Pods are a set of agentic workflows triggered by using the /pod command within Discord. Follow these steps to get started:

  1. Join a Discord Server: Ensure you are a member of a Discord server where the bot is installed.
  2. Use the /pod Command: In any text channel, type /pod followed by a prompt you want the agents to work. For example, /pod prompt: Make a Table of Sakura Season in past 10 years in Tokyo and add Future Projections to initiate a agent run.
  3. Follow Prompts: The bot will guide you through the necessary steps to complete the agent run. Follow the prompts and provide any required information.
  4. Complete the agent Run: Once all steps are completed, the bot will confirm the successful execution of the agent run.

Getting Started

StreamPod Agent is Discord bot for interacting with AI agents and models through natural language prompts. Supports model switching, agent type selection, and provides detailed execution metrics.

Overview

  • Commands: /pod, /setmodel, /listmodels, /setagent, /listagents
  • Agents: CodeAgent, ToolCallingAgent
  • Models: Multiple AWS Bedrock models including Meta Llama 3 variants and Claude 3 models
  • Features:
    • Dynamic model/agent switching
    • Response formatting with code blocks
    • Execution metrics (tokens, duration, costs)
    • Autocomplete for models/agents

Commands

/pod [prompt]

Submit a prompt to the active AI agent.

Parameters:

prompt: Your query/instruction Response Features:

/setmodel [model_id]

Switch the active AI model (autocomplete supported).

Example:

/setmodel model_id: bedrock/anthropic.claude-3-haiku-20240307-v1:0

/listmodels

Show all available models.

/setagent [agent_type]

Switch agent type (autocomplete supported).

/listagents

Show available agent types.

Configuration

AVAILABLE_MODELS = [
    "bedrock/us.meta.llama3-2-1b-instruct-v1:0",
    "bedrock/us.meta.llama3-2-3b-instruct-v1:0",
    "bedrock/us.meta.llama3-2-11b-instruct-v1:0",
    "bedrock/us.meta.llama3-3-70b-instruct-v1:0",
    "bedrock/amazon.nova-micro-v1:0",
    "bedrock/amazon.nova-lite-v1:0",
    "bedrock/amazon.nova-pro-v1:0",
    "bedrock/anthropic.claude-3-haiku-20240307-v1:0",
    "bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"
]

Agent Types

AVAILABLE_AGENTS = ["CodeAgent", "ToolCallingAgent"]

Best Practices

  1. Use code blocks in prompts for technical questions
  2. Monitor token usage with /pod metrics
  3. Test different agent types for complex tasks
  4. Use model autocomplete (type partial model names)
  5. Check execution logs for troubleshooting