Prêt-à-Submit is a tool that allows researchers to use their own local LLMs to review their papers against conference's checklists. Each question is answered with Yes / Partial / No / NA, accompanied by an explanation — giving reviewers a structured, consistent first-pass assessment.

---

Features

• 📋 Conference management — define conferences with custom evaluation question
• 🤖 LLM-powered evaluation — sends papers to any Ollama-compatible model for structured review
• 📊 Visual results — doughnut chart summarising the distribution of Yes / Partial / No / NA answers
• 🔒 Privacy-first — all processing runs locally through your own Ollama instance; no data leaves your infrastructure
• 🐳 Docker-ready — single-command deployment

---

Tech Stack

Layer	Technology
Backend	Django 5, Python 3.12
LLM client	Ollama Python
PDF parsing	pymupdf4llm
Frontend	TailwindCSS v4, DaisyUI
Charts	Chart.js
Containerisation	Docker

---

Prerequisites

• Docker or Python 3.12+ with uv
• A running Ollama instance with at least one model pulled (e.g. ollama pull llama3). For better results we recommend using a model with a large context length, around 64K tokens or more, depending on the length of the paper. To change the context length of the Ollama server check the documentation here.

---

Getting Started

With Docker (recommended)

You may build it locally

# Build the image
docker build -t pret-a-submit .

# Run (host network is required to reach a local Ollama instance)
docker run --network=host pret-a-submit

or pull from the repository

# Run (host network is required to reach a local Ollama instance)
docker run --network=host ghcr.io/rowiz49/pret-a-submit:latest

The app will be available at http://localhost:8965.

Local development

# Install dependencies
uv sync

# Apply migrations
uv run python manage.py migrate

# Create a superuser to access the admin panel
uv run python manage.py createsuperuser

# Start the development server
uv run python manage.py tailwind dev

---

Configuration

1. Set up a Conference and Questions

1. Go to http://localhost:8965
2. Create a Conference with a name
3. Add Questions to the conference — each question has a position (order) and the question text the LLM will answer

2. Evaluate a Paper

1. Navigate to the home page
2. Select a conference
3. Enter your Ollama server URL (e.g. http://localhost:11434), model name (e.g. llama3), and optionally an API key
4. Upload a PDF paper
5. Submit — the app converts the PDF to Markdown, sends it to the LLM with the conference questions, and renders the structured evaluation

---

How It Works

PDF upload → PyMuPDF → Markdown text
                              ↓
              Ollama prompt with conference questions
                              ↓
              Structured JSON response (position, question, rating, explanation)
                              ↓
              Results page with per-question badges + summary chart

The LLM is constrained to reply in a JSON schema derived from a Pydantic model, ensuring parseable, structured output. Ratings are validated and sorted by question position before rendering.