SuperLocalMemory: Switch Profile

Name: slm-switch-profile
Rating: 92
Author: varun369

Switch between memory profiles to maintain separate contexts for different projects or use cases.

Usage

bash

slm switch-profile <name> [--create]

What are Profiles?

Profiles = Isolated Memory Spaces

Each profile has its own:

•✅ Separate database - Zero context bleeding
•✅ Independent knowledge graph - Project-specific relationships
•✅ Unique patterns - Different coding preferences per project
•✅ Isolated history - No cross-contamination

Think of profiles as workspaces:

•default - General use
•work - Work projects
•personal - Personal projects
•client-acme - Specific client work
•experiment - Testing/learning

Examples

Example 1: Switch to Existing Profile

bash

$ slm switch-profile work

Output:

code

🔄 Switching profile: default → work

Profile Information:
  Name: work
  Created: 2026-01-20
  Memories: 342
  Last Used: 2026-02-05 09:15

✅ Switched to profile "work"

Restart your AI tools (Cursor, ChatGPT, etc.) to use the new profile.

Example 2: Create New Profile

bash

$ slm switch-profile client-acme --create

Output:

code

🆕 Creating new profile: client-acme

Profile created successfully!
  Location: ~/.claude-memory/profiles/client-acme/
  Database: memory.db (empty)
  Status: Active

✅ Switched to profile "client-acme"

Next steps:
  • Start saving memories: slm remember "..."
  • AI tools will use this profile automatically

Example 3: List All Profiles

bash

$ slm list-profiles

Output:

code

📁 Available Profiles (4 total)

→ work (active)
  Created: 2026-01-20
  Memories: 342
  Size: 1.8 MB
  Last used: 2 minutes ago

  default
  Created: 2026-01-15
  Memories: 1,247
  Size: 4.2 MB
  Last used: 3 hours ago

  personal
  Created: 2026-01-22
  Memories: 89
  Size: 0.4 MB
  Last used: Yesterday

  client-acme
  Created: 2026-02-07
  Memories: 0
  Size: 0 KB
  Last used: Never

Example 4: Return to Default

bash

$ slm switch-profile default

Always works (default profile always exists)

Arguments

Argument	Type	Required	Description
`<name>`	string	Yes	Profile name (lowercase, hyphens allowed)
`--create`	flag	No	Create profile if doesn't exist
`--force`	flag	No	Switch even if current profile has unsaved work

Profile Naming Rules

Valid:

•Lowercase letters: a-z
•Numbers: 0-9
•Hyphens: -
•Length: 1-64 characters

Examples:

•✅ work
•✅ client-acme
•✅ project-2024
•✅ personal

Invalid:

•❌ Work (uppercase)
•❌ client_acme (underscore)
•❌ client acme (space)
•❌ client.acme (period)

Use Cases

1. Work/Personal Separation

bash

# Morning: Start work
slm switch-profile work
slm remember "Sprint planning: focus on auth system"

# Evening: Personal projects
slm switch-profile personal
slm remember "Learning Rust - ownership is tricky"

Benefit: No mixing of work and personal context

2. Multi-Client Consulting

bash

# Client A project
slm switch-profile client-a
slm remember "Client A uses AWS, prefers TypeScript"

# Client B project
slm switch-profile client-b
slm remember "Client B uses Azure, prefers C#"

Benefit: Client-specific preferences and patterns

3. Experimentation

bash

# Create throwaway profile for learning
slm switch-profile experiment --create
slm remember "Testing new framework..."

# Later: delete when done
slm delete-profile experiment

Benefit: Safe experimentation without polluting main memory

4. Project Isolation

bash

# Different tech stacks
slm switch-profile backend-api  # Python/FastAPI
slm switch-profile frontend-app  # React/TypeScript
slm switch-profile ml-pipeline   # Python/PyTorch

Benefit: Pattern learning adapts to each stack

5. Team Shared Profiles (Advanced)

bash

# Sync profile to team folder
slm export-profile work > team-shared/work-profile.db

# Team members import
slm import-profile team-shared/work-profile.db --name work

Benefit: Shared knowledge base for team

How Switching Works

Behind the Scenes

1. Profile Structure:

code

~/.claude-memory/
├── memory.db               ← Current active profile (symlink)
├── profiles/
│   ├── default/
│   │   └── memory.db       ← Default profile database
│   ├── work/
│   │   └── memory.db       ← Work profile database
│   └── personal/
│       └── memory.db       ← Personal profile database
└── current_profile         ← Text file with active profile name

