Squill

Video Demo (Click the banner below!):

If you only want to interact with the first frontend pages, here is the deployed version: https://sopcopilot.vercel.app/

Squill is an application that guides students through the Statement of Purpose writing process by generating thoughtful questions rather than writing the content for them. Squill will take some personal information such as application details and resume points and formulate a set of personalized and beneficial questions to inspire its users to write and be more clear about the identity they portray through their applications.

---

Project Structure (Root Level)

.
main.py # Backend entrypoint (API + agent orchestration)
agents/                # LangGraph-based question generation pipeline
econf/                 # Environment configuration & runtime assets
frontend/              # Next.js (App Router) frontend application
public/                # Static assets
.env.example           # Environment variable template
README.md              # Project overview and architecture

Frontend Project Structure

app/            # Next.js App Router pages
components/     # Reusable UI components
public/         # Static assets (images, SVGs, icons)

Request Flow

User (UI)
   ↓
Frontend (Next.js)
   ↓
Backend API (main.py)
   ↓
LangGraph Agent Pipeline
   ↓
Generated Questions
   ↓
Frontend 
   ↓
User (UI)

Tech Stack Summary

Frontend

• Next.js (App Router), TypeScript
  • Chosen for fast iteration, server components, and strong typing.
  • And it is easy to vibe code.

Backend

• Python (main.py)
  • Simple, explicit orchestration layer that integrates cleanly with LangGraph.
  • And it is easy to vibe code.

LLM Provider

• Cohere (COHERE_API_KEY, see .env.example)
  • High-quality text generation with strong instruction-following.
  • And to support this great Canadian startup, eh.

Orchestration

• LangGraph (agent workflows)
  • Provides explicit stateful graphs instead of brittle prompt chains.
  • And to win the Forresters Challenge: https://www.linkedin.com/feed/update/urn:li:activity:7419442041513963520/

Quick Start

You will need to run both a frontend and backend server. Start with the backend.

Download the Directory and Start a Terminal

Click the download button, then start a terminal at the directory you just downloaded, SOPcopilot.

Initialize a virtual environment and install dependencies

Ensure your new terminal is in the root directory, SOPcopilot. Then, run:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create a .env file and fill it out

Duplicate the .env.example file. Rename it to .env. Then, replace COHERE_API_KEY=your_api_key with your API key from Cohere. If you do not have one, Cohere offers free keys (limit 20 calls/minute): sign up here. As a caveat, we unfortunately cannot guarantee full functionality with these free keys.

The first line should look something like COHERE_API_KEY=abcdefghijk1234jlkjfIJo3nFD(pa98sdUFJUCI (<— this is a fake key. Don't share your API keys!)

Run main.py to start the backend server

Run the command:

python3 main.py

Then, leave this terminal running. NOTE: if you see an error message like "AttributeError: module 'tensorflow' has no attribute 'contrib'", run pip uninstall agents and try this step again.

Open a new terminal

Open a new terminal in which to run the frontend server.

Navigate to frontend

In the terminal, move to the /frontend folder. If you're starting from the root directory, SOPcopilot, run cd frontend.

Install Dependencies

From the frontend directory, run:

npm install

Create a .env.local file and fill it out

Duplicate the file called .env.local.example. Rename it .env.local. Then, going back to the backend terminal, copy the bottom message that looks like this: Running on http://123.45.67.890:10000 Paste that IP address into .env.local in place of YOUR_BACKEND_URL. For example, the file should look like BACKEND_URL=http://123.45.67.890:10000.

Run the Frontend Server

Start the server with:

npm run dev

View the Application

Once the server is running, open your browser and navigate to:

http://localhost:3000

And begin using the application. We hope you enjoy!