Paper Research Assistant

Overview

A web-based research assistant that helps users explore and analyze academic papers. The application features a modern web interface with a tab-based research system and supports multiple AI models for paper analysis.

There are a few options:

• Single-paper Summarization: Summarize a paper, pull out key topics for analysis and allow for open-ended questions of that paper.
• Explore: As open-ended questions against a corpus of papers crawled into our vector store (see backend/apptasks/)
• Deep Research: A OpenAI/Google deep research-style searching of the web for pages related to a given paper. Can use either text-based browsing or a web agent like browser-use.

Uses:

• smolagents for the agents
• QDrant as a Vector DB
• LiteLLM for interacting with AI providers
• Railway for deployments
• FastAPI and Next.JS

Key Features

• Interactive research interface with multiple tabs
• Paper exploration and deep research capabilities
• Support for multiple AI models
• Persistent storage of research sessions
• Real-time paper analysis

Development Setup

Prerequisites

• Python 3.13+
• Node.js 18+
• uv for Python dependency management
• Various API keys for AI services

Backend Setup

1. Clone the repository
2. Navigate to the backend directory
3. Set up the environment with uv:

curl -LsSf https://astral.sh/uv/install.sh | sh
uv venv
source .venv/bin/activate

4. Install dependencies:

uv pip install -e .

5. Copy .env.example to .env and configure the keys.

6. Start the backend server:

fastapi dev main.py

Frontend Setup

1. Navigate to the frontend directory
2. Install dependencies:

npm install

3. Start the development server:

npm run dev

The frontend is built with:

• Next.js 15.2.0
• React 19
• TailwindCSS
• TypeScript

Development Commands

Backend:

• Install dev dependencies: uv pip install -e ".[dev]"
• Format code: black .
• Lint: ruff check
• Type checking: mypy
• Run tests: pytest

Frontend:

• Development server: npm run dev
• Build: npm run build
• Start production server: npm run start
• Lint: npm run lint
• Format code: npm run format

CI/CD

GitHub Actions

The project uses three main GitHub Actions workflows which run on PRs:

1. Python Static Analysis: Runs black (formatting), ruff (lint) and mypy (type checking) against all python code.
2. Frontend Linting: Runs eslint against all code in frontend.
3. Python Tests: Runs pytest suite in backend/app/tests.

Deployment

The application automatically deploys to Railway when changes are merged to the main branch.