Langfuse Via Openinference

import base64
import os

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.tools.duckduckgo import DuckDuckGoTools
from openinference.instrumentation.agno import AgnoInstrumentor
from opentelemetry import trace as trace_api
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import SimpleSpanProcessor

LANGFUSE_AUTH = base64.b64encode(
    f"{os.getenv('LANGFUSE_PUBLIC_KEY')}:{os.getenv('LANGFUSE_SECRET_KEY')}".encode()
).decode()
os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"] = (
    "https://us.cloud.langfuse.com/api/public/otel"  # 🇺🇸 US data region
)
# os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"]="https://cloud.langfuse.com/api/public/otel" # 🇪🇺 EU data region
# os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"]="http://localhost:3000/api/public/otel" # 🏠 Local deployment (>= v3.22.0)

os.environ["OTEL_EXPORTER_OTLP_HEADERS"] = f"Authorization=Basic {LANGFUSE_AUTH}"


tracer_provider = TracerProvider()
tracer_provider.add_span_processor(SimpleSpanProcessor(OTLPSpanExporter()))
trace_api.set_tracer_provider(tracer_provider=tracer_provider)

# Start instrumenting agno
AgnoInstrumentor().instrument()


agent = Agent(
    name="Stock Price Agent",
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[DuckDuckGoTools()],
    instructions="You are an internet search agent. Find and provide accurate information on any topic.",
    debug_mode=True,
)

agent.print_response("What are the latest developments in artificial intelligence?")