LinkSeek: Agentic Graph Analysis App

Overview

LinkSeek is an agentic graph analysis application built for the ArangoDB Hackathon. It enables users to explore and analyze complex event relationships using GDELT data, stored and queried in ArangoDB. The system is built with Jupyter Notebook, Gradio, and Flask, offering an intuitive way to interact with graph data through natural language queries.

Features

• Graph-Based Querying: Retrieve event insights using GraphRAG and ArangoDB queries.
• Natural Language Processing: Convert user queries into meaningful graph operations.
• Interactive Visualizations: Explore graphs dynamically using Cosmograph.
• Multiple Hosting Options: Deploy on AWS EC2, Hugging Face, GCP, or locally.

Tech Stack

• Database: ArangoDB (Graph database for event storage and querying)
• Backend: Flask (or Ngrok for tunneling local deployments)
• Frontend: Gradio (for easy UI exposure)
• Notebook: Jupyter (Main development environment)
• Visualization: Cosmograph (Graph visualization tool)
• Cloud Deployment: AWS, GCP, Hugging Face Spaces

---

Setup Guide

1. Clone the Repository

git clone https://github.com/Daenarys8/linkseek.git
cd linkseek

2. Install Dependencies

Ensure you have Python 3.8+ installed, then install dependencies: N.B: requirements.txt was generated from google colab.

pip install -r requirements.txt

You can also install the dependencies from the notebook.

3. Configure Environment Variables

Create a .env file using the provided .env.example as a reference.

cp .env.example .env

Modify the .env file to include: Only add ngrok if you'd like to use it to expose the flask app.

AWS_ACCESS_KEY_ID=xxxxxxxxxxxxxxxx
AWS_SECRET_ACCESS_KEY=xxxxxxxxxxxxxxxxxxxxxxx
AWS_REGION=xxxxxxxxxx
ARANGO_HOST=xxxxxxxxxxxxxxxxxxxxxxxxxxx
ARANGO_USER=xxxxxxxxxxx
ARANGO_PASS=xxxxxxxxxxxxxxxxxxxxxxx
ARANGO_DB=xxxxxxxxxx
NGROK_AUTH_TOKEN=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

4. Set Up ArangoDB

Local Setup:

• Install ArangoDB locally.
• Start the database and create a collection for GDELT data.

Cloud Setup:

• Use ArangoDB Oasis for a managed instance.
• Ensure the credentials in .env match the database setup.

5. Load the GDELT Dataset

Open the notebook to ingest the GDELT dataset into ArangoDB or use arango-datasets to load your graph:

6. Start the Flask API

Execute the ingestion cells in ArangoHackathon.ipynb to populate the database.

Deployment Options

1. AWS EC2

• Launch an Ubuntu EC2 instance.
• Install Python, ArangoDB, and required libraries.
• Use gunicorn or nohup to keep the Flask app running.

2. Hugging Face Spaces

• Create a new Gradio Space.
• Upload the repository and configure environment variables.
• Deploy and access the app via the Hugging Face UI.

3. Google Cloud Platform (GCP)

• Deploy using Google Compute Engine (GCE).
• Use a VM instance with Python, Flask, and ArangoDB installed.
• Expose the Gradio UI via a public IP.

---

Usage

1. Run Queries

   • Enter a natural language query in the Gradio UI.
   • Example: "Show all political events in 2015."
   • The app processes the query, fetches relevant graph data, and visualizes it.

2. Explore Graphs

   • Use Cosmograph for dynamic visualization.
   • Hover over nodes to view relationships and connections.

3. Extend & Customize

   • Modify the agent logic in the notebook to support additional query types.
   • Extend visualization options using available Plotly or NetworkX functions in the notebookfor smaller graph.

---

Future Enhancements

• Real-time GDELT Integration for continuous event updates.
• Advanced Graph Algorithms to improve query accuracy.
• Multi-Cloud Deployments for better scalability.

---

License

This project is open-source under the MIT License. Cosmograph is used under the CC BY-NC 4.0 license.

---

Credits

• Graph Database & Infrastructure: ArangoDB Community
• UI/UX Design: Freepik (Image Assets), NoCrypt (Miko Gradio Theme)
• Dataset: GDELT Project

For any issues or contributions, please submit a pull request or open an issue in the repository.