DeepEval OpenTelemetry Skill

Overview

The deepeval-otel Agent Skill teaches your coding agent how to export raw OpenTelemetry traces from an AI application to Confident AI's Observatory — no deepeval package required. Because the mechanism is just OTLP attribute keys plus an exporter endpoint, it works with any OpenTelemetry SDK in any language, not only Python. It ships in the confident-ai/deepeval repository alongside the deepeval and deepeval-tracing skills.

The job is exactly two things: point an OTLP/HTTP span exporter at the correct Confident AI endpoint, and set the confident.span.* and confident.trace.* attributes Confident AI reads off each span. Parent/child nesting comes from native OpenTelemetry span context.

There is one OTLP/HTTP traces endpoint per region, authenticated with the x-confident-api-key header:

The skill picks the endpoint from your API key's prefix — only confident_eu_... keys use the EU endpoint.

When It Triggers

The skill activates on prompts like:

Send our OpenTelemetry traces to Confident AI.
Wire our OTel Collector to export AI spans to Confident AI's Observatory.
Set confident.span.type on the LLM spans in our TypeScript agent.
Which Confident AI OTLP endpoint should our EU deployment use?

Installation

npx skills add confident-ai/deepeval --skill "deepeval-otel"

/plugin marketplace add confident-ai/deepeval
/plugin install deepeval@deepeval-plugins
/reload-plugins

git clone https://github.com/confident-ai/deepeval
cp -r deepeval/skills/deepeval-otel .claude/skills/

Prerequisites

• An exported CONFIDENT_API_KEY
• An OpenTelemetry SDK for your language (Python: opentelemetry-sdk and opentelemetry-exporter-otlp-proto-http)

What Changes in Your Codebase

1. Ask for OTLP export

   The agent checks for an existing TracerProvider, span exporters, or an OpenTelemetry Collector and repoints what exists rather than adding a parallel pipeline.

   Send our OpenTelemetry traces to Confident AI's Observatory.

2. Wire the exporter

   The agent points an OTLP/HTTP span exporter at the region endpoint, authenticated with the x-confident-api-key header — or, with no code changes at all, through the standard environment variables:

   export OTEL_EXPORTER_OTLP_ENDPOINT="https://otel.confident-ai.com"
   export OTEL_EXPORTER_OTLP_HEADERS="x-confident-api-key=<CONFIDENT_API_KEY>"

   See what the agent writes

   import os
   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
   
   provider = TracerProvider()
   provider.add_span_processor(
       BatchSpanProcessor(
           OTLPSpanExporter(
               endpoint="https://otel.confident-ai.com/v1/traces",
               headers={"x-confident-api-key": os.environ["CONFIDENT_API_KEY"]},
           )
       )
   )
   trace.set_tracer_provider(provider)

3. Set the confident.* attributes

   The agent sets confident.span.* on each AI span and confident.trace.* for trace-wide fields, honoring OTLP data-type rules (JSON-encode dicts, native arrays for string lists) and falling back to standard gen_ai.* semantic conventions where the app already emits them.

   See what the agent writes

   tracer = trace.get_tracer(__name__)
   
   with tracer.start_as_current_span("my-llm-app") as span:
       span.set_attribute("confident.span.type", "agent")
       span.set_attribute("confident.trace.name", "my-llm-app")

4. Keep non-AI spans out

   When the process also runs auto-instrumentation or an APM agent (HTTP/DB spans, Datadog, and the like), the agent sets up a dedicated pipeline or span filter so only AI spans reach Confident AI — then verifies the traces render in the Observatory.

FAQs

Next Steps

Last updated on August 14, 2026