Core3

Understand any topic through its 3 most important principles

A multi-agent AI system that distills complex topics into exactly 3 first principles with curated resources. The application allows you to study a topic, then uses silence to let ideas assimilate. At the end, Dōgen and Wittgenstein prompt a moment of introspection on the topic.

---

🎯 Problem

Modern AI assistants are verbose by design—endless conversations when clarity requires few words. Users face:

• Information Overload — Too much content, not enough clarity
• Lost Fundamentals — Surface-level explanations miss core principles
• Conversation Fatigue — Multi-turn dialogues when single answers suffice

💡 Solution

Inspired by Zen philosophy: maximum insight through minimum words, followed by silence.

Buddhist Oracle delivers:

• 3 First Principles — The fundamental concepts underlying any topic
• Curated Resources — Authoritative links for each principle
• Visual Learning — A relevant YouTube video
• Markdown Output — Results saved as readable .md files in outputs/
• Philosophical Closure — Ends definitively, no follow-up prompts
• Reflective Questions — Two philosophers challenge you to think deeper

---

🏗️ Architecture

User Query
    │
    ▼
┌─────────────────────────────────────────┐
│     PHASE 1: SEQUENTIAL PIPELINE        │
│                                         │
│  Strategist → Librarian → Editor →      │
│  Projectionist → Sage                   │
│                                         │
│  (5 agents passing state via output_key)│
└─────────────────────────────────────────┘
    │
    ▼
        * 5-Second Silence *
    │
    ▼
┌─────────────────────────────────────────┐
│     PHASE 2: PARALLEL PHILOSOPHERS      │
│                                         │
│   Monk Dōgen    ║    Wittgenstein       │
│   (Zen koan)    ║    (Language game)    │
│                                         │
│   (asyncio.gather - run in parallel)    │
└─────────────────────────────────────────┘
    │
    ▼
Session Complete (no follow-ups)

Agent Roles

Agent	Role	Tools
TheStrategist	Extract 3 first principles	—
TheLibrarian	Find authoritative sources	google_search
TheEditor	Format as Markdown	—
TheProjectionist	Find YouTube video	google_search
TheSage	Present final answer + closure	—
MonkDōgen	Ask Zen koan question	—
LudwigWittgenstein	Ask language/meaning question	—

---

🔧 Technical Concepts (3/3)

Concept	Implementation
Multi-Agent (Sequential)	SequentialAgent with 5 sub-agents, state passing via output_key
Multi-Agent (Parallel)	asyncio.gather() runs Dōgen + Wittgenstein simultaneously
Tools + Sessions	google_search tool + InMemoryRunner with session management

---

🚀 Setup

Prerequisites

• Python 3.10+
• Google API Key (Gemini)

Installation

1. Clone the repository

   git clone https://github.com/EyalIv/agents-llm-silence.git
   cd agents-llm-silence

2. Create virtual environment

   python -m venv venv
   venv\Scripts\activate  # Windows
   # source venv/bin/activate  # macOS/Linux

3. Install dependencies

   pip install -r requirements.txt

4. Configure API key

   Create a .env file in the root directory:

   GOOGLE_API_KEY=your_api_key_here
   

Run

python main.py

---

📖 Usage Example

══════════════════════════════════════════════════════
 BUDDHIST ORACLE - FIRST PRINCIPLES AGENT
══════════════════════════════════════════════════════

What topic do you seek to understand? > How does blockchain work?

[The Oracle is contemplating...]

1. **Decentralization** - No single authority controls the network.
   → https://ethereum.org/decentralization

2. **Cryptographic Hashing** - Data integrity through one-way functions.
   → https://www.investopedia.com/crypto-hash

3. **Consensus Mechanisms** - Network agrees on valid transactions.
   → https://consensys.net/consensus-explained

**Visual Guide:** https://youtube.com/watch?v=blockchain-explained

*These are the answers.*
In silence, understanding deepens.

                    * Silence *

──────────────── The Masters Speak ────────────────

Monk Dōgen asks:
"When you speak of 'trust' in a trustless system, 
 where does your body feel that trust reside?"

Ludwig Wittgenstein asks:
"In what language game does 'decentralized' 
 gain its meaning—and who taught you to play?"

══════════════════════════════════════════════════════

📄 Results saved to: outputs/2025-11-30_14-30-00_blockchain.md

---

📁 Project Structure

buddhist-oracle/
├── agent.py         # 7 agent definitions
├── main.py          # Runner, CLI, async execution
├── outputs/         # Saved markdown results (auto-created)
├── requirements.txt # Dependencies
├── PRD.md           # Product Requirements Document
├── README.md        # This file
└── .env             # API key (not committed)

---

🛠️ Technology Stack

Component	Technology
Framework	Google ADK (Agent Development Kit)
LLM	Gemini 2.5 Flash
Tools	Google Search (built-in ADK)
Sessions	InMemoryRunner
Error Handling	Exponential backoff (5 retries)
Async	Python asyncio

---

📚 Dependencies

google-adk>=0.1.0
google-generativeai>=0.3.0
python-dotenv>=1.0.0

---

🎨 Design Principles

1. Minimalism — Exactly 3 principles, no more
2. Completeness — Answer is self-sufficient
3. Closure — Ends with "These are the answers"
4. Reflection — Philosophers invite deeper thinking

---

📄 License

MIT License

---

🙏 Acknowledgments

• Google ADK Team
• Kaggle Agents Competition
• Zen Master Eihei Dōgen (1200-1253)
• Ludwig Wittgenstein (1889-1951)