Knowledge Garden

Introduction

This is an AI agent that can sort and give report automatically according to your daily thoughts(short inputs).

Applications

Nowadays, many people read articles, listen to podcasts, and browse news every day, but this knowledge is fragmented and forgotten after a few months. The system acts like an "AI gardener", helping to organize, connect, and foster new ideas.

Construction

knowledge-garden-ai/
├── api/                # Backend service powered by FastAPI/Flask
│   ├── main.py             # Entry point: API route definitions and orchestration
│   ├── agents/             # Core AI logic modules
│   │   ├── agent1.py    # Agent 1 ingestion: Content processing and summarization
│   │   ├── agent2.py    # Agent 2 synthesis: Relationship discovery and graph linking
│   │   ├── agent3.py       # Agent 3 growth: Spaced repetition and output suggestions
│   │   └── agent4.py       # Agent 4 Powerpoint Generator
│   └── requirements.txt    # Python dependencies
├── ./                 # Frontend application (Root for web files)
│   ├── index.html          # Homepage: Inspiration capture and input
│   ├── garden.html         # Visualization: Knowledge graph and card grid
│   ├── js/             # Static assets (Custom CSS, JS logic)
│   │   └──  lang.js    # JS file for multi-language
└── README.md               # Project documentation and setup guide

*： garden_backup_sample.json is for demo use. You can download and import this file at Garden Page to see sample outputs.

Deployment

Or you can clone this repos and deploy it on your own server, you should set the following environment variables, or change them in api/main.py.

Environment Variables Keys	Description
DEEPSEEK_BASE_URL	input your openai url, not limited to deepseek, any model is ok.
DEEPSEEK_API_KEY	input your openai api, not limited to deepseek, any model is ok.

Demo

Demo Website

Agent Roles:

• Ingestion Agent: Helps users input/summarize what they've learned today (supports links or text). → Automatic summary.

• Connection & Synthesis Agent: Discovers hidden connections between knowledge points, verifying/expanding them with code or search. (Uses simple embeddings or rules + code execution to create mind maps).

• Growth & Reminder Agent: Generates a "knowledge growth plan" (next review time, application scenarios, output ideas such as short articles).

• Powerpoint Agent: Generate Markdown for Powerpoint Generator according to the contents generated by the agents above.