Skip to content
FloTorch

FloTorch ADK Plugin Overview

The FloTorch ADK Plugin streamlines the development of Google ADK agents by providing managed infrastructure and centralized configuration through the FloTorch Console. Instead of managing complex configurations in code, developers can leverage a powerful set of pre-configured services:

  • Centralized Agent Management - Configure and manage agents through the FloTorch Console
  • Managed LLM Access - Seamless integration with FloTorch Gateway for model inference
  • Persistent Memory Services - Optional memory and session capabilities with provider setup
  • Automated Tool Integration - Support for MCP tools and custom tool implementations
  • Distributed Tracing - Built-in observability with configurable tracing for production monitoring
  • Configurable Logging - Flexible logging configuration for debugging and monitoring (see SDK Logging Configuration)

Prerequisites

Section titled “Prerequisites”

Before getting started with the FloTorch ADK Plugin, ensure you have completed the following:

  1. FloTorch Account - Create an account at console.flotorch.cloud
  2. Agent Configuration - Set up your agent following the Gateway Agents documentation
  3. API Credentials - Generate your API key from API Keys settings
  4. Memory Provider (Optional) - Configure if using memory features, as detailed in the Memory documentation

Installation

Section titled “Installation”

Install the FloTorch ADK Plugin using pip:

Terminal window
pip install flotorch[adk]

Environment Configuration

Section titled “Environment Configuration”

Configure your environment variables to avoid hardcoding credentials:

Terminal window
export FLOTORCH_API_KEY="your_api_key"
export FLOTORCH_BASE_URL="https://gateway.flotorch.cloud"

Optional Logging Configuration:

Terminal window
# Enable debug logging (optional)
export FLOTORCH_LOG_DEBUG=true
export FLOTORCH_LOG_PROVIDER="console"  # or "file"
export FLOTORCH_LOG_FILE="flotorch_logs.log"  # if provider is "file" (default: "flotorch_logs.log")

For comprehensive logging configuration details, see the SDK Logging Configuration guide.

Quick Start

Section titled “Quick Start”

Basic Agent Setup

Section titled “Basic Agent Setup”
from flotorch.adk.agent import FlotorchADKAgent
from google.adk import Runner


# Initialize the agent manager with your FloTorch Console configuration
agent_manager = FlotorchADKAgent(
    agent_name="your-agent-name",  # Must match the agent name in FloTorch Console
    custom_tools=[your_tool],      # Optional: Add custom tools
    base_url="https://gateway.flotorch.cloud",
    api_key="your_api_key"
    # Note: Agent description(goal) and instructions(system_prompt) should be configured
    # in the FloTorch Console when creating the agent
)


# Get the configured agent
agent = agent_manager.get_agent()


# Execute the agent with Runner
runner = Runner(agent=agent)

Agent Creation: FloTorch vs ADK

Section titled “Agent Creation: FloTorch vs ADK”

The following comparison illustrates the simplified configuration approach offered by FloTorch ADK:

FloTorch ADK Agent

Section titled “FloTorch ADK Agent”
# Configuration managed centrally in FloTorch Console
agent_manager = FlotorchADKAgent(
    agent_name="weather-agent",     # Reference agent from Console
    custom_tools=[weather_tool],    # Optional custom tools
    base_url="https://gateway.flotorch.cloud",
    api_key="<your_api_key>"
    # Note: Agent description(goal) and instructions(system_prompt) should be configured
    # in the FloTorch Console during agent creation
)


agent = agent_manager.get_agent()
runner = Runner(agent=agent)

ADK Agent

Section titled “ADK Agent”
# All configuration defined in code
agent = Agent(
    name="weather_agent",
    model=your_llm_model,
    description="A helpful weather assistant...",
    instruction="You are a professional...",
    tools=[weather_tool]
)


runner = Runner(agent=agent)

Core Components

Section titled “Core Components”

FlotorchADKAgent

Section titled “FlotorchADKAgent”

The main entry point for loading agent configurations from the FloTorch Console. It provides fully configured, ADK-compatible agents with minimal code setup.

FlotorchADKLLM

Section titled “FlotorchADKLLM”

An ADK-compatible LLM wrapper that integrates with FloTorch Gateway for model inference, providing seamless access to managed language models.

Memory Services

Section titled “Memory Services”
  • FlotorchMemoryService - Traditional keyword-based memory search and storage
  • FlotorchADKVectorMemoryService - Semantic vector-based memory search for advanced retrieval

FlotorchADKSession

Section titled “FlotorchADKSession”

Manages persistent session storage through FloTorch Gateway, enabling conversation continuity across multiple interactions.

Next Steps

Section titled “Next Steps”

Explore the detailed documentation for each component:

References

Section titled “References”