opentelemetry-instrumentation-haystack

What it is and what it does

This package integrates OpenTelemetry tracing into Haystack-based LLM applications, automatically capturing the flow of data through your pipeline. It instruments Haystack components to emit spans that record prompts, completions, and embeddings, giving you visibility into how your LLM application behaves at runtime.

By default, the instrumentation logs the actual content of prompts and responses to span attributes, which helps with debugging and quality evaluation but may expose sensitive user data. You can disable content logging by setting the TRACELOOP_TRACE_CONTENT environment variable to false, reducing both privacy risk and trace size. The instrumentation is activated with a single call to HaystackInstrumentor().instrument() after importing.

Use it for:

• Debug LLM application behavior by inspecting prompts, completions, and intermediate embeddings in distributed traces
• Monitor production Haystack pipelines by sending traces to a backend for centralized observability
• Evaluate LLM output quality by analyzing trace data to correlate inputs with outputs across your application
• Reduce trace storage costs by disabling content logging while preserving structural tracing information
• Correlate LLM operations with other services in a microservices architecture using OpenTelemetry's standard trace context

Worth the install?

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

Adds distributed tracing to Haystack LLM applications by instrumenting them with OpenTelemetry, capturing prompts, completions, and embeddings as span attributes for visibility and debugging.

Yes, if you are building or operating Haystack LLM applications and need observability. The low install friction, active maintenance, and permissive license make it a straightforward addition. Be aware that content logging is on by default—set TRACELOOP_TRACE_CONTENT=false in production if your prompts or completions contain sensitive data.

Install

opentelemetry-instrumentation-haystack on PyPI

pip install opentelemetry-instrumentation-haystack

uv add opentelemetry-instrumentation-haystack

poetry add opentelemetry-instrumentation-haystack

Installing opentelemetry-instrumentation-haystack

Before you install

Low install friction with four runtime dependencies, all from the OpenTelemetry ecosystem. Active maintenance with a release 4 days old and recent commits signal ongoing support.

License in practice

Apache-2.0 permissive license allows use in most commercial and open-source projects without significant restrictions.

Quickstart

pip install opentelemetry-instrumentation-haystack

from opentelemetry.instrumentation.haystack import HaystackInstrumentor

HaystackInstrumentor().instrument()

Requires Python 3.10 or later (supports_current); set TRACELOOP_TRACE_CONTENT environment variable to control whether prompts and completions are logged to spans.

Verify before relying

• Whether tracing overhead is acceptable for production workloads at scale
• Compatibility with specific Haystack versions beyond the stated Python 3.10+ requirement
• Performance impact when TRACELOOP_TRACE_CONTENT is enabled vs. disabled

Package facts

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

Tags

More Monitoring packages