7.4 KiB

Raw Blame History

CLI Agent Guide: Create Agents from Command Line

Create, configure, and run AI agents directly from your terminal without writing Python code.

Basic Agent Creation

Step 1: Define Your Agent

Create an agent with required parameters:

swarms agent \
    --name "Research-Agent" \
    --description "An AI agent that researches topics and provides summaries" \
    --system-prompt "You are an expert researcher. Provide comprehensive, well-structured summaries with key insights." \
    --task "Research the current state of quantum computing and its applications"

Step 2: Customize Model Settings

Add model configuration options:

swarms agent \
    --name "Code-Reviewer" \
    --description "Expert code review assistant" \
    --system-prompt "You are a senior software engineer. Review code for best practices, bugs, and improvements." \
    --task "Review this Python function for efficiency: def fib(n): return fib(n-1) + fib(n-2) if n > 1 else n" \
    --model-name "gpt-4o-mini" \
    --temperature 0.1 \
    --max-loops 3

Step 3: Enable Advanced Features

Add streaming, dashboard, and autosave:

swarms agent \
    --name "Analysis-Agent" \
    --description "Data analysis specialist" \
    --system-prompt "You are a data analyst. Provide detailed statistical analysis and insights." \
    --task "Analyze market trends for electric vehicles in 2024" \
    --model-name "gpt-4" \
    --streaming-on \
    --verbose \
    --autosave \
    --saved-state-path "./agent_states/analysis_agent.json"

Complete Parameter Reference

Required Parameters

Parameter	Description	Example
`--name`	Agent name	`"Research-Agent"`
`--description`	Agent description	`"AI research assistant"`
`--system-prompt`	Agent's system instructions	`"You are an expert..."`
`--task`	Task for the agent	`"Analyze this data"`

Model Parameters

Parameter	Default	Description
`--model-name`	`"gpt-4"`	LLM model to use
`--temperature`	`None`	Creativity (0.0-2.0)
`--max-loops`	`None`	Maximum execution loops
`--context-length`	`None`	Context window size

Behavior Parameters

Parameter	Default	Description
`--auto-generate-prompt`	`False`	Auto-generate prompts
`--dynamic-temperature-enabled`	`False`	Dynamic temperature adjustment
`--dynamic-context-window`	`False`	Dynamic context window
`--streaming-on`	`False`	Enable streaming output
`--verbose`	`False`	Verbose mode

State Management

Parameter	Default	Description
`--autosave`	`False`	Enable autosave
`--saved-state-path`	`None`	Path to save state
`--dashboard`	`False`	Enable dashboard
`--return-step-meta`	`False`	Return step metadata

Integration

Parameter	Default	Description
`--mcp-url`	`None`	MCP server URL
`--user-name`	`None`	Username for agent
`--output-type`	`None`	Output format (str, json)
`--retry-attempts`	`None`	Retry attempts on failure

Use Case Examples

Financial Analyst Agent

swarms agent \
    --name "Financial-Analyst" \
    --description "Expert financial analysis and market insights" \
    --system-prompt "You are a CFA-certified financial analyst. Provide detailed market analysis with data-driven insights. Include risk assessments and recommendations." \
    --task "Analyze Apple (AAPL) stock performance and provide investment outlook for Q4 2024" \
    --model-name "gpt-4" \
    --temperature 0.2 \
    --max-loops 5 \
    --verbose

Code Generation Agent

swarms agent \
    --name "Code-Generator" \
    --description "Expert Python developer and code generator" \
    --system-prompt "You are an expert Python developer. Write clean, efficient, well-documented code following PEP 8 guidelines. Include type hints and docstrings." \
    --task "Create a Python class for managing a task queue with priority scheduling" \
    --model-name "gpt-4" \
    --temperature 0.1 \
    --streaming-on

Creative Writing Agent

swarms agent \
    --name "Creative-Writer" \
    --description "Professional content writer and storyteller" \
    --system-prompt "You are a professional writer with expertise in engaging content. Write compelling, creative content with strong narrative flow." \
    --task "Write a short story about a scientist who discovers time travel" \
    --model-name "gpt-4" \
    --temperature 0.8 \
    --max-loops 2

Research Summarizer Agent

swarms agent \
    --name "Research-Summarizer" \
    --description "Academic research summarization specialist" \
    --system-prompt "You are an academic researcher. Summarize research topics with key findings, methodologies, and implications. Cite sources when available." \
    --task "Summarize recent advances in CRISPR gene editing technology" \
    --model-name "gpt-4o-mini" \
    --temperature 0.3 \
    --verbose \
    --autosave

Scripting Examples

Bash Script with Multiple Agents

#!/bin/bash
# run_agents.sh

# Research phase
swarms agent \
    --name "Researcher" \
    --description "Research specialist" \
    --system-prompt "You are a researcher. Gather comprehensive information on topics." \
    --task "Research the impact of AI on healthcare" \
    --model-name "gpt-4o-mini" \
    --output-type "json" > research_output.json

# Analysis phase
swarms agent \
    --name "Analyst" \
    --description "Data analyst" \
    --system-prompt "You are an analyst. Analyze data and provide insights." \
    --task "Analyze the research findings from: $(cat research_output.json)" \
    --model-name "gpt-4o-mini" \
    --output-type "json" > analysis_output.json

echo "Pipeline complete!"

Loop Through Tasks

#!/bin/bash
# batch_analysis.sh

TOPICS=("renewable energy" "electric vehicles" "smart cities" "AI ethics")

for topic in "${TOPICS[@]}"; do
    echo "Analyzing: $topic"
    swarms agent \
        --name "Topic-Analyst" \
        --description "Topic analysis specialist" \
        --system-prompt "You are an expert analyst. Provide concise analysis of topics." \
        --task "Analyze current trends in: $topic" \
        --model-name "gpt-4o-mini" \
        >> "analysis_results.txt"
    echo "---" >> "analysis_results.txt"
done

Tips and Best Practices

!!! tip "System Prompt Tips" - Be specific about the agent's role and expertise - Include output format preferences - Specify any constraints or guidelines

!!! tip "Temperature Settings" - Use 0.1-0.3 for factual/analytical tasks - Use 0.5-0.7 for balanced responses - Use 0.8-1.0 for creative tasks

!!! tip "Performance Optimization" - Use gpt-4o-mini for simpler tasks (faster, cheaper) - Use gpt-4 for complex reasoning tasks - Set appropriate --max-loops to control execution time

!!! warning "Common Issues" - Ensure API key is set: export OPENAI_API_KEY="..." - Wrap multi-word arguments in quotes - Use --verbose to debug issues

Next Steps

CLI YAML Configuration - Run agents from YAML files
CLI Multi-Agent Guide - LLM Council and Heavy Swarm
CLI Reference - Complete command documentation

7.4 KiB Raw Blame History