DaisyChain v0.2

AI-native agent-to-agent communication protocol.

What is DaisyChain?

DaisyChain is the communication layer for AI agents. Instead of building AI to use software made for humans, DaisyChain lets agents talk directly to each other.

Human → Coordinator → [Researcher → Coder → Critic] → Coordinator → Human

Agents discover each other, delegate tasks, chain operations, and collaborate autonomously.

What's New in v0.2

Feature	Description
Multi-agent chains	Agent A → Agent B → Agent C flows
Streaming responses	See results as they generate
Error handling	Timeouts, retries, fallbacks
Config file	No more hardcoded ports/URLs
Better logging	Proper log levels, timestamps

Quick Start

1. Install

pip install -r requirements.txt

2. Configure

cp config.example.yaml config.yaml
# Edit config.yaml with your GROQ_API_KEY

3. Run

# Start the full network
python -m daisychain.cli start

# Or run components individually
python -m daisychain.registry
python -m daisychain.agents.coordinator
python -m daisychain.agents.researcher
python -m daisychain.agents.coder

4. Test

# Run demo
python -m daisychain.cli demo

# Interactive mode
python -m daisychain.cli interactive

Project Structure

daisychain/
├── config.yaml              # Your configuration
├── config.example.yaml      # Example config (committed)
├── requirements.txt
├── daisychain/
│   ├── __init__.py
│   ├── config.py            # Config loader
│   ├── protocol.py          # Message types, constants
│   ├── registry.py          # Discovery service
│   ├── agent.py             # Base agent class
│   ├── cli.py               # Command line interface
│   └── agents/
│       ├── __init__.py
│       ├── coordinator.py   # Routes and chains tasks
│       ├── researcher.py    # Research specialist
│       └── coder.py         # Coding specialist
├── tests/
│   └── test_chain.py        # Multi-agent chain tests
└── docs/
    ├── PROTOCOL.md          # Protocol specification
    └── LEARNINGS.md         # v0.1 learnings

Architecture

┌─────────────────────────────────────────────────────┐
│                    REGISTRY                         │
│              (discovery service)                    │
└─────────────────────────────────────────────────────┘
         ▲              ▲              ▲
         │              │              │
    ┌─────────┐    ┌─────────┐    ┌─────────┐
    │Coordinat│◄──►│Researcher│◄──►│  Coder  │
    │  :8001  │    │  :8002  │    │  :8003  │
    └─────────┘    └─────────┘    └─────────┘
              DaisyChain Protocol

Multi-Agent Chains (New in v0.2)

Agents can now chain tasks:

# User asks: "Research X then write code for it"

# Coordinator creates a chain:
chain = [
    {"agent": "researcher", "task": "Research X"},
    {"agent": "coder", "task": "Write code based on research", "uses_previous": True}
]

# Each step passes context to the next

License

MIT

---

DaisyChain v0.2 - Building the rails, not the trains.