litellm-adk 0.1.3

Production-grade multiservice Agent Development Kit

LiteLLM ADK (Agent Development Kit)

A production-grade, highly flexible multiservice Agent Development Kit.

Built for developers who need to swap models, API keys, and base URLs dynamically while maintaining a robust structure for tool usage, modular memory persistence, and observability.

Features

• Model Agnostic: Access 100+ LLMs (OpenAI, Anthropic, OCI Grok-3, Llama, etc.) seamlessly.
• Easy Tools: Register Python functions with the @tool decorator. No manual JSON schema management.
• Modular Memory: Native support for conversation persistence:
  • InMemoryMemory: Fast, ephemeral storage.
  • FileMemory: Simple JSON-based local persistence.
  • MongoDBMemory: Scalable, remote persistence.
• Parallel & Sequential Execution: Built-in support for parallel tool calls with robust stream accumulation.
• Dynamic Configuration: Global defaults via .env or per-agent/per-request overrides.
• Async & Streaming: Native support for ainvoke, stream, and astream.

Installation

pip install litellm-adk

Quick Start

Simple Conversational Agent

from litellm_adk.agents import LiteLLMAgent
from litellm_adk.memory import FileMemory

# Setup persistent memory
memory = FileMemory("chat_history.json")

agent = LiteLLMAgent(
    model="gpt-4", 
    memory=memory,
    session_id="user-123"
)

response = agent.invoke("My name is Alice.")
print(agent.invoke("What is my name?")) # Alice

Registering Tools

from litellm_adk.tools import tool

@tool
def get_weather(location: str):
    """Get the current weather for a location."""
    return f"The weather in {location} is sunny."

agent = LiteLLMAgent(tools=[get_weather])
agent.invoke("What is the weather in London?")

Configuration

The ADK uses pydantic-settings. Configure via .env:

• ADK_MODEL: Default model (e.g., gpt-4o).
• ADK_API_KEY: Default API key.
• ADK_BASE_URL: Global base URL override.
• ADK_LOG_LEVEL: DEBUG, INFO, etc.

Documentation

• Example: Basic Tools
• Example: Persistent Memory

License

MIT