opentelemetry-instrumentation-watsonx

What it is and what it does

This package instruments IBM Watsonx and Watson Machine Learning library calls to emit OpenTelemetry spans, capturing prompts, completions, and embeddings automatically. It hooks into the official IBM libraries and exports trace data to any OpenTelemetry-compatible backend (Jaeger, Datadog, etc.) for centralized observability of LLM application behavior.

By default it logs full prompt and completion content to span attributes for visibility, but you can disable this via the TRACELOOP_TRACE_CONTENT environment variable to protect sensitive user data. It depends on opentelemetry-api, opentelemetry-instrumentation, and AI-specific semantic conventions to standardize how LLM interactions are recorded.

Use it for:

• Debug LLM application behavior by inspecting prompts and completions in a centralized trace backend.
• Monitor token usage, latency, and error rates across Watsonx API calls in production.
• Correlate LLM requests with upstream application traces for end-to-end observability.
• Evaluate output quality by reviewing prompts and completions alongside application context in traces.
• Audit LLM interactions for compliance by exporting trace data to a secure backend.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Automatically captures and traces IBM Watsonx prompts, completions, and embeddings to OpenTelemetry spans for observability and debugging of LLM applications.

Yes, if you use IBM Watsonx or Watson Machine Learning and need observability. The package is actively maintained, has no known vulnerabilities, and low install friction. Be aware that by default it logs full prompt and completion content—set TRACELOOP_TRACE_CONTENT=false if you handle sensitive user data.

Install

opentelemetry-instrumentation-watsonx on PyPI

pip install opentelemetry-instrumentation-watsonx

uv add opentelemetry-instrumentation-watsonx

poetry add opentelemetry-instrumentation-watsonx

Installing opentelemetry-instrumentation-watsonx

Before you install

Low friction install with four stable runtime dependencies. Package is actively maintained with a recent release and no known vulnerabilities.

License in practice

Licensed under Apache-2.0 (permissive), allowing use in commercial and private projects with minimal restrictions.

Quickstart

pip install opentelemetry-instrumentation-watsonx

from opentelemetry.instrumentation.watsonx import WatsonxInstrumentor

WatsonxInstrumentor().instrument()

Requires Python 3.10 or later. Requires an OpenTelemetry exporter configured to receive spans (e.g., via OTEL_EXPORTER_OTLP_ENDPOINT).

Verify before relying

• Whether the instrumentation works with all versions of the IBM Watson Machine Learning library and watsonx.ai library, or only specific versions.
• Performance overhead of tracing on typical LLM request latency.
• Whether TRACELOOP_TRACE_CONTENT=false fully prevents sensitive data leakage or only suppresses span attributes.

Package facts

Evidence: opentelemetry_instrumentation_watsonx-0.62.3-py3-none-any.whl

Tags

More Monitoring packages