Multi-Agent Chat
Using the Naptha SDK and node infrastructure, we implemented a multiplayer chat app where each of the conversational agents runs on a different node using various LLMs.
Prerequisites
- Make sure you have the Naptha SDK installed.
Modules Used
This multi-agent, multi-node workflow is made of the following components, which you can find on the Naptha GitHub:
Module | Description | Repository |
---|---|---|
Multi-Agent Chat Orchestrator | Multi-agent chat across a network of multiple nodes. | GitHub |
Simple Chat Agent | Agent for running simple chat with LLMs. | GitHub |
Run
You can run it using the Naptha SDK with the following command:
naptha run multiagent_chat -p "prompt='I would like to count up to ten, one number at a time. ill start. one.'" --worker_nodes "http://node.naptha.ai:7001,http://node1.naptha.ai:7001"
Configuration Breakdown:
prompt
: Your initial message to start the conversationworker_nodes
: Comma-separated list of worker node URLs
This runs the flow across three nodes in total - one orchestrator node (whichever you have set as the NODE_URL
in the .env file of the Naptha SDK),
and two worker nodes (that you have set using the --worker_nodes
flag).
Example Conversation Starters
Try these prompts to see different interaction patterns:
- Counting game:
"I would like to count up to ten, one number at a time. ill start. one."
- Story creation:
"Let's create a story together. I'll start: Once upon a time..."
- Word association:
"Let's play word association. I'll say a word, and you respond with related words. Starting with: Ocean"
Start with simple interactions to understand the flow, then try more complex conversations as you get familiar with the system.
Each agent in this chat system runs independently on its own node, demonstrating true distributed AI interaction!