Vedanshipanda/layer10-api

🧠 Layer10 Memory Graph

An organizational knowledge graph built from GitHub issues using LLMs — visualized as an interactive force-directed graph.

📁 Project Structure

layer10-memory-graph/
├── Dockerfile              # Backend container for Hugging Face
├── README.md               # This file
├── requirements.txt        # Python dependencies
├── src/
│   ├── main.py             # FastAPI backend
│   ├── graph.py            # Graph building logic
│   ├── extraction.py       # LLM entity extraction
│   └── ingestion.py        # Data ingestion
├── data/
│   └── memory_graph.json   # Knowledge graph (139 nodes, 81 edges)
└── ui/                     # React frontend
    ├── src/
    │   ├── App.js
    │   └── App.css
    └── package.json

🔧 Backend — FastAPI

Endpoints

Stack

• —FastAPI — REST API
• —NetworkX — graph data structure
• —Uvicorn — ASGI server
• —Docker — containerized deployment

Local Development

# Create virtual environment
python -m venv venv
venv\Scripts\activate       # Windows
source venv/bin/activate    # Mac/Linux

# Install dependencies
pip install -r requirements.txt

# Run server
uvicorn src.main:app --reload --port 8000

API docs at http://localhost:8000/docs

Deployment — Hugging Face Spaces (Docker)

1. 1.Go to huggingface.co → New Space
2. 2.Name: layer10-api · SDK: Docker
3. 3.Push repo with Dockerfile, src/, data/, README.md
4. 4.Space auto-builds and runs on port 7860

Live at: https://Vedanshipanda-layer10-api.hf.space

🖥️ Frontend — React

Features

• —Knowledge Graph — interactive force-directed graph with glowing nodes and animated edges
• —Filters — filter by entity type, relationship type, min degree, max nodes
• —Analytics — type distribution charts, top relationship types, most connected nodes table
• —Node Explorer — inspect any node's incoming and outgoing relationships
• —Query — search the graph via backend API and highlight matching nodes

Stack

• —React — frontend framework
• —react-force-graph-2d — WebGL graph rendering
• —Axios — API calls
• —Orbitron + Share Tech Mono — robotic HUD-style fonts

Local Development

cd ui
npm install
npm start

Make sure the backend is running at http://localhost:8000 first.

Deployment — Hugging Face Spaces (Static)

1. 1.Go to huggingface.co → New Space
2. 2.Name: layer10-ui · SDK: Static
3. 3.Build the frontend:

cd ui
npm run build

1. 1.Upload everything inside ui/build/ to the Space

Live at: https://Vedanshipanda-layer10-ui.hf.space

🚀 Full Stack Setup

📊 Graph Stats

• —139 nodes — Bugs, Developers, Features, Issues, Repositories, Technologies
• —81 edges — relationships extracted from GitHub issues via LLM
• —6 entity types — color-coded with neon glow in the visualization