carrasquelalex1/hipocampo
🔌 MCP Servercarrasquelalex1
A dual-memory MCP server using PostgreSQL and pgvector for advanced sparse selective caching and retrieval.
Hipocampo represents a significant advancement in AI memory systems by integrating structured relational databases with vector search capabilities. At its core, it utilizes PostgreSQL 17 and the pgvector extension to manage long-term memory, while incorporating NVIDIA NIM for high-performance inference tasks. The standout feature is its Sparse Selective Caching (SSC) mechanism, which employs a 4-phase progressive retrieval strategy to ensure that the most relevant context is surfaced efficiently during AI interactions.
The server provides a comprehensive suite of 11 tools designed for operational excellence: search, save, profile, health, auto-repair, stats, tune, dedup, checkpoint, and maintenance. These tools allow developers to maintain data integrity, optimize retrieval performance, and manage memory states effectively. By supporting bilingual inputs (English/Spanish), Hipocampo is highly versatile for global applications. It is designed to be easily deployed via a Python-based server script, providing a plug-and-play memory layer for any MCP-compliant AI agent or application.
💡Highlights
- ├─PostgreSQL 17 & pgvector backend
- ├─Sparse Selective Caching (SSC)
- └─11 specialized memory tools
🎯For
- ├─AI Engineers
- └─Backend Developers