coder_db

Coder DB acts as a structured memory system for AI coding assistants, integrating a Qdrant vector store for semantic memory, a SQLite database for structured algorithm storage and versioning, and a Knowledge Graph to model relationships between algorithms, patterns, and concepts. The MCP server exposes tooling to store and retrieve memory items, query patterns by language or complexity, and traverse related concepts via the knowledge graph. You can leverage the Qdrant interface to save code patterns and explanations with rich metadata, use the SQLite-backed catalog to manage algorithm versions and performance data, and explore the knowledge graph to discover related patterns, problems, and solutions. Typical workflows include querying for code patterns related to a task, inspecting version histories of algorithms, and navigating related concepts to identify alternative approaches.

Prerequisites:

• Node.js (14+)
• Access to a running Qdrant instance (or a container image) on localhost:6333
• SQLite writable path (or container with mounted volume)
• Access to a Knowledge Graph (Neo4j) at localhost:7687 with credentials

1. Clone the MCP server repository (or use your existing MCP workspace): git clone https://github.com/angrysky56/coder_db.git cd coder_db

2. Install Node.js dependencies: npm install

3. Configure environment variables (example):

   • Create a .env file or export variables in your shell: QDRANT_HOST=localhost QDRANT_PORT=6333 SQLITE_PATH=./data/sqlite.db KG_GRAPH_DB=neo4j://localhost:7687 KG_USER=neo4j KG_PASSWORD=password LOG_LEVEL=info

4. Start the MCP server: npm start

5. Verify the server is running and accessible (default port depends on your setup). You can begin by storing a memory item in Qdrant or querying for related patterns via the provided CLI or API endpoints.

6. Optional: Initialize and seed the SQLite database and knowledge graph with initial schemas and sample data per your deployment strategy.

Tips and considerations:

• Ensure Qdrant is reachable before starting; connectivity issues will prevent memory storage if the vector store is unavailable.
• Keep environmental credentials for the knowledge graph secure; prefer using a secrets manager in production.
• Use controlled vocabularies for tags to improve retrieval quality in Qdrant.
• Regularly back up the SQLite database and the knowledge graph data.
• If you update schemas (algorithms, categories, or metrics), document changes in the change_logs section of SQLite to preserve version history.
• For performance tuning, monitor Qdrant response times and SQLite query plans; optimize indices and vocabulary terms as needed.