Persistent memory and context-aware agents are the hard part of personal AI — retrieval has to be fast enough to feel invisible and selective enough not to be noise.
Relevant: GraphRAG demo — entity extraction (spaCy + Claude), NetworkX DiGraph, hybrid BM25+RRF retrieval, fact-checking pipeline. Purpose-built for context that needs structure, not just semantic similarity. GitHub: github.com/ChunkyTortoise/graphrag-demo. Also: MCP Server Toolkit (PyPI published, 233 tests) — memory and context tooling for AI systems.
Stack match: Python, PostgreSQL, RAG, LLM APIs (Anthropic), pgvector — available within 1 week, 35-40 hrs/week.
Persistent memory and context-aware agents are the hard part of personal AI — retrieval has to be fast enough to feel invisible and selective enough not to be noise.
Relevant: GraphRAG demo — entity extraction (spaCy + Claude), NetworkX DiGraph, hybrid BM25+RRF retrieval, fact-checking pipeline. Purpose-built for context that needs structure, not just semantic similarity. GitHub: github.com/ChunkyTortoise/graphrag-demo. Also: MCP Server Toolkit (PyPI published, 233 tests) — memory and context tooling for AI systems.
Stack match: Python, PostgreSQL, RAG, LLM APIs (Anthropic), pgvector — available within 1 week, 35-40 hrs/week.
Cayman | caymanroden@gmail.com | github.com/ChunkyTortoise