Summary

Engineering Figure Agent addresses the challenge of generating publication-quality figures for engineering and computer science papers, offering specialized modes for conceptual diagrams and precise quantitative plots. It targets researchers seeking an efficient, agent-native workflow for figure production, benefiting from bilingual support and an editable handoff process.

How It Works

The project employs a dual-mode approach: 'Image Mode' for conceptual visuals like system architectures and algorithms using prompt templates, and 'Plot Mode' for exact quantitative data representation (charts, curves) via local plotting. This distinction allows for accurate data rendering while enabling creative exploration of visual layouts. It supports backends including Gemini, OpenAI, and local plotting libraries, facilitating mixed-mode figures by rendering numeric panels locally before composing conceptual elements.

Quick Start & Requirements

Installation and testing can be initiated via a PowerShell script: "$HOME/.codex/skills/engineering-figure-agent/scripts/install_and_test.ps1" -RunSetupCheck. A wizard is available at "$HOME/.codex/skills/engineering-figure-agent/scripts/wizard.ps1". Direct image generation uses scripts/generate_image.py. Windows is the primary tested platform, with macOS and Linux support expected. API keys for providers like OpenAI or Gemini are required and should be managed externally. Links to English and Chinese guides, an example gallery, and a showcase are provided within the repository's documentation.

Highlighted Details

• Agent-native figure production tailored for academic workflows.
• Supports bilingual labeling and Chinese academic paper requirements.
• Distinct 'Image Mode' for conceptual diagrams and 'Plot Mode' for precise quantitative figures.
• Flexible backend support: Gemini, OpenAI, and local plotting.
• Focuses on researchers needing a streamlined figure production path.

Maintenance & Community

The provided README does not detail specific contributors, sponsorships, partnerships, or community channels like Discord or Slack.

Licensing & Compatibility

The project is released under the MIT License, which is permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

This tool is not a comprehensive paper-upload platform and does not replace the need for manual verification of scientific accuracy, labels, or numeric values. It explicitly avoids treating plots, diagrams, and schematics as a single, undifferentiated prompt task. Users must manage API keys securely outside the repository and should never rely on image generation for the precise geometry or values in quantitative figures.