PaperBanana: Academic Illustration Pipeline

Automates publication-ready academic illustrations via 5 specialized agents, each implemented as a separate Gemini API call: Retriever (categorize & select references) -> Planner (multimodal description) -> Stylist (polish) -> Visualizer (render) -> Critic (evaluate & refine).

Two output modes:

• DIAGRAM MODE: Each agent is a Python script calling Gemini VLM/image APIs. Run scripts/orchestrate.py for end-to-end execution.
• PLOT MODE: Statistical plots generated as executable Python matplotlib/seaborn code (code-based to eliminate data hallucination).

Requirements: GOOGLE_API_KEY env var (used for VLM calls in retriever/planner/stylist/critic AND image generation in visualizer), Python 3.10+ with google-genai, matplotlib, seaborn, numpy, pillow.

Paper: PaperBanana: Automating Academic Illustrations with Multi-Agent Systems (arXiv:2601.23265, Google/PKU)

Step 1: Determine Output Mode

Decide which track to follow: