Skip to content
Back to Blog

Memory · CLI · Best Practices · Doctor · Diff

Keeping Memory Healthy: vem summarize, diff, and doctor

February 10, 20265 min read

Stale memory misleads agents. Learn the three tools that keep your .vem/ directory accurate, current, and diagnosable.

Why Memory Goes Stale

Your codebase changes every day. Without maintenance, .vem/CONTEXT.md starts describing architecture that no longer exists, CURRENT_STATE.md refers to tasks that were finished weeks ago, and agents confidently give wrong answers based on outdated facts.

vem provides three focused tools to combat this: summarize detects new changes and proposes memory updates, diff shows you what has drifted from the cloud, and doctor checks the overall health of your memory setup.

vem summarize: Stay Current with Git Changes

vem summarize analyses your recent git log and compares it against your current .vem/ content. It then proposes targeted updates — new changelog entries, task status changes, and context refreshes — based only on what actually changed.

Run it after a big merge or at the start of a new sprint to quickly catch up your memory to the current state of the code.

Summarize recent changes

# Analyse the last 20 commits and propose memory updates
vem summarize

# Analyse a specific range
vem summarize --since 2026-02-01

vem diff: See What Has Drifted from the Cloud

Every vem push sends a snapshot to the cloud. vem diff compares your current local .vem/ state against the last verified cloud snapshot, showing you exactly what is out of sync.

This is especially useful before pushing — you can review the diff to make sure you are not overwriting important cloud-side changes made by a teammate.

Diff local vs cloud memory

# Compare local .vem/ to the last cloud snapshot
vem diff

# Diff a specific section
vem diff --section tasks

vem doctor: Full Health Check

vem doctor runs a comprehensive diagnostic across your entire setup: authentication status, project configuration, git integration, local .vem/ data integrity, and cloud sync status. It reports each check as pass, warn, or fail with actionable suggestions.

Make it part of your onboarding checklist for new team members and your first step whenever something feels off.

Run a full health check

# Full diagnostic
vem doctor

# Quick check (auth and project only)
vem doctor --quick

A Simple Memory Hygiene Habit

The most effective teams build a two-minute memory check into their workflow: run vem doctor at the start of a sprint to confirm everything is wired up, vem summarize after landing a significant PR, and vem diff before any vem push. This takes less time than writing a Slack update and keeps every agent — human or AI — working from accurate context.

  • vem doctor — check auth, config, git, and data integrity
  • vem summarize — propose memory updates from recent git history
  • vem diff — compare local .vem/ to the last cloud snapshot
  • vem push — publish verified, healthy memory to the cloud