2. Switching Process:

bash

$ slm switch-profile work

•Step 1: Verify "work" profile exists
•Step 2: Update symlink: memory.db → profiles/work/memory.db
•Step 3: Write "work" to current_profile file
•Step 4: Notify user to restart AI tools

3. AI Tools Read:

•MCP server reads ~/.claude-memory/memory.db
•Symlink automatically points to active profile
•Transparent to AI tools

Important Notes

Restart Required

After switching, restart your AI tools:

•Cursor: Cmd+Q, reopen
•Claude Desktop: Quit, reopen
•ChatGPT: Close, reopen
•VS Code: Reload window

Why: Tools cache database connection at startup

Profile Isolation is Complete

Each profile has:

•❌ NO access to other profiles' memories
•❌ NO shared knowledge graph
•❌ NO cross-profile pattern learning

This is by design for true isolation.

Default Profile Special

•Always exists
•Cannot be deleted
•Safe fallback

Advanced Usage

Profile Management

Create multiple profiles:

bash

slm switch-profile work --create
slm switch-profile personal --create
slm switch-profile client-a --create

Backup profile:

bash

cp -r ~/.claude-memory/profiles/work ~/.claude-memory/backups/work-2026-02-07

Delete profile:

bash

slm delete-profile experiment
# Warning: This permanently deletes all memories!

Rename profile:

bash

slm rename-profile old-name new-name

Scripting & Automation

Auto-switch by project directory:

bash

# Add to .bashrc or .zshrc
function cd() {
  builtin cd "$@"

  # Check for .slm-profile file
  if [ -f ".slm-profile" ]; then
    profile=$(cat .slm-profile)
    slm switch-profile "$profile" --quiet
  fi
}

Usage:

bash

# In project directory
echo "myapp" > .slm-profile

# CD into directory
cd ~/projects/myapp
# Automatically switches to "myapp" profile

Daily profile switching:

bash

#!/bin/bash
# cron: 0 9 * * 1-5 (Mon-Fri at 9 AM)

slm switch-profile work
echo "Switched to work profile" | notify-send "SuperLocalMemory"

Profile Statistics

Compare profiles:

bash

slm list-profiles --stats

Output:

code

Profile          Memories  Size     Nodes  Edges  Patterns
default          1,247     4.2 MB   892    2,134  34
work             342       1.8 MB   256    789    12
personal         89        0.4 MB   67     145    4

Export profile summary:

bash

slm profile-summary work > work-summary.txt

Troubleshooting

"Profile not found"

Cause: Typo or profile doesn't exist

Solution:

bash

# List existing profiles
slm list-profiles

# Create if needed
slm switch-profile myprofile --create

"Switch failed: Database locked"

Cause: AI tool still using old profile

Solution:

bash

# Kill all MCP servers
killall python3

# Try again
slm switch-profile work

# Restart AI tools

"Memories disappeared after switching"

Cause: Switched to empty/different profile (working as intended)

Solution:

bash

# Check current profile
slm status | grep "Current Profile"

# Switch back to profile with memories
slm switch-profile default

# Or list all profiles to find the right one
slm list-profiles

"AI tools still see old profile"

Cause: Didn't restart tools

Solution:

bash

# Must completely restart (not just reload)
# macOS: Cmd+Q, then reopen
# Windows/Linux: Close all windows, reopen

Performance

Operation	Time	Notes
Switch profile	~100ms	Instant
Create profile	~200ms	Fast
List profiles	~50ms	Very fast

Switching is fast because it's just updating a symlink.

Notes

•No data loss: Switching never deletes memories
•Profile persistence: Memories stay in original profile
•Tool transparency: AI tools don't need special config per profile
•Backup-friendly: Each profile is separate directory

Related Commands

•slm list-profiles - Show all profiles
•slm delete-profile <name> - Delete a profile
•slm rename-profile <old> <new> - Rename profile
•slm status - Shows current profile
•slm export-profile <name> - Export to file
•slm import-profile <file> - Import from file

Created by: Varun Pratap Bhardwaj (Solution Architect) Project: SuperLocalMemory V2 License: MIT with attribution requirements (see ATTRIBUTION.md) Repository: https://github.com/varun369/SuperLocalMemoryV2

Open source doesn't mean removing credit. Attribution must be preserved per MIT License terms.