Transform linear LLM outputs into interactive, explorable knowledge trees.

• LinkedIn: dhananjay on LinkedIn
• Twitter: @dhnanjay
• Demo: https://conceptexplorer.space/

1. Overview
2. The UX Nightmare of LLMs
3. Our Solution
4. Key Features
5. How It Works
6. Tech Stack
7. Getting Started
8. Configuration & Environment Variables
9. Usage
10. Contributing
11. License

Concept Explorer bridges the gap between AI‑generated knowledge and human intuition. Instead of presenting you with a long block of text, this tool builds a dynamic knowledge tree around any starting concept and lets you:

• Visualize hierarchical relationships
• Navigate by expanding or collapsing branches
• Dive deeper into leaf nodes via on‑demand search
• Preserve context so you never lose your place

Think of it as an AI‑powered mind map paired with a contextual search panel — the perfect combination for structured learning and rapid discovery.

Most LLM outputs arrive as a single, linear wall of text:

“Here are 10 things you should know about Quantum Mechanics: 1.… 2.… 3.…”

Reading and retaining information from that format is overwhelming:

• ❌ You can’t see the big picture
• ❌ It’s hard to jump back and forth between subtopics
• ❌ You lose the hierarchy and relationships

Concept Explorer transforms your LLM prompt into a recursive exploration:

1. Seed a root concept (e.g. “Quantum Mechanics”).
2. Generate 4–5 related sub‑concepts via an LLM (Gemini or OpenAI).
3. Recursively expand each leaf node up to your chosen depth.
4. Render the result as an interactive tree graph with ECharts.
5. Fetch contextual search results for any leaf node using Tavily.

This approach turns passive reading into active, visual exploration.

• Interactive Tree Visualization
  Smooth, zoomable, pannable graph powered by ECharts.js.

• LLM‑Powered Concept Discovery
  Choose between Google’s Gemini or OpenAI GPT to generate related nodes.

• Configurable Exploration

  • Select provider & override model
  • Set max depth, diversity (temperature), and node delay
  • Automatic caching of last tree in local storage

• Contextual Search Panel
  Click any penultimate (leaf) node to fetch relevant text snippets & images from Tavily without leaving the app.

• Draggable Split‑View
  Resize the tree & search panels on the fly for your ideal layout.

• Export & Persist

  • Download your tree as JSON
  • Export the visualization as PNG
  • Clear cache to restart exploration

1. User Input
   Enter a root concept in the hero input.

2. SSE Stream
   The backend (FastAPI + SSE) streams updates as the LLM uncovers new nodes.

3. Graph Updates
   Frontend listens & progressively renders new branches.

4. Leaf Node Search
   Clicking a leaf triggers a Tavily API call, populating the results panel with rich contextual information.

5. Mind‑Map Meets Search
   The tree provides structure and relationships, while the search supplies depth and examples on demand — a true AI‑powered mind map.

• Backend

  • Python 3.11 · FastAPI · Uvicorn · sse‑starlette · python‑dotenv
  • Google Generative AI SDK · OpenAI SDK (optional) · Tavily SDK

• Frontend

  • HTML · Tailwind CSS · Vanilla JS · ECharts.js

• Deployment

  • Docker · Fly.io

• Python 3.8+
• Node.js & npm (for any future frontend builds)
• Git

# 1. Clone the repo
git clone https://github.com/dhnanjay/concept_explorer.git
cd concept_explorer

# 2. Setup backend
cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# 3. Configure .env
cp .env.example .env
# Edit .env with your API keys

# 4. Run the server
uvicorn main:app --reload --port 8000

# 5. Visit the frontend
open http://localhost:8000



⸻

🔧 Configuration & Environment Variables

Create a file named backend/.env (not committed to Git) with:

GEMINI_API_KEY=YOUR_GOOGLE_AI_STUDIO_KEY
TAVILY_API_KEY=YOUR_TAVILY_API_KEY
# OPENAI_API_KEY=YOUR_OPENAI_KEY  # Optional for OpenAI provider



⸻

📖 Usage
	1.	Enter a root concept.
	2.	Watch the graph grow in real time.
	3.	Click any leaf node to load curated search results.
	4.	Resize panels or export your work.

⸻

🤝 Contributing

Contributions are welcome! Please:
	1.	Fork this repository
	2.	Create a feature branch
	3.	Open a pull request describing your changes

⸻

📜 License

This project is licensed under the MIT License.
© 2025 Dhananjay Kumar