VisPile

Analyze unstructured text data with LLMs and Knowledge Graphs! 🕵️

What is VisPile?

Intelligence analysts perform sensemaking over collections of documents using various visual and analytic techniques to gain insights from large amounts of text. As data scales grow, our work explores how to leverage two AI technologies, large language models (LLMs) and knowledge graphs (KGs), in a visual text analysis tool, enhancing sensemaking and helping analysts keep pace.

Collaborating with intelligence community experts, we developed a visual analytics system called VisPile. VisPile integrates an LLM and a KG into various UI functions that assist analysts in grouping documents into piles, performing sensemaking tasks like summarization and relationship mapping on piles, and validating LLM- and KG-generated evidence.

This code accompanies the research paper:

VisPile: A Visual Analytics System for Analyzing Multiple Text Documents With Large Language Models and Knowledge Graphs
Adam Coscia, Alex Endert
The 59th Hawaii International Conference on System Sciences (HICSS), 2026
| 📖 Paper | ▶️ Live Demo | 🎞️ Demo Video | 🧑‍💻 Code |

Features

🧭 A UI for interactive LLM- and KG-driven document piling, synthesis, and analysis:

🚀 Intuitive LLM and KG operations to create and analyze piles of documents:

✅ Pile operations to explain, validate, and contextualize LLM-generated evidence:

Tutorial Video

🎞️ Watch the tutorial video for a demo of how to use VisPile here: https://youtu.be/Yql1nfkrzmU

Live Demo

🚀 To try out a live demo of VisPile, visit: https://adamcoscia.com/papers/vispile/demo/

Getting Started

🌱 If you want to customize VisPile for your own project, start here!

• Install Node.js >= v22.x and npm >= v11.x by visiting (latest release)
• Install Python v3.9.x (latest release)
• Clone this repo to your computer (instructions)

git clone git@github.com:AdamCoscia/VisPile.git

# use --depth if you don't want to download the whole commit history
git clone --depth 1 git@github.com:AdamCoscia/VisPile.git

Interface

• A frontend Vue 3 web app to visualize data in the browser.
• Additional details can be found in interface/README.md

Navigate to the interface folder:

cd interface

Install dependencies:

npm install

Then run VisPile:

npm run dev

Navigate to localhost:3000. You should see VisPile running in your browser :)

Server

• A backend Python 3 Flask app to query the OpenAI API and serve data.
• Additional details can be found in server/README.md

Navigate to the server folder:

cd server

Download the data:

• Download data.zip
• Move into the server/data directory
• Unzip the file
• You should see 3 new folders in server/data: News Articles/, embeddings/, and models/

Install dependencies:

• If you are running Windows (replace -3.9 with your Python version):

# Start a virtual environment
py -3.9 -m venv venv

# Activate the virtual environment
.\venv\Scripts\activate

# Install dependencies
python -m pip install -r requirements.txt

• If you are running MacOS / Linux (replace python3.9 with your Python version):

# Start a virtual environment
python3.9 -m venv venv

# Activate the virtual environment
source venv/bin/activate

# Install dependencies
python -m pip install -r requirements.txt

Next, create a .env file in this directory and put the following in it:

OPENAI_API_KEY=<<YOUR API KEY HERE>>

Finally, run the server:

python main.py

Credits

Led by Adam Coscia, VisPile is a result of a collaboration between visualization experts in human centered computing and interaction design from Georgia Tech. VisPile is created by Adam Coscia and Alex Endert.

Data

• The example dataset provided comes from the IEEE Visual Analytics Science and Technology (VAST) Challenge 2021, Mini-Challenge 1 (https://vast-challenge.github.io/2021/MC1.html)

Citation

To learn more about VisPile, please read our research paper (published at HICSS '26).

@inproceedings{Coscia:2026:VisPile,
  author = {Coscia, Adam and Endert, Alex},
  title = {VisPile: A Visual Analytics System for Analyzing Multiple Text Documents With Large Language Models and Knowledge Graphs},
  year = {2026},
  booktitle = {Proceedings of the 59th Hawaii International Conference on System Sciences},
  location = {Lahaina, HI, USA},
  series = {HICSS-59}
}

License

The software is available under the MIT License.

Contact

If you have any questions, feel free to open an issue or contact Adam Coscia.