Plott SDK for LangGraph

Automatic analytics tracking for LangGraph applications. Just wrap your graph and analytics are collected transparently.

Installation

pip install plott-sdk-langgraph

Quick Start

from langgraph.graph import StateGraph
from langchain_core.messages import HumanMessage
from plott_langgraph import plott_tracked_graph

# Create and compile your graph
graph = workflow.compile(checkpointer=memory)

# Wrap with Plott - that's it!
tracked_graph = plott_tracked_graph(graph, {
    "api_key": "cpk_...",  # or set PLOTT_API_KEY env var
    "environment": "development",
})

# Use exactly like the original graph - analytics are automatic
async for event in tracked_graph.astream_events(
    {"messages": [HumanMessage(content="Hello!")]},
    config={"configurable": {"thread_id": "123", "run_id": "run-456"}},
):
    # Your normal event handling
    if event["event"] == "on_chat_model_stream":
        print(event["data"]["chunk"].content, end="", flush=True)

What Gets Tracked

The SDK automatically tracks:

Configuration

tracked_graph = plott_tracked_graph(graph, {
    "api_key": "cpk_...",           # Required (or PLOTT_API_KEY env var)
    "environment": "production",     # production, staging, development, test
    "retry_attempts": 3,             # Number of retry attempts on failure
    "retry_delay": 1.0,              # Seconds between retries
    "debug": False,                  # Enable debug logging
})

Environment Variables

• PLOTT_API_KEY - API key (if not provided in config)

Tracked Methods

The following methods have analytics tracking:

• astream_events() - Full event streaming with detailed tracking
• astream() - State streaming with lifecycle tracking
• invoke() - Synchronous invocation with lifecycle tracking
• ainvoke() - Async invocation with lifecycle tracking

License

Apache 2.0