Skip to main content

LangChain Agent

Using WaxellLangChainHandler to capture LLM calls within LangChain chains.

Additional dependencies
pip install waxell-observe[langchain] langchain-openai
Environment variables

This example requires OPENAI_API_KEY, WAXELL_API_KEY, and WAXELL_API_URL.

from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from waxell_observe.integrations.langchain import WaxellLangChainHandler

handler = WaxellLangChainHandler(agent_name="langchain-qa")

llm = ChatOpenAI(model="gpt-4o")

# Build prompts
analyze_prompt = ChatPromptTemplate.from_template(
    "Analyze the key themes in: {topic}"
)
summarize_prompt = ChatPromptTemplate.from_template(
    "Summarize this analysis in 2 sentences: {analysis}"
)

# Step 1: Analyze
analysis = (analyze_prompt | llm).invoke(
    {"topic": "The impact of large language models on software engineering"},
    config={"callbacks": [handler]},
)

# Step 2: Summarize
summary = (summarize_prompt | llm).invoke(
    {"analysis": analysis.content},
    config={"callbacks": [handler]},
)

# Flush telemetry
handler.flush_sync(result={"summary": summary.content})
print(f"Run ID: {handler.run_id}")

What this demonstrates​

  • LangChain callback handler -- WaxellLangChainHandler plugs into LangChain's callback system to capture every LLM call.
  • Chain composition -- prompt-template-to-LLM chains connected via | (LCEL) are traced automatically.
  • Automatic LLM call capture -- input prompts, model parameters, and completions are recorded without extra code.
  • flush_sync -- ensures all telemetry is sent before the process exits, and attaches a final result to the trace.

Run it​

export OPENAI_API_KEY="sk-..."
export WAXELL_API_KEY="your-waxell-api-key"
export WAXELL_API_URL="https://api.waxell.ai"

python langchain_agent.py