamp_integrity.py

This checks for:

• Impossible timestamps (created_at > updated_at)
• Suspicious timestamp clusters (bulk reset indicators)
• Timestamp distribution anomalies

2. Recover Corrupted Timestamps

# Preview recovery (dry-run)
python scripts/maintenance/recover_timestamps_from_cloudflare.py --dry-run

# Apply recovery
python scripts/maintenance/recover_timestamps_from_cloudflare.py --apply

This restores correct created_at timestamps from Cloudflare backups.

Upgrade Instructions

For all hybrid backend users:

1. Upgrade immediately:

   pip install --upgrade mcp-memory-service
   # or
   uv pip install --upgrade mcp-memory-service

2. Validate your database:

   python scripts/validation/validate_timestamp_integrity.py

3. If validation shows issues, run recovery:

   # Preview first
   python scripts/maintenance/recover_timestamps_from_cloudflare.py --dry-run
   
   # If preview looks good, apply
   python scripts/maintenance/recover_timestamps_from_cloudflare.py --apply

4. Re-validate after recovery:

   python scripts/validation/validate_timestamp_integrity.py

Changed Semantics

The preserve_timestamps parameter now has clearer behavior:

• preserve_timestamps=True (default): Only updates updated_at to current time, always preserves created_at
• preserve_timestamps=False: Uses timestamps from updates dict if provided, always preserves existing created_at
• created_at is never reset to current time (this was the bug)

What's Fixed

• Timestamp preservation during hybrid sync operations
• Drift detection no longer corrupts created_at timestamps
• Bidirectional sync preserves historical timestamp data
• Cloudflare ↔ SQLite-vec sync maintains timestamp integrity
• Comprehensive regression tests prevent future issues

Affected Versions

• v8.25.0: Introduced drift detection with timestamp bug
• v8.27.0: Enhanced bidirectional sync amplified the issue

For More Information

• See CHANGELOG.md for full technical details
• Test suite: tests/test_timestamp_preservation.py
• Recovery tools in scripts/maintenance/ and scripts/validation/

This is a critical update. All hybrid backend users should upgrade immediately and run validation tools.

Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude noreply@anthropic.com