Send traces with OpenInference

This guide sends traces from an application instrumented with OpenInference, which has instrumentations for OpenAI, Anthropic, LangChain, LlamaIndex, DSPy and other libraries. Tracing reads the model, tokens and messages of LLM spans, and the input and output of chain, tool and retriever spans.

Before you start

You need an API key with the traces:write scope. Create one under Settings in the dashboard, or run div login.

Send traces

Install the OpenTelemetry SDK, the OTLP exporter and the instrumentation for your library. This example uses OpenAI’s.
```
pip install opentelemetry-sdk opentelemetry-exporter-otlp-proto-http openinference-instrumentation-openai
```

Set the exporter’s endpoint and key.

export OTEL_EXPORTER_OTLP_ENDPOINT=https://api.divergentlabs.xyz/api/otlp
export OTEL_EXPORTER_OTLP_HEADERS="Authorization=Bearer%20$DIVERGENT_API_KEY"
export OTEL_SERVICE_NAME=my-app

Set up the tracer provider and instrument the library.

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from openinference.instrumentation.openai import OpenAIInstrumentor

provider = TracerProvider()
provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter()))
trace.set_tracer_provider(provider)
OpenAIInstrumentor().instrument(tracer_provider=provider)

Run your application, then list the traces.
```
div traces --since 10m
```

Sessions and users

OpenInference’s using_session and using_user context managers set session.id and user.id, which tracing reads:

from openinference.instrumentation import using_session, using_user

with using_session(session_id), using_user(user_id):
    client.chat.completions.create(model="gpt-4o-mini", messages=messages)

Keep content out of traces

Set OPENINFERENCE_HIDE_INPUTS=true and OPENINFERENCE_HIDE_OUTPUTS=true.