LangDB ADK - Agent Development Kit for LangDB

A comprehensive Google ADK (Agent Development Kit) integration package for LangDB. This package provides seamless integration of LangDB's LLM capabilities, enhanced agent functionality, and distributed tracing into the Google ADK framework.

Features

• LangDBLlm: Native integration with Google ADK's BaseLlm interface
• LangDBAgent: Enhanced agent class with automatic LangDB callbacks and session management
• LangDBTracing: Distributed tracing integration with OpenTelemetry
• Session tracking and thread management across agent invocations
• Comprehensive callback system for model, agent, and tool interactions

Installation

pip install langdb_adk

Usage with Google ADK

Basic Usage with LangDBAgent

import os
import asyncio
from langdb_adk import LangDBLlm, LangDBAgent, LangDBTracing
from google.adk.runners import InMemoryRunner

# Initialize tracing (optional but recommended)
LangDBTracing(collector_endpoint="https://api.staging.langdb.ai:4317")

async def main():
    # Create a LangDB agent with automatic callbacks and session management
    agent = LangDBAgent(
        model=LangDBLlm(model="openai/gpt-4.1"),
        name="my_agent",
        description="A simple agent with LangDB integration",
        instruction="You are a helpful assistant."
    )
    
    # Create a runner to execute the agent
    runner = InMemoryRunner()
    
    # Run the agent with a prompt
    response = await runner.run(agent, "Hello, how are you?")
    print(response.text)

if __name__ == "__main__":
    asyncio.run(main())

Basic Usage with Standard Agent (Legacy)

from langdb_adk import LangDBLlm, LangDBAgent
from langdb_adk import LangDBTracing
LangDBTracing()


root_agent = LangDBAgent(
    model=LangDBLlm(model="openai/gpt-4.1"),
    name="root_agent",
    description="A Travel Conceirge using the services of multiple sub-agents",
    instruction="You are a travel concierge that coordinates with specialist agents.",
)

Complete Example with Sub-Agents and Tracing

from langdb_adk import LangDBLlm, LangDBAgent, LangDBTracing

# Initialize tracing with LangDB collector
LangDBTracing()

# Create sub-agents for different capabilities
planning_agent = LangDBAgent(
    model=LangDBLlm(model="openai/gpt-4.1"),
    name="planning_agent",
    description="Handles travel planning tasks",
    instruction="You help users plan their travel itineraries."
)

booking_agent = LangDBAgent(
    model=LangDBLlm(model="openai/gpt-4.1"),
    name="booking_agent", 
    description="Handles travel booking tasks",
    instruction="You help users book flights, hotels, and activities."
)

# Create main agent with sub-agents
root_agent = LangDBAgent(
    model=LangDBLlm(model="openai/gpt-4.1"),
    name="root_agent",
    description="A Travel Concierge using multiple sub-agents",
    instruction="You are a travel concierge that coordinates with specialist agents.",
    sub_agents=[planning_agent, booking_agent]
)

With MCP Servers

# Configure MCP servers for LangDB
mcp_servers = [
    {
        "server_url": "server_url",
        "type": "sse",
        "name": "search",
        "description": "Web search capabilities via DuckDuckGo"
    }
]


    # Initialize the LangDB LLM with MCP servers
llm = LangDBLlm(
        model="anthropic/claude-sonnet-4",
        api_key=os.getenv("LANGDB_API_KEY"),
        project_id=os.getenv("LANGDB_PROJECT_ID"),
        mcp_servers=mcp_servers
    )
    
    # Create a LangDB agent with MCP capabilities
root_agent = LangDBAgent(
        model=llm,
        name="search_agent",
        description="Agent with web search capabilities")

Configuration

Environment Variables

• LANGDB_API_KEY: Your LangDB API key (required)
• LANGDB_PROJECT_ID: Your LangDB project ID (required)
• LANGDB_TRACING_BASE_URL: Custom tracing collector endpoint (optional, defaults to https://api.us-east-1.langdb.ai:4317)

LangDBLlm Parameters

• model: The name of the model to use (e.g., "anthropic/claude-sonnet-4", "openai/gpt-4.1")
• api_key: Your LangDB API key (defaults to LANGDB_API_KEY env var)
• project_id: Your LangDB project ID (defaults to LANGDB_PROJECT_ID env var)
• extra_headers: Additional headers to include in requests
• mcp_servers: List of MCP server configurations for extended capabilities

LangDBTracing Parameters

• collector_endpoint: OpenTelemetry collector endpoint (defaults to LANGDB_TRACING_BASE_URL or https://api.us-east-1.langdb.ai:4317)
• api_key: Your LangDB API key (defaults to LANGDB_API_KEY env var)
• project_id: Your LangDB project ID (defaults to LANGDB_PROJECT_ID env var)
• client_name: Client identifier for tracing (defaults to "adk" for LangDBAdkTracing)

Advanced Features

Session Management

LangDBAgent automatically manages session tracking across agent invocations:

• Thread ID: Maintains consistent session IDs across agent calls within the same conversation
• Invocation Tracking: Tracks sequence of invocations for debugging and analytics
• State Persistence: Maintains state across callbacks and sub-agent interactions

Distributed Tracing

LangDBTracing provides comprehensive observability through OpenTelemetry:

• Automatic Span Creation: Traces all agent, model, and tool interactions
• Attribute Mapping: Maps ADK attributes to LangDB-specific attributes
• Session Correlation: Links spans across different agents using consistent thread IDs
• Run ID Tracking: Provides unique run identifiers for each execution

Custom Tracing Setup

from langdb_adk import LangDBTracing

# Use default endpoints and credentials from environment
tracing = LangDBTracing()

# Or configure custom settings
tracing = LangDBTracing(
    collector_endpoint="https://custom-collector.example.com:4317",
    api_key="your-api-key",
    project_id="your-project-id",
    client_name="adk"
)

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Acknowledgements

• Google ADK Team for the Agent Development Kit
• OpenTelemetry community for distributed tracing standards