Mini Agents

Run multi-step workflows in a few lines with one Agent per role, coordinated by AgentTeam

Mini agents run multi-step workflows in a few lines — one Agent per role, coordinated by AgentTeam.

from praisonaiagents import Agent, AgentTeam

researcher = Agent(name="Researcher", instructions="Research topics briefly.")
writer = Agent(name="Writer", instructions="Summarise findings.")
team = AgentTeam(agents=[researcher, writer])
team.start("Research and summarise quantum computing.")

The user describes a goal; the team runs each agent in sequence and returns a combined result.

How It Works

Quick Start

Create multiple AI agents that can work together in just a few lines of code!

Code

1. Install Package

   First, install the PraisonAI Agents package:

   pip install praisonaiagents duckduckgo_search
   
2. Set API Key

   Set your OpenAI API key as an environment variable in your terminal:

   export OPENAI_API_KEY="${OPENAI_API_KEY:?Set OPENAI_API_KEY in your shell}"
   
3. Create Your Agents

   Single Agent

   from praisonaiagents import Agent, AgentTeam
   summarise_agent = Agent(instructions="Summarise Photosynthesis")
   agents = AgentTeam(agents=[summarise_agent])
   agents.start()
   

   # Agent with Internet Search Tool
   from praisonaiagents import Agent, AgentTeam
   research_agent = Agent(instructions="Research about AI 2024", tools=[Tools.internet_search])
   summarise_agent = Agent(instructions="Summarise research agent's findings")
   agents = AgentTeam(agents=[research_agent, summarise_agent])
   agents.start()
   
4. Run Your Agents

   Execute your agents by running:

   python app.py
   

No Code

1. Install Package

   First, install the PraisonAI Agents package:

   pip install praisonaiagents duckduckgo_search
   
2. Set API Key

   Set your OpenAI API key as an environment variable in your terminal:

   export OPENAI_API_KEY="${OPENAI_API_KEY:?Set OPENAI_API_KEY in your shell}"
   
3. Create Your Agents

   Single Agent

   # Create a new file `agents.yaml` with the following content:
   agents:  # Canonical
     summarise_agent
       instructions: Summarise Photosynthesis
   

   # Create a new file `agents.yaml` with the following content:
   # Agent with Internet Search Tool
   agents:  # Canonical
     research_agent
       instructions: Research about AI 2024
       tools:
         - internet_search
     summarise_agent
       instructions: Summarise research agent's findings
   
4. Run Your Agents

   Execute your agents by running:

   python app.py
   

Prerequisites

• Python 3.10 or higher
• OpenAI API key. Generate OpenAI API key here. Use Other models using this guide.

Understanding Mini AI Agents

What are Mini AI Agents?

Mini AI Agents are simplified yet powerful AI agents that can:

• Work together to accomplish tasks
• Use tools like internet search
• Process and summarize information
• Execute tasks sequentially

Key Components

Agent

Individual AI agents with specific roles and capabilities

Agent(instructions="Your agent's role description")

Tools

Built-in tools that agents can use

tools=[Tools.internet_search]

Agents Manager

Coordinates multiple agents

AgentTeam(agents=[agent1, agent2])

Sequential Flow

Agents work in sequence, passing results to each other

Available Tools

Internet Search

Tools.internet_search

Search the internet using DuckDuckGo

Requires: pip install duckduckgo_search

Custom Instructions

Basic

# Simple instructions
research_agent = Agent(
    instructions="Research about climate change"
)

# Detailed instructions
research_agent = Agent(
    instructions="""
    You are a research agent focused on gathering information about:
    1. Latest AI developments in 2024
    2. Major breakthroughs in machine learning
    3. New AI applications in industry
    
    Provide detailed and accurate information from reliable sources.
    """
)

Best Practices

Agent Instructions

Write clear and specific instructions:

# Good
Agent(instructions="Research and analyze the latest AI developments in 2024")

# Too vague
Agent(instructions="Research AI")

Tool Usage

Provide tools only to agents that need them:

# Research agent needs search
research_agent = Agent(
    instructions="Research AI",
    tools=[Tools.internet_search]
)

# Summary agent doesn't need search
summary_agent = Agent(
    instructions="Summarize findings"
)

Common Patterns

Research and Analysis

# Research agent
researcher = Agent(
    instructions="Research latest developments in quantum computing",
    tools=[Tools.internet_search]
)

# Analysis agent
analyst = Agent(
    instructions="Analyze and explain the research findings in simple terms"
)

agents = AgentTeam(agents=[researcher, analyst])

Information Processing

# Data collector
collector = Agent(
    instructions="Collect information about renewable energy",
    tools=[Tools.internet_search]
)

# Summarizer
summarizer = Agent(
    instructions="Create a concise summary of the collected information"
)

# Report writer
writer = Agent(
    instructions="Write a detailed report based on the summary"
)

agents = AgentTeam(agents=[collector, summarizer, writer])

Troubleshooting

Tool Not Available

If using Tools.internet_search, install required package:

pip install duckduckgo_search

Agent Communication

Ensure agent instructions are clear and complementary

Next Steps

Advanced Agents

Learn about advanced agent configurations

Custom Tools

Create your own agent tools

Mini AI Agents are designed to be simple yet powerful. They’re perfect for quick prototypes and straightforward automation tasks.

Related

Build full multi-agent teams when tasks outgrow a single mini agent.

Add custom tools to give mini agents new capabilities.