NemoClaw Knowledge Wiki

EdgeQuake: Local Rust Graph-RAG with Ollama for Improved Knowledge Retrieval

3 min read

🗂️ Tools, Platforms & Infrastructure · View mindmap

Generated: 2026-05-22 · API: Gemini 2.5 Flash · Modes: Summary

EdgeQuake: Local Rust Graph-RAG with Ollama for Improved Knowledge Retrieval

Clip title: EdgeQuake - 100% Local with Ollama: Fixes Broken RAG Author / channel: Fahd Mirza URL: https://www.youtube.com/watch?v=kkSVZfGzHyQ

Summary

The video introduces EdgeQuake, a high-performance Graph-RAG (Retrieval Augmented Generation) framework developed in Rust, aiming to address the limitations of conventional RAG systems. The presenter argues that many existing RAG tools are fundamentally flawed, often providing unhelpful or “garbage” responses to complex queries because they primarily rely on finding similar text snippets without understanding the underlying relationships between ideas or concepts. EdgeQuake seeks to overcome this by building an “actual knowledge graph” from documents, allowing for more sophisticated reasoning and superior retrieval.

EdgeQuake’s core functionality involves processing user-uploaded documents by first chunking them, then extracting entities and their relationships using a local Large Language Model (LLM). This structured information is then stored as a traversable graph in a PostgreSQL database, with embeddings managed by PGVector. The system is designed to run entirely locally, offering an open-source solution that avoids expensive proprietary APIs. The architecture consists of a React 19 + TypeScript frontend, a Rust-based backend API, and integrates with local LLM providers like Ollama, supporting various inference and embedding models.

The demonstration showcases EdgeQuake’s installation and usage. The presenter sets up the system using a quickstart Docker compose script on an Ubuntu server, configuring it to use Ollama for LLM inference and embeddings. After resolving an initial networking challenge that prevented the Dockerized application from reaching the local Ollama daemon, a personal biography text file is successfully uploaded. EdgeQuake processes this document, extracting 37 entities and visualizing them in an interactive knowledge graph. When queried about the tools and technologies used by the speaker, the system leverages this graph to provide a comprehensive, structured, and context-aware answer, highlighting its ability to derive meaningful insights beyond simple keyword matching.

In conclusion, EdgeQuake presents a promising approach to RAG by shifting the paradigm from purely vector-based similarity search to a more intelligent, graph-aware retrieval mechanism. Its ability to create and query knowledge graphs from custom documents locally and open-source offers a powerful tool for enhanced understanding and reasoning over complex data. While minor setup intricacies related to local model integration might be present, the framework’s fundamental design provides a significant step forward in addressing the contextual shortcomings of traditional RAG pipelines.

Video Description & Links

Description

This video installs EdgeQuake with local ollama model. It’s GraphRAG inspired from LightRag written in Rust.

🔥 Buy Me a Coffee to support the channel: https://ko-fi.com/fahdmirza

edgequake

PLEASE FOLLOW ME: ▶ LinkedIn: https://www.linkedin.com/in/fahdmirza/ ▶ YouTube: https://www.youtube.com/@fahdmirza ▶ Blog: https://www.fahdmirza.com

RESOURCES:

▶ https://github.com/raphaelmansuy/edgequake

All rights reserved © Fahd Mirza

URLs

Graph View

Backlinks