This repository contains the implementation for the paper "PeerCopilot: A Language Model-Powered Assistant for Behavioral Health Organizations" (IAAI 2025).

Paper: arXiv:2511.21721 | PDF

This work was done by Gao Mo*, Naveen Raman*, Megan Chai, Cindy Peng, Shannon Pagdon, Nev Jones, Hong Shen, Peggy Swarbrick, Fei Fang.

TL;DR

Peer-run behavioral health organizations offer holistic wellness support by combining mental health services with assistance for needs such as housing, employment, and income. However, staffing and expertise limitations hinder behavioral health organizations, making meeting all service user needs difficult. We address this issue through PeerCoPilot, a large language model (LLM)-powered assistant that helps peer providers create wellness plans, construct step-by-step goals, and find resources for these goals. Because information reliability is critical for peer providers, we designed PeerCoPilot to rely on information verified by peer providers via retrieval augmented generation. We conducted human evaluations with 15 peer providers and 6 service users and found that both groups overwhelmingly supported using PeerCoPilot. We show that PeerCoPilot provides more reliable and specific information than a baseline LLM. PeerCoPilot is now used by a group of peer providers at cspnj, a large behavioral health organization serving over 10,000 service users, and we are actively expanding PeerCoPilot's use.

To set up the backend environment, first clone this repository:

git clone https://github.com/chengolivia/community_services_llm.git
cd community_services_llm

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

Install Python dependencies:

cd backend
pip install -r requirements.txt

Install frontend dependencies:

cd ../frontend
npm install

Set up environment variables. Create a .env file in the backend directory with:

OPENAI_API_KEY=your_key_here
DATABASE_URL=your_database_url
RESOURCE_DB_URL=your_resource_database_url
GOOGLE_API_KEY=your_google_api_key
BRAVE_API_KEY=your_brave_api_key
SECRET_KEY=your_secret_key
HF_TOKEN=your_huggingface_token

Start the backend server:

cd backend
npm run start

In a separate terminal, start the frontend:

cd frontend
npm start

PeerCoPilot will be available at:

http://localhost:3000/ (frontend) http://127.0.0.1:8000/ (backend API)

To extend this to new organizations, prepare a file called <name>_resources.txt in the backend/data folder Next, scrape the resources by running

python scrape_resources.py --org_name {name} --location "<location/state>"

Finally, add this new organization to Home.js and submodules.py.