🧠 Linguist-Core: Sovereign Distributed Knowledge Graph

Linguist-Core is a high-performance, edge-first Knowledge Graph and RAG (Retrieval-Augmented Generation) engine designed for the AMD Slingshot 2026. It allows users to ingest complex documents, extract semantic relationships in real-time, and synchronize that knowledge across a distributed peer-to-peer network—all running locally on AMD hardware.

Important

This system is designed for Sovereign AI. No cloud data, no external APIs. Every computation happens on your silicon.

---

🚀 Key Features

• Fast Semantic Extraction: Optimized regex-based relationship extraction that captures technical facts instantly.
• Edge-First GraphRAG: Local Graph Traversal combined with flan-t5 for precise, multi-hop question answering.
• Distributed Infinity Fabric Sync: ZeroMQ-based P2P layer that broadcasts knowledge nodes across your LAN.
• Cross-Platform Ready: Fully compatible with Windows, Linux, and macOS.
• AMD Hardware Optimized: Designed for Ryzen AI NPUs and Radeon GPUs via ROCm.

---

🏗 Architecture

1. Ingestion Layer: Parses PDF/DOCX/TXT and splits them into semantic chunks.
2. Extraction Engine: Identifies Entities and meaningful Verbs (e.g., leverages, requires, supports).
3. Local Graph Store: A persistent NetworkX-based multi-directed graph with entity embeddings.
4. Sync Layer: A publisher-subscriber model that replicates graph topology between peers.
5. Query Engine: Performs semantic search + graph BFS to build context for the local RAG LLM.

---

💻 Installation

Prerequisites

• Python 3.9+
• pip and venv

1. Clone & Setup

git clone https://github.com/GiGiKoneti/AMDss.git
cd AMDss
python3 -m venv venv

2. Install Dependencies

Windows (PowerShell):

.\venv\Scripts\activate
pip install -r requirements.txt

macOS / Linux (Terminal):

source venv/bin/activate
pip install -r requirements.txt

---

🏃 Running the System

To experience the Distributed Sync, you should run the system on two separate machines (System A and System B).

Step 1: Start the API Backend

Identify the IP address of your peer machine.

System A (at 192.168.0.107):

export PEER_IPS="192.168.0.112"
python -m linguist_core.api_server

System B (at 192.168.0.112):

export PEER_IPS="192.168.0.107"
python -m linguist_core.api_server

Step 2: Launch the UI

In a new terminal window (with venv active):

python -m linguist_core.ui_app

Access the UI at: http://localhost:7860

---

🛠 Advanced Configuration

Toggling Extraction Mode

In linguist_core/extractor.py, you can toggle between Fast Mode (Default) and LLM Mode (Experimental):

• Fast Mode: Immediate results using semantic regex.
• LLM Mode: High-accuracy extraction using flan-t5 (Requires GPU/NPU for performance).

---

📄 License

Tailored for the AMD Slingshot Hackathon 2026. Built with ❤️ for Sovereign AI.