edgequake: what it is, what problem it solves & why it's gaining traction

What it solves

EdgeQuake is a high-performance Graph-RAG framework designed to overcome the limitations of traditional RAG systems. While standard vector-based retrieval often fails at multi-hop reasoning and thematic queries because it loses structural relationships between concepts, EdgeQuake transforms documents into intelligent knowledge graphs. This allows the system to combine semantic similarity with graph traversal to answer complex questions about how different entities relate to one another.

How it works

Built in Rust for maximum concurrency and performance, EdgeQuake implements the LightRAG algorithm. It decomposes documents into a knowledge graph of entities and relationships using LLM-powered extraction. At query time, it utilizes six different query modes (Naive, Local, Global, Hybrid, Mix, and Bypass) to traverse both the vector space and the graph structure. It uses PostgreSQL AGE for graph storage and pgvector for embeddings, and includes a specialized PDF pipeline that can use multimodal LLMs to read pages as images for complex layouts.

Who it’s for

Developers and organizations needing a production-ready RAG system that can handle complex reasoning, multi-column PDFs, and high concurrent user loads. It is also suitable for those who want to integrate knowledge graphs into AI agents via the Model Context Protocol (MCP).

Highlights

• Graph-RAG Architecture: Combines vector search with knowledge graph traversal for superior reasoning.
• Rust Performance: Async-first Tokio architecture with zero-copy operations for high throughput.
• Multimodal PDF Pipeline: Supports vision-capable LLMs (GPT-4o, Claude, Gemini) to process scanned documents and complex tables.
• Six Query Modes: Offers flexible retrieval strategies ranging from fast naive vector search to comprehensive hybrid graph queries.
• Knowledge Injection: Allows the injection of domain glossaries and synonyms to expand queries automatically.
• Custom Entity Configuration: Supports up to 50 domain-specific entity types per workspace.
• Agent Integration: Compatible with the Model Context Protocol (MCP) for use with AI agents.