Instrumentation

Overview

respan-tracing auto-instruments supported libraries to capture LLM calls, vector DB queries, and framework operations as spans — no code changes required. When a supported library is installed, its calls are automatically traced.

Configure

1
from respan_tracing import RespanTelemetry, Instruments

Enable all (default)

1
telemetry = RespanTelemetry(api_key="your-api-key")
2
# All installed libraries are auto-instrumented

Enable specific instruments only

1
telemetry = RespanTelemetry(
2
    api_key="your-api-key",
3
    instruments={Instruments.OPENAI, Instruments.ANTHROPIC},
4
)

Block specific instruments

1
telemetry = RespanTelemetry(
2
    api_key="your-api-key",
3
    block_instruments={Instruments.REQUESTS, Instruments.URLLIB3},
4
)

Disable all auto-instrumentation

1
telemetry = RespanTelemetry(
2
    api_key="your-api-key",
3
    instruments=set(),
4
)

Disable auto-instrumentation (when using plugins)

When using instrumentation plugins with the Respan entry point, set auto_instrument=False on RespanTelemetry to avoid duplicate spans:

1
from respan import Respan, propagate_attributes
2
from respan_instrumentation_openai_agents import OpenAIAgentsInstrumentor
3
4
respan = Respan(
5
    api_key="your-api-key",
6
    auto_instrument=False,
7
    instrumentations=[OpenAIAgentsInstrumentor()],
8
)

Available instruments

AI/ML providers

Cloud AI services

Vector databases

Frameworks

Infrastructure

Environment variables

Troubleshooting

If you see warnings like Failed to initialize Requests instrumentation, install the optional OpenTelemetry instrumentations:

$
pip install opentelemetry-instrumentation-requests opentelemetry-instrumentation-urllib3

This is optional — tracing works without them. Only needed if your app uses requests or urllib3 and you want those calls traced.