mcp_sqlite_memory_bank memory.instructions.md

MEMORY MANAGEMENT INSTRUCTIONS (General Guidelines)

PRIME DIRECTIVE FOR MEMORY USAGE

• ALWAYS SEARCH KNOWLEDGE BASE FIRST: Before proceeding with any task or answering questions, you MUST search available memory banks for relevant context, previous decisions, project structure, and user preferences.
• You MUST use available memory systems to store all important contextual information learned during conversations.
• You MUST retrieve relevant context from memory systems at the start of each new interaction.
• Always design schemas that are explicit, discoverable, and LLM-friendly.
• ALWAYS verify data storage and handle errors gracefully.
• Use MCP tools consistently for all memory operations.

MEMORY MANAGEMENT PROTOCOL

SETUP & INITIALIZATION

1. Check if required tables exist at the beginning of each session.
2. Create missing tables with appropriate schemas before storing data.
3. When creating tables, follow these naming conventions:
   • Use snake_case for table and column names
   • Use descriptive, semantic names
   • Include appropriate primary keys and constraints

STORAGE & RETRIEVAL

1. Store information immediately after learning it.
2. Categorize information appropriately in the correct tables.
3. Before answering questions, query relevant tables for context.
4. Use specific WHERE clauses to retrieve only relevant information.
5. LEVERAGE SEMANTIC SEARCH: Use semantic search capabilities for intelligent knowledge discovery when exact matches aren't sufficient.

SCHEMA MAINTENANCE

1. Maintain consistent schemas across sessions.
2. Update existing records rather than creating duplicates.
3. Use appropriate data types for columns.
4. Create relationships between tables when appropriate.
5. Drop tables only when they are no longer needed, and document the reason for deletion.
6. Regularly review and refactor schemas to improve clarity and efficiency.

SEMANTIC SEARCH CAPABILITIES (IF AVAILABLE)

OVERVIEW

Many memory systems support intelligent semantic search using sentence-transformers for natural language queries and content discovery. This enables agents to find conceptually similar content even when exact keyword matches don't exist.