AgentsExample

Multi-Agent Dashboard Demo

A demonstration of multiple agents working on scientific goals with a full-screen Rich dashboard showing live progress.

Source code: View on GitHub

Features

• Full-screen Rich dashboard with per-agent, per-goal live updates
• Configurable scientific goals in agents4science/workflows/goals.yaml
• Adjustable pacing via environment variables
• Multiple LLM backends supported

The Application

Five specialized agents collaborate on scientific goals defined in YAML:

Agent	Role
Scout	Identifies research opportunities
Planner	Designs workflows and allocates resources
Operator	Executes workflows safely
Analyst	Summarizes results and quantifies uncertainty
Archivist	Records provenance for reproducibility

Goals are defined in agents4science/workflows/goals.yaml and include examples like:

• “Find catalysts that improve CO2 conversion at room temperature”
• “Identify polymer compositions with high thermal stability”
• “Discover metal-organic frameworks for hydrogen storage”

Quick Start

cd Capabilities/local-agents/AgentsExample
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python main.py

LLM Configuration

The example supports three modes:

Mode	Environment Variable	Description
OpenAI	OPENAI_API_KEY	Uses OpenAI (gpt-4o-mini by default)
FIRST	FIRST_API_KEY	Uses FIRST HPC inference service
Mock	(none)	Uses mock responses for demonstration

# OpenAI mode
export OPENAI_API_KEY=<your_key>
python main.py

# FIRST mode (for HPC environments)
export FIRST_API_KEY=<your_token>
export FIRST_API_BASE=https://your-first-endpoint/v1
export FIRST_MODEL=meta-llama/Meta-Llama-3.1-70B-Instruct
python main.py

# Mock mode (no API key needed)
python main.py

Other Configuration

Environment Variable	Description	Default
A4S_LATENCY	Delay between agent steps (seconds)	0.4
A4S_TOOL_LATENCY	Delay for tool execution (seconds)	0.2
A4S_UI	Enable dashboard UI (0 to disable)	1
A4S_MODEL	Override model name for OpenAI	gpt-4o-mini

Directory Structure

AgentsExample/
├── main.py                      # Entry point
├── requirements.txt             # Dependencies
└── agents4science/
    ├── base_agent.py            # Base agent class with LLM mode detection
    ├── roles/                   # Agent implementations
    │   ├── scout.py
    │   ├── planner.py
    │   ├── operator.py
    │   ├── analyst.py
    │   └── archivist.py
    ├── tools/                   # Tool implementations
    │   ├── execution.py
    │   ├── analysis.py
    │   └── provenance.py
    └── workflows/
        └── goals.yaml           # Scientific goals

This site is open source. Improve this page.