Payed Agents

A CLI application that implements a pay-per-query system using the CDP (Coinbase Developer Docs) to facilitate transactions between consumers and providers in AI agent workflows.

Overview

This application enables secure and transparent payment processing for AI agent interactions, where consumers can submit queries and pay for responses from AI providers. The system uses blockchain-based transactions via CDP to handle payments.

Features

• Pay-per-query system for AI agent interactions
• Secure wallet management for both consumers and providers
• LLM integration using LangChain and Groq
• Workflow management with LangGraph
• Performance metrics and transaction monitoring
• CLI interface for easy interaction

Prerequisites

Before using this application, you need:

1. Pre-filled wallets with ETH and USDC

   • Both consumer and provider wallets must have ETH for gas fees
   • Consumer wallet must have sufficient USDC for payments
   • You can use testnet faucets to get test ETH and USDC

2. API Keys

   • Create and add your CDP API key to api-key/cdp_api_key.json
   • ArXiv tool requires no API key (used by paper_researcher agent)
   • For Google Search, set environment variables GOOGLE_API_KEY and GOOGLE_CSE_ID (optional)

Configuration

1. The application uses a YAML configuration file (config.yaml) to manage settings
2. Key configuration includes:
   • LLM providers and models
   • Wallet settings
   • File paths
   • Agent definitions

You can modify the config.yaml file to adjust these settings.

Usage

Process a query with default settings:

python main.py --query "Your query text here"

Process a query from a file:

python main.py --file path/to/query.txt

Generate a report:

python main.py --query "Your query" --export-report

Specify a different LLM provider:

python main.py --query "Your query" --provider anthropic

Specify both provider and model:

python main.py --query "Your query" --provider openai --model gpt-4o-mini

Use a specific agent for processing:

python main.py --query "What are the latest papers on LLMs?" --agent paper_researcher

List available agents:

python main.py --list-agents

List available models for a provider:

python main.py --list-models --provider anthropic

Use a custom configuration file:

python main.py --query "Your query" --config my_custom_config.yaml

Available LLM Providers

The default configuration includes the following providers and models:

• groq (default)

  • llama3-8b-8192 (default)
  • llama3-70b-8192
  • mixtral-8x7b-32768

• anthropic

  • claude-3-5-sonnet-20240620 (default)
  • claude-3-opus-20240229
  • claude-3-haiku-20240307

• openai

  • gpt-4o (default)
  • gpt-4o-mini
  • gpt-3.5-turbo

Available Agents

The system includes specialized agents for different tasks:

• basic_llm (default): Simple LLM-based agent without tools
• paper_researcher: Scientific paper research agent using arXiv (fully functional for demonstration)
• web_researcher: Web search agent with Google Search (demo purposes only, requires API keys)
• code_assistant: Programming assistant with access to documentation (demo purposes only)

Note: Currently, only the paper_researcher agent is fully functional for demonstration purposes without additional configuration. The other specialized agents are included as examples and would require additional API keys and configuration to work properly.

Custom Agents

You can define custom agents in the config.yaml file:

agents:
  definitions:
    my_custom_agent:
      description: "Custom agent description"
      type: "tool_augmented"
      provider: "groq"
      model: "llama3-70b-8192"
      temperature: 0.5
      prompt_template: "You are a custom assistant that helps with {request}."
      tools:
        - "tool_name"

(OUTPUT EXAMPLE)

Dependencies

• cdp-sdk: Handles blockchain transactions
• langchain: Framework for LLM applications
• langgraph: Workflow orchestration
• langchain-groq: Groq LLM integration
• pandas: Data analysis tools
• pyyaml: YAML file processing
• langchain-openai, langchain-anthropic: Additional LLM provider integrations