parlot-instrumentation-langgraph

Parlot’s lead product is production voice observability on LiveKit. Use this package when you already run voice and want the same debugger for LangGraph / LangChain — same turn model, not a separate product.

Prefer parlot-instrumentation-livekit for voice agents.

Docs: LangGraph guide · Quick Start · parlot.ai

Install

pip install parlot-instrumentation-langgraph
# or: uv add parlot-instrumentation-langgraph
# or: pip install "parlot[langgraph]"

Instrument

from parlot.instrumentation.langgraph import parlotize

parlotize("support-bot", version="0.1.0", channel="webchat")

# Normal LangGraph usage — no per-call callbacks required
graph.invoke(
    {"messages": [("user", "Hello")]},
    config={"configurable": {"thread_id": "demo-1"}},
)

Pass channel (e.g. webchat, sms) when LangGraph owns the session. If thread_id is omitted, Parlot mints one automatically. When used inside a LiveKit agent that already called LiveKit parlotize(), this nests GenAI spans under that session — see the LangGraph guide.

Configure

export PARLOT_ENDPOINT=https://ingest.parlot.ai
export PARLOT_API_KEY=<org-scoped-key>

Mint the API key in Parlot Settings → API Keys. Session logs and generative AI content capture are configured in Settings or via parlotize(...) kwargs — see the LangGraph guide.

Next

• LangGraph guide
• Concepts
• Example: examples/langgraph/minimal-agent/