Summary

ManimCat is an AI-assisted workspace for generating mathematical animations and static visuals from natural language. It targets educators and researchers needing efficient tools for explanatory content, offering a dual-mode workflow (direct generation vs. collaborative agent studio) and dual-engine approach (Manim for animation, matplotlib for plots). The project simplifies complex visual creation with AI-powered code fixing and iterative rendering.

How It Works

The system uses dual modes: Workflow for rapid generation, and Agent mode for collaborative studio sessions (Manim Studio for animation, Plot Studio for static visuals). It leverages Manim for animations and matplotlib for charts. AI is central, featuring two-stage generation (concept design, code generation), static analysis with AI auto-patching, and AI-driven retries for render failures.

Quick Start & Requirements

Installation involves Node.js package management (npm install, npm run dev) and navigating to the frontend directory. Docker deployment is supported via wingflow/manimcat. Key dependencies include Node.js, Docker, and an OpenAI-compatible API endpoint for AI. Environment variables and deployment specifics are in the deployment guide. Local development runs at http://localhost:3000.

Highlighted Details

• AI-driven code generation, static analysis, and auto-patching with multi-pass retry.
• Dual-stage AI generation: concept designer then code generator.
• Agent Mode features a Studio runtime with builder, designer, reviewer roles.
• Dedicated Manim Studio (animation) and Plot Studio (