Same agent Smaller bill. The tools and infrastructure to get more work from your AI. Fewer tokens. Less compute. Lower cost.
TypeScript agent-harnessagent-harness-engineeringai-engineeringai-observabilityamd-gpuclaude-codeclickhousecodexcoding-agentcoding-agentscursorgenaigpu-monitoringllm-observabilityllmopsnvidia-smiobservabilityopen-sourceopenaiopentelemetry
"""LangGraph agent sample — instrumented with openlit.init() only."""
import os
import time
from pathlib import Path
from typing import Annotated, TypedDict
from dotenv import load_dotenv
load_dotenv(Path(__file__).resolve().parents[2] / ".env")
import openlit
if not os.environ.get("OPENAI_API_KEY"):
raise SystemExit("OPENAI_API_KEY missing — set it in the repo root .env")
OTLP_ENDPOINT = os.environ.get(
"OTEL_EXPORTER_OTLP_ENDPOINT", "http://127.0.0.1:4318"
)
SERVICE_NAME = os.environ.get("OPENLIT_SERVICE_NAME", "langgraph-agent-app")
ENVIRONMENT = os.environ.get("OPENLIT_ENVIRONMENT", "default")
INTERVAL = int(os.environ.get("REQUEST_INTERVAL_SECONDS", "30"))
MODEL = os.environ.get("OPENAI_MODEL", "gpt-4o-mini")
openlit.init(
service_name=SERVICE_NAME,
environment=ENVIRONMENT,
otlp_endpoint=OTLP_ENDPOINT,
)
from langchain_core.messages import HumanMessage # noqa: E402
from langchain_openai import ChatOpenAI # noqa: E402
from langgraph.graph import END, START, StateGraph # noqa: E402
from langgraph.graph.message import add_messages # noqa: E402
class AgentState(TypedDict):
messages: Annotated[list, add_messages]
llm = ChatOpenAI(model=MODEL, temperature=0.2)
def researcher(state: AgentState) -> dict:
"""First node: gather a short research note."""
response = llm.invoke(
[
{
"role": "system",
"content": (
"You are a research node. Reply with one short factual sentence."
),
},
*state["messages"],
]
)
return {"messages": [response]}
def summarizer(state: AgentState) -> dict:
"""Second node: compress the research into a tweet-length summary."""
response = llm.invoke(
[
{
"role": "system",
"content": (
"You are a summarizer. Turn the prior research into one tweet "
"(under 200 characters)."
),
},
*state["messages"],
]
)
return {"messages": [response]}
builder = StateGraph(AgentState)
builder.add_node("researcher", researcher)
builder.add_node("summarizer", summarizer)
builder.add_edge(START, "researcher")
builder.add_edge("researcher", "summarizer")
builder.add_edge("summarizer", END)
graph = builder.compile()
PROMPTS = [
"Interesting fact about LLM observability in production.",
"Why agent traces help debug tool-calling failures.",
"One benefit of OpenTelemetry for multi-agent systems.",
]
print(
f"[langgraph-agent-app] starting → OTLP {OTLP_ENDPOINT} "
f"service={SERVICE_NAME} every {INTERVAL}s"
)
i = 0
while True:
prompt = PROMPTS[i % len(PROMPTS)]
try:
result = graph.invoke({"messages": [HumanMessage(content=prompt)]})
last = result["messages"][-1]
text = getattr(last, "content", str(last))[:120]
print(f"[{i}] -> {text}")
except Exception as exc: # pylint: disable=broad-except
print(f"[{i}] Error: {exc}")
i += 1
time.sleep(INTERVAL)