RAGU introduces multi-step GraphRAG engine with compact Meno-Lite-0.1 extractor

RAGU is an open-source modular GraphRAG engine that improves structured knowledge integration by separating entity extraction from consolidation. It utilizes a two-stage typed extraction process, DBSCAN-backed deduplication, LLM summarization, and Leiden community detection to reduce noise and brittleness found in single-pass systems.

• The system employs Meno-Lite-0.1, a 7B model optimized for language skills, which outperforms Qwen2.5-32B on knowledge-graph construction by +12.5% relative harmonic mean.
• On GraphRAG-Bench (Medical), RAGU achieves evidence recall up to 0.84 and overtakes HippoRAG2 on synthesis tasks.
• The engine is installable via pip, runs on a single GPU, and is released under the MIT license.

This approach demonstrates that compact models can effectively handle in-pipeline language skills like comprehension and reasoning, offering a resource-efficient alternative to larger models for GraphRAG tasks.