About

Partitioned Graph Synthesis

PGS is a standalone, zero-dependency npm package for coverage-optimized querying of large knowledge graphs. It partitions your graph using Louvain community detection, sweeps each partition at full fidelity, and synthesizes cross-domain connections -- reporting not just what it found, but what's absent.

Where standard retrieval examines the nearest handful of nodes, PGS examines every node in the graph, organized by community structure. This makes it possible to detect convergence across distant clusters, identify structural gaps, and surface connections that no top-K window would reveal.

View on GitHub

COSMO -- The Research System

COSMO is the AI research system that builds the knowledge graphs PGS queries. It uses multi-agent orchestration to conduct deep research investigations, building a persistent knowledge graph with spreading-activation memory, Hebbian learning, and state-dependent topology maintenance.

COSMO runs autonomous research "brains" that accumulate knowledge over dozens of cycles, producing graphs with hundreds of nodes and thousands of edges that PGS can then query comprehensively. Each brain represents a sustained investigation into a domain -- physics, biology, engineering, policy -- where the graph grows organically as agents discover, validate, and connect findings.

Open Source

PGS Engine is MIT licensed and available on GitHub. It was extracted from COSMO 2.3 as a standalone package with zero runtime dependencies. Install it, pass in your graph and an LLM caller, and get full-coverage synthesis.

Get in Touch

For questions, feedback, or collaboration: