Commands for your workflow

Using Commands Within the Chat Input

All commands introduced here are designed to be used directly within the Agent Flow TUI, specifically in the Chat Input where you interact with agents or models. This integration allows you to seamlessly switch between conversing with the AI, refining outputs, and managing your agent development workflow—all within the same interface.

How It Works

The Chat Input serves as the main input field for communicating with agents. It supports a range of commands that can be used to:

• Directly send messages to the agent for responses.

• Issue commands for exporting responses.

• Feed files into the conversation as input, effectively reusing content.

By using commands within the Chat Input, you can control your workflow dynamically without needing to switch contexts or navigate through multiple menus.

---

Supported Commands Overview

1. Exporting Agent Responses to Markdown:

   • Command:

     /markdown [offset] [folder_alias]

   • Use Case: This command allows you to export specific responses from the chat history into a Markdown file for documentation, review, or reuse.

   • Example:

     /markdown reports

     This saves the latest response to the folder associated with the alias reports.

2. Feeding Input from Files:

   • Command:

     file [path-to-file]

   • Use Case: This command lets you import content from a file directly into the ongoing chat. It is especially useful when you want to reuse the output from one agent as input for another agent or task.

   • Example:

     file ~/documents/reports/llm_response_20241117_134512_offset_0.md

     This command loads the content of the specified Markdown file into the chat, allowing the current agent to process or respond to that content.

---

How These Commands Enhance the Agent Flow TUI

• Unified Workflow: By using commands directly within the Chat Input, you can interact with agents, export data, and feed inputs without leaving the chat context. This streamlines the process, making it faster and more intuitive.

• Agent Development & Integration: The ability to quickly export outputs and reuse them as inputs promotes a modular, iterative approach to agent development. This is particularly useful when using the Timeline Editor to create dialog scripts, Agent Clarity to refine responses, and Agent Assessment Builder to evaluate performance.

• Manual Control for High Precision: The manual nature of these commands gives you full control over the agent interaction process, allowing for fine-tuning before agents are fully integrated into automated scripts using the underlying Agent Dialog Manager in the UnderdogCowboy library.

---

Example Workflow

1. Develop a dialog with an agent in the Timeline Editor and refine the responses using Agent Clarity.

2. Export the refined response:

   /markdown assessments

   • This saves the response into the folder associated with the alias assessments.

3. Feed the saved response into a new agent using the file command:

   file ~/documents/assessments/llm_response_20241117_142030_offset_0.md

   • This loads the previous agent's output into the new chat context, enabling the second agent to continue processing or build upon the original conversation.

By leveraging these commands within the Chat Input of the Agent Flow TUI, you can create a robust, repeatable workflow for developing, refining, and integrating AI agents, ensuring that each step aligns perfectly with your project’s requirements.

By keeping the approach simple and intuitive, it allows users to immediately see the practical value of connecting different streams of intelligence. The modular process of exporting, refining, and reusing content not only empowers them to build better agents but also demonstrates how AI can seamlessly integrate into their existing workflows.

By making it tangible and hands-on, users can start to appreciate that it's not just about having a powerful model—it's about orchestrating interactions between multiple agents to achieve nuanced, high-quality outcomes. This unlocks the potential for collaborative AI workflows, where each agent becomes a specialized contributor in a larger system, much like members of a well-coordinated team.

We think this approach aligns perfectly with the vision behind UnderdogCowboy and Agent Flow—guiding users to harness AI in a way that’s not only effective but also empowering and adaptable to their unique needs.

Workshopping the Streamlining Streams of Intelligence

For many, the concept of streamlining streams of intelligence—connecting various AI agents and models to work together seamlessly—is still new and uncharted territory. It's not just about harnessing the capabilities of a single AI model but about orchestrating multiple agents to collaborate, refine, and build upon each other’s outputs. This process unlocks new levels of efficiency and creativity, yet it can feel complex and abstract without practical experience.

That's why we believe that hands-on experience is essential. In our workshops, participants will have the opportunity to experiment directly with the tools and techniques provided by UnderdogCowboy and the Agent Flow TUI. By actively working through real-world scenarios—refining agent responses, exporting conversations, and feeding data back into new workflows—participants will gain a tangible understanding of how to connect and leverage streams of intelligence.

This immersive, interactive approach not only demystifies the process but also empowers participants to see the immediate impact on their own workflows. By the end of the workshop, attendees will leave with the confidence to navigate this new frontier, equipped with practical strategies to streamline and optimize their own AI-driven processes.