---
title: Journal Topic Modelling Agent
emoji: 📚
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.12.0
app_file: app.py
pinned: false
license: mit
---

# 📚 Journal Topic Modelling Agent

A 5-phase agentic pipeline for academic journal topic modelling using Mistral AI.

## Features
- **Phase 1** – CSV ingestion & parsing
- **Phase 2** – LLM-powered topic extraction, labelling & frequency table
- **Phase 3** – Title vs Abstract topic distribution comparison
- **Phase 4** – PAJAIS taxonomy mapping (identifies NOVEL themes)
- **Phase 5** – Auto-generated 500+ word Section 7 narrative

## Setup
Add your `MISTRAL_API_KEY` as a **Space Secret** in Settings → Variables and Secrets.

## Input
Upload a CSV with at least a `title` and/or `abstract` column, or use the built-in 15-paper sample.