MindGenius AI offers an autonomous agent for generating editable mind maps directly from PDF documents. It targets users needing efficient document analysis and knowledge visualization, providing a transparent, multi-step AI decision-making process and supporting various LLM providers for flexibility and privacy.

How It Works

This project implements a true AI agent, distinct from simple prompt-to-markdown tools, by utilizing a multi-step, chained tool-calling loop powered by the Vercel AI SDK v5. The agent autonomously decides to search uploaded documents via a RAG (Retrieval-Augmented Generation) pipeline, generate map structures, and expand branches, with each step streamed live as a visible, expandable tool card. Its RAG implementation is notably concise (~100 lines), performing PDF parsing, overlap chunking, embedding, and cosine retrieval in-process without relying on heavy frameworks like LangChain.

Quick Start & Requirements

• Requirements: Node.js >= 20, pnpm >= 9.
• Installation: Clone the repository, install dependencies with pnpm install.
• Configuration: Copy apps/server/.env.example to apps/server/.env and set API keys (e.g., OPENAI_API_KEY).
• Running: Execute pnpm dev to start the web (port 5173) and server (port 3000) concurrently. Docker: cp apps/server/.env.example apps/server/.env then docker compose up --build (web on 8080, api on 3000).
• Links: docs/REFACTOR_PLAN.md for detailed design.

Highlighted Details

• Visible Tool-Calling Loop: Real-time streaming of agent actions (search, generate, expand) with input/output details in expandable cards.
• In-Process RAG: Efficient PDF document analysis, embedding, and retrieval directly within the application, citing sources in the generated map.
• Multi-Provider LLM Support: Runtime switching between OpenAI, Anthropic (Claude), and DeepSeek, with support for custom gateways and Bring Your Own Key (BYOK) stored client-side.
• Incremental Refactor: Successfully migrated stack from Vue/Koa/LangChain to React/Hono/AI SDK, maintaining backward compatibility for existing SSE endpoints.

Maintenance & Community

The project includes CI pipelines for linting, type-checking, testing, and building on every push, along with workspace-aware Docker builds. No specific community channels (Discord/Slack) or notable contributors/sponsorships are mentioned in the provided README.

Licensing & Compatibility

The project is released under the MIT license, permitting commercial use and integration within closed-source applications.

Limitations & Caveats

The RAG implementation is designed for in-memory processing, which may present scalability challenges for extremely large documents. While the README highlights a successful incremental refactor, potential complexities from managing legacy compatibility might exist. Specific hardware requirements beyond Node.js/pnpm versions are not detailed.