CoQuest: Agent LLM for Research Question Co-Creation

This is the code for the system introduced in SIGCHI 2024 paper "CoQuest: Exploring Research Question Co-Creation with an LLM-based Agent".

About The Project

We proposed a novel system called CoQuest, which allows an AI agent to initiate research question (RQ) generation by tapping the power of LLMs and taking humans' feedback into a co-creation process.

Major features of the CoQuest system:

• RQ Flow Editor that facilitates a user’s major interactions, such as generating RQs, providing input and feedback to AI, and editing the RQ flow (e.g., drag and delete).
• Paper Graph Visualizer that displays the literature space related to each RQ.
• AI Thoughts that explains AI’s rationale of why each RQ is generated.

(back to top)

Demo Video

coquest-demo-github.mp4

Built With

We thankfully acknowledge the following projects/libraries based on the which this prototype is made possible.

• 
• 
• xyflow: https://github.com/xyflow/xyflow
• AutoGPT: https://github.com/Significant-Gravitas/AutoGPT
• LangFlow: https://github.com/logspace-ai/langflow

(back to top)

Getting Started

This is an example of how you may give instructions on setting up your project locally. To get a local copy up and running follow these simple example steps.

Prerequisites

Recommended: Install Docker (https://docs.docker.com/get-docker/) and docker-compose.

Alternatively, to install from source, you would need to install Node.js and Python>=3.10.

Installation

Adding API configs

This is required for both running from docker or source.

1. First create a config file backend/.env based on your API choice and info. Refer to the backend/.env.example for an example.

2. IF you are using Azure OpenAI API, create another Azure config file at backend/azure.yaml. Refer to the backend/azure.yaml.example for an example.

From docker (Recommended)

1. To deploy locally, build and run docker containers with docker-compose (Note that this runs a DEV server)

docker-compose up

From source

1. Clone the repo

git clone https://github.com/yiren-liu/coquest.git

2. Install and run server for the frontend

cd frontend/rq-flow
npm install
npm start

3. Install backend Python server requirements

cd backend/
pip install -r requirements.txt
python -m spacy download en_core_web_sm

4. Run backend server (Dev mode)

python main.py

The backend DB for logging change be changed to any self-hosted postgres DB by modifying the .env configs.

(back to top)

Usage

After deploying the service locally, visit: http://localhost:3000/app

Replacing the paper pool with your own

The paper pool used in the search function is vectorized and stored using ChromaDB under backend/paper_graph/db. Currently, the embedding model we used is OpenAI Ada 2, but you could use any other models if needed. When swapping the paper pool, modify and run the backend/paper_graph/get_embeddings.py.

(back to top)

Contact

Yiren Liu - @yirenl2 - [email protected]

SALT Lab @ UIUC: https://socialcomputing.web.illinois.edu/index.html

(back to top)