Notes on AgentFlow

edit

Category:AgentFlow
Documentation —Orchestrating Multi-Agent Workflows —Notes on AgentFlow

AgentFlow is a RESTful endpoint that launches agents by orchestrated-intent of a top-holon, or mere mortal.

LLM-Launcher.exe Help Listing

C:\xenoKOS\AgentFlow>bin\llm-launcher.exe --help

AgentFlow Launcher - CLI Bridge for MindChat Multi-Agent Orchestration

Usage:
  llm-launcher.exe <command> [options]

Commands:
  launch-pipeline          Launch a multi-agent pipeline workflow
  chat                     Send a message between two agents (inter-loquation)
  monitor-emergence        Monitor for emergence events (Zenkin flash moments)
  help                     Show this help message

launch-pipeline Options:
  --agents <list>          Comma-separated agent IDs (required)
                          Example: CodeAssistant,Narrator,TestAgent
  --input <text>           Pipeline input text (required)
  --output-file <path>     Where to save output (optional)
  --output-format <fmt>    Output format: markdown, json, text (default: markdown)
  --topology <topo>        MindChat topology: interloquation, seminar, chorus,
                          convergence, whisper-chain, oracle (default: seminar)
  --turns <n>              Number of dialogue turns (default: 6)
  --participants <n>       Number of participants (overrides agent count)
  --vector-diffusion       Enable semantic vector alignment (default: true)
  --verbose                Enable verbose output

Examples:
  # Two-agent pipeline
  llm-launcher.exe launch-pipeline \
    --agents CodeAssistant,DocumentationAgent \
    --input "Design a caching system" \
    --output-file output.md \
    --turns 8

  # Code review consensus (3 agents)
  llm-launcher.exe launch-pipeline \
    --agents CodeAssistant,TestAgent,SecurityAgent \
    --input "Review this authentication code" \
    --topology seminar \
    --turns 4

chat Options:
  --from <agent>           Source agent ID (required)
  --to <agent>             Destination agent ID (required)
  --message <text>         Message to send (required)
  --vector-diffusion       Enable semantic alignment (default: true)

Examples:
  # Send message from one agent to another
  llm-launcher.exe chat \
    --from CodeAssistant \
    --to DocumentationAgent \
    --message "Explain this function in plain English"

monitor-emergence Options:
  --agents <list>          Comma-separated agent IDs (required)
  --query <text>           Query for emergence detection (required)
  --timeout <seconds>      Detection timeout (default: 120)

Examples:
  # Detect novel solutions from multi-agent collaboration
  llm-launcher.exe monitor-emergence \
    --agents CodeAssistant,SecurityAgent,ArchitectAgent \
    --query "Design a zero-trust authentication system" \
    --timeout 300

Global Options:
  -h, --help               Show this help message
  --verbose                Enable verbose output

Environment Variables:
  MINDCHAT_PORT            MindChat HTTP server port (default: 1000)
  VECTOR_DIFFUSION         Enable vector alignment globally
  VERBOSE                  Enable verbose logging

For more information, see: docs/LAUNCHER_GUIDE.md

C:\xenoKOS\AgentFlow>

See also

• The Social Brain of Paul S. Prueitt, Ph.D —http://www.blueberry-brain.org/winterchaos/PreuittTheSocialBrain.pdf

---

https://groupkos.com/dev/index.php/Notes_on_AgentFlow?action=purge