Flowwise AI Agent Teams

Overview

• Flowwise offers a solution for creating AI agent teams without coding.
• Users can drag and drop nodes onto a canvas to configure teams.
• Team members collaborate to solve user tasks.
• Similar to Crew AI and AutoGen in functionality.

How It Works

• Utilizes LangRoth for orchestrating team's tasks.
• A supervisor agent delegates tasks among workers.
• Video tutorial covers fundamental steps for creating agent teams in Flowwise.

Example Projects

1. Software Development Team
2. Lead Research Team

Setup Instructions

• For new users, set up Flowwise locally using provided video tutorial.
• Ensure that your Flowwise instance is updated to access agent flows.

Creating the Software Development Team

1. Click on "Add New" in the Agent Flows page.
2. Name the team (e.g., "Software Development Team").
3. Add a supervisor node from the multi-agent section.
   • The supervisor manages task delegation and orchestration.
   • Optional: Add input moderation to the supervisor node.
4. Add a chat model (i.e., "Chat OpenAI Model") and connect it to the supervisor.
   • Set Model Name to GPT40 and adjust temperature from 0.9 to 0.
5. Add Worker nodes for team members (e.g., product designer, software developer, technical writer).
   • Each worker node can take input tools and is assigned roles.

Worker Prompts

• Product Designer: Responsible for feature set creation.
  • Prompt: "You work for a software development company. Your role as product designer is to come up with the features and product design based on user requirements."
• Software Developer: Writes code for the application.
  • Prompt: "As an experienced software developer, you build software solutions using React, Next.js, Tailwind, CSS, Express, JavaScript, and Node."
• Technical Writer: Creates user manuals in Markdown format.
  • Prompt: "Write a detailed user manual based on the product design and code written by the team members."

Testing the Team

• Test the performance by asking the team to build a to-do list app.
• Observe interactions among the supervisor, product designer, developer, and writer.
  • Supervisor passes requests and organizes task execution based on role sequences (product design, then development, then documentation).

Creating the Lead Research Team

1. Start by creating a new agent flow named "Lead Outreach Team".
2. Set up a supervisor node and add a chat model with similar parameters as above.
3. Add two worker nodes: Lead Researcher & Lead Sales Representative.
   • Lead Researcher: Builds detailed reports based on online research; requires Google search access.
   • Lead Sales Representative: Crafts personalized messages based on the researcher's findings.
4. Use a Google Custom Search tool connecting it to the Lead Researcher node to enable internet searches.
   • Steps for API and search engine ID setup include creating a Google project and enabling the Custom Search API.
5. Test the flow to ensure the researcher can access up-to-date information and generate an accurate report.

Conclusion

• AI agent teams can be easily constructed using Flowwise.
• Offer feedback on scenarios or use cases for future tutorials.