Configuration

LoreHub can be configured through environment variables and a configuration file.

Configuration File

Location: ~/.lorehub/config.json

{
  "searchMode": "hybrid",
  "semanticSearchThreshold": 0.5,
  "defaultConfidence": 80
}

Available Settings

Environment Variables

All configuration can also be set via environment variables:

# Add to ~/.bashrc or ~/.zshrc
export LOREHUB_DB="$HOME/.lorehub/lorehub.db"
export LOREHUB_SEARCH_MODE="semantic"

# Add to ~/.config/fish/config.fish
set -x LOREHUB_DB "$HOME/.lorehub/lorehub.db"
set -x LOREHUB_SEARCH_MODE "semantic"

# Add to $PROFILE
$env:LOREHUB_DB = "$HOME\.lorehub\lorehub.db"
$env:LOREHUB_SEARCH_MODE = "semantic"

Search Configuration

Search Modes

1. Literal - Traditional substring matching

   • Fast and precise
   • Good for finding specific terms
   • Case-insensitive

2. Semantic - AI-powered conceptual search

   • Finds related concepts
   • Uses embeddings for similarity
   • Best for natural language queries

3. Hybrid - Combined approach (default)

   • 30% weight on literal matches
   • 70% weight on semantic similarity
   • Best overall results

Semantic Search Threshold

The semanticSearchThreshold controls how similar a lore must be to appear in semantic search results:

• 0.0 - Show all results (not recommended)
• 0.5 - Moderate similarity (default)
• 0.7 - High similarity
• 1.0 - Exact matches only

File Locations

LoreHub stores all data locally:

~/.lorehub/
├── lorehub.db          # SQLite database with all lores
├── config.json         # User configuration
└── exports/            # Export directory

Database Schema

The SQLite database contains:

1. realms - Tracked codebases
2. lores - Knowledge entries
3. lore_relations - Relationships between lores
4. lores_vec - Embeddings for semantic search
5. change_log - Change tracking for sync

MCP Server Configuration

When using LoreHub with AI assistants like Claude Desktop, configure the MCP server:

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "lorehub": {
      "command": "node",
      "args": ["/path/to/global/node_modules/lorehub/dist/mcp/index.js"]
    }
  }
}

To find the correct path:

npm list -g lorehub

Performance Tuning

Embeddings

LoreHub uses the all-MiniLM-L6-v2 model for semantic search embeddings. This provides a good balance of:

• Size: ~80MB model
• Speed: Fast inference
• Quality: Good semantic understanding

To regenerate embeddings after changes:

lh migrate-embeddings --force

Database Optimization

SQLite is configured with:

• WAL mode for better concurrency
• Memory-mapped I/O for performance
• Full-text search indexes

Troubleshooting

Common Issues

1. Semantic search not working

   # Generate missing embeddings
   lh migrate-embeddings

2. Database locked errors

   • Ensure only one LoreHub process is running
   • Check file permissions: ls -la ~/.lorehub/

3. MCP connection issues

   • Verify the path in Claude Desktop config
   • Check that LoreHub is installed globally
   • Restart Claude Desktop

Debug Mode

Enable debug logging by setting:

export DEBUG=lorehub:*

This will show detailed logs for troubleshooting.

Backup and Restore

Backup

Regular backups are recommended:

# Export all lores
lh export --output backup-$(date +%Y%m%d).json
 
# Or backup the entire database
cp ~/.lorehub/lorehub.db ~/.lorehub/lorehub.db.backup

Restore

# From export
lh import backup-20240115.json
 
# From database backup
cp ~/.lorehub/lorehub.db.backup ~/.lorehub/lorehub.db

Best Practices