Support using an external OpenTelemetry trace provider

[berkegocmen]

Hello,

I would like to use traceloop-sdk alongside an external tracing library that installs its own OpenTelemetry-compatible TracerProvider.

For example, Datadog’s Python tracing library, ddtrace, installs its own provider. It creates spans and manages distributed trace context, but intentionally does not expose the OpenTelemetry SDK method add_span_processor() because it handles export, sampling, and propagation internally.

Currently, Traceloop.init() requires the global provider to support add_span_processor(). Initializing Traceloop after such a provider therefore fails with:

Cannot add span processor to the default provider since it doesn't support it

Initializing Traceloop first is also not a solution, because it creates a separate provider and trace context from the external tracing library.

Would you consider supporting an “existing provider” or “instrument-only” initialization mode? Conceptually:

from opentelemetry import trace
from traceloop.sdk import Traceloop

Traceloop.init(
    tracer_provider=trace.get_tracer_provider(),
    instrument_only=True,
)

Expected behavior:

• Reuse the supplied or global provider for span creation.
• Do not call add_span_processor().
• Do not configure an exporter or replace the global provider.
• Let the external tracing library own export, sampling, and W3C propagation.

This would make Traceloop compatible with external OpenTelemetry tracing implementations while keeping all spans in one distributed trace.

Would this be possible?