SuperLocalMemory: Remember

Save content to your local memory system with automatic indexing, knowledge graph integration, and pattern learning.

Usage

slm remember "<content>" [--tags tag1,tag2] [--project name] [--importance 1-10]

Examples

Example 1: Basic Memory

slm remember "We use FastAPI for REST APIs"

What happens:

• •Content saved to SQLite database
• •TF-IDF vectors generated for semantic search
• •Entities extracted and added to knowledge graph
• •Pattern learning analyzes for coding preferences
• •Memory ID returned (e.g., 42)

Example 2: With Tags

slm remember "JWT tokens expire after 24 hours" --tags security,auth,jwt

Tags help with:

• •Organization
• •Filtering
• •Related memory discovery

Example 3: With Project

slm remember "Database uses PostgreSQL 15 with UUID primary keys" --project myapp --tags database,postgresql

Project isolation:

• •Separate memories per project
• •Switch profiles with slm switch-profile
• •No context bleeding

Example 4: Important Memory

slm remember "CRITICAL: Production deploy requires approval from @lead" --importance 10 --tags deployment,production

Importance (1-10):

• •1-3: Low priority (notes, ideas)
• •4-6: Normal (coding patterns, decisions)
• •7-9: High priority (critical info, warnings)
• •10: Critical (blockers, security issues)

Arguments

Output

Memory added with ID: 42

✅ Memory saved successfully

Next steps:
  • Use `slm recall <query>` to search this memory
  • Use `slm list` to see recent memories

What Happens Behind the Scenes

1. •Content Storage: Saved to SQLite (~/.claude-memory/memory.db)
2. •Semantic Indexing: TF-IDF vectors generated for similarity search
3. •Knowledge Graph: Entities extracted and nodes/edges created
4. •Pattern Learning: Analyzes content for coding preferences (frameworks, style, testing)
5. •Full-Text Index: FTS5 index updated for fast keyword search
6. •Timestamp: Created timestamp recorded

Advanced Usage

Natural Language (in AI chat)

Most AI assistants will automatically invoke this skill when you say:

• •"Remember that..."
• •"Save this for later..."
• •"I want to store..."
• •"Keep track of..."

Example in Cursor/Claude:

You: "Remember that we decided to use React hooks over class components"
AI: [Automatically invokes slm-remember skill]
✓ Memory saved

Bulk Import

Save multiple memories from a file:

# From text file (one memory per line)
while IFS= read -r line; do
  slm remember "$line" --project bulk-import
done < memories.txt

# From CSV (content,tags,project)
while IFS=',' read -r content tags project; do
  slm remember "$content" --tags "$tags" --project "$project"
done < memories.csv

Integration with Git Hooks

Pre-commit hook (save commit messages):

#!/bin/bash
# .git/hooks/post-commit

commit_msg=$(git log -1 --pretty=%B)
commit_hash=$(git log -1 --pretty=%H)

slm remember "Commit: $commit_msg (${commit_hash:0:7})" \
  --tags git,commit \
  --project "$(basename $(git rev-parse --show-toplevel))"

Error Handling

Notes

• •100% local: Nothing leaves your machine
• •Cross-tool sync: All AI tools access same database (Cursor, ChatGPT, Claude, etc.)
• •Unlimited: No memory limits, no quotas
• •Privacy: Your data stays on your computer
• •Profiles: Use slm switch-profile for project isolation

Related Commands

• •slm recall "<query>" - Search memories semantically
• •slm list - List recent memories
• •slm status - Check system health
• •slm build-graph - Rebuild knowledge graph
• •slm switch-profile <name> - Switch memory profile

Technical Details

Database Schema:

• •Table: memories
• •Fields: id, content, tags, project_name, importance, created_at, etc.
• •Indexes: Full-text search (FTS5), TF-IDF vectors, timestamps

Performance:

• •Add memory: ~50ms
• •With knowledge graph: ~300ms
• •Large content (10KB): ~1s

Limits:

• •Max content size: 1MB
• •Max tags: 50 per memory
• •Max project name: 64 characters

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.