8.9 KiB

Raw Blame History

CLI Heavy Swarm Guide: Comprehensive Task Analysis

Run Heavy Swarm from command line for complex task decomposition and comprehensive analysis with specialized agents.

Overview

Heavy Swarm follows a structured workflow:

Task Decomposition: Breaks down tasks into specialized questions
Parallel Execution: Executes specialized agents in parallel
Result Synthesis: Integrates and synthesizes results
Comprehensive Reporting: Generates detailed final reports

Basic Usage

Step 1: Run a Simple Analysis

swarms heavy-swarm --task "Analyze the current state of quantum computing"

Step 2: Customize with Options

swarms heavy-swarm \
    --task "Research renewable energy market trends" \
    --loops-per-agent 2 \
    --verbose

Step 3: Use Custom Models

swarms heavy-swarm \
    --task "Analyze cryptocurrency regulation globally" \
    --question-agent-model-name gpt-4 \
    --worker-model-name gpt-4 \
    --loops-per-agent 3 \
    --verbose

Command Options

Option	Default	Description
`--task`	Required	The task to analyze
`--loops-per-agent`	1	Execution loops per agent
`--question-agent-model-name`	gpt-4o-mini	Model for question generation
`--worker-model-name`	gpt-4o-mini	Model for worker agents
`--random-loops-per-agent`	False	Randomize loops (1-10)
`--verbose`	False	Enable detailed output

Specialized Agents

Heavy Swarm includes specialized agents for different aspects:

Agent	Role	Focus
Question Agent	Decomposes tasks	Generates targeted questions
Research Agent	Gathers information	Fast, trustworthy research
Analysis Agent	Processes data	Statistical analysis, insights
Writing Agent	Creates reports	Clear, structured documentation

Use Case Examples

Market Research

swarms heavy-swarm \
    --task "Comprehensive market analysis of the electric vehicle industry in North America" \
    --loops-per-agent 3 \
    --question-agent-model-name gpt-4 \
    --worker-model-name gpt-4 \
    --verbose

Technology Assessment

swarms heavy-swarm \
    --task "Evaluate the technical feasibility and ROI of implementing AI-powered customer service automation" \
    --loops-per-agent 2 \
    --verbose

Competitive Analysis

swarms heavy-swarm \
    --task "Analyze competitive landscape for cloud computing services: AWS vs Azure vs Google Cloud" \
    --loops-per-agent 2 \
    --question-agent-model-name gpt-4 \
    --verbose

Investment Research

swarms heavy-swarm \
    --task "Research investment opportunities in AI infrastructure companies for 2024-2025" \
    --loops-per-agent 3 \
    --worker-model-name gpt-4 \
    --verbose

Policy Analysis

swarms heavy-swarm \
    --task "Analyze the impact of proposed AI regulations on tech startups in the United States" \
    --loops-per-agent 2 \
    --verbose

Due Diligence

swarms heavy-swarm \
    --task "Conduct technology due diligence for acquiring a fintech startup focusing on payment processing" \
    --loops-per-agent 3 \
    --question-agent-model-name gpt-4 \
    --worker-model-name gpt-4 \
    --verbose

Workflow Visualization

┌─────────────────────────────────────────────────────────────────┐
│                         User Task                                │
│  "Analyze the impact of AI on healthcare"                       │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                      Question Agent                              │
│  Decomposes task into specialized questions:                     │
│  - What are current AI applications in healthcare?              │
│  - What are the regulatory challenges?                          │
│  - What is the market size and growth?                          │
│  - What are the key players and competitors?                    │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────┬─────────────┬─────────────┬─────────────┐
│  Research   │  Analysis   │  Research   │   Writing   │
│   Agent 1   │   Agent     │   Agent 2   │    Agent    │
└─────────────┴─────────────┴─────────────┴─────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                      Synthesis & Integration                     │
│              Combines all agent outputs                          │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                    Comprehensive Report                          │
│  - Executive Summary                                            │
│  - Detailed Findings                                            │
│  - Analysis & Insights                                          │
│  - Recommendations                                              │
└─────────────────────────────────────────────────────────────────┘

Configuration Recommendations

Quick Analysis (Cost-Effective)

swarms heavy-swarm \
    --task "Quick overview of [topic]" \
    --loops-per-agent 1 \
    --question-agent-model-name gpt-4o-mini \
    --worker-model-name gpt-4o-mini

Standard Research

swarms heavy-swarm \
    --task "Detailed analysis of [topic]" \
    --loops-per-agent 2 \
    --verbose

Deep Dive (Comprehensive)

swarms heavy-swarm \
    --task "Comprehensive research on [topic]" \
    --loops-per-agent 3 \
    --question-agent-model-name gpt-4 \
    --worker-model-name gpt-4 \
    --verbose

Exploratory (Variable Depth)

swarms heavy-swarm \
    --task "Explore [topic] with varying depth" \
    --random-loops-per-agent \
    --verbose

Best Practices

!!! tip "Task Formulation" - Be specific about what you want analyzed - Include scope and constraints - Specify desired output format

!!! tip "Loop Configuration" - Use --loops-per-agent 1 for quick overviews - Use --loops-per-agent 2-3 for detailed analysis - Higher loops = more comprehensive but slower

!!! tip "Model Selection" - Use gpt-4o-mini for cost-effective analysis - Use gpt-4 for complex, nuanced topics - Match model to task complexity

!!! warning "Performance Notes" - Deep analysis (3+ loops) may take several minutes - Higher loops increase API costs - Use --verbose to monitor progress

Comparison: LLM Council vs Heavy Swarm

Feature	LLM Council	Heavy Swarm
Focus	Collaborative decision-making	Comprehensive task analysis
Workflow	Parallel responses + peer review	Task decomposition + parallel research
Best For	Questions with multiple viewpoints	Complex research and analysis tasks
Output	Synthesized consensus	Detailed research report
Speed	Faster	More thorough but slower

Next Steps

CLI LLM Council Guide - Collaborative decisions
CLI Reference - Complete command documentation
Heavy Swarm Python API - Programmatic usage

8.9 KiB Raw Blame History