Skip to content

Logging, Tracing, and Observability

OpenTelemetry-compatible LLM tracing, structured log schema, error reporting, prompt-versioned observability.

Şükrü Yusuf KAYA
12 min read
Advanced
OpenTelemetry trace: gen-ai span'leri

Structured Log Şeması

Her LLM çağrısı için minimum log alanları:
{
  "ts": "...",
  "trace_id": "...",
  "span_id": "...",
  "prompt_id": "support_classifier",
  "prompt_version": "1.4.1",
  "model": "claude-sonnet-4-6",
  "temperature": 0.0,
  "max_tokens": 256,
  "input_tokens": 348,
  "output_tokens": 200,
  "cache_read_tokens": 300,
  "latency_ms": 740,
  "stop_reason": "end_turn",
  "tool_calls": [],
  "user_id_hash": "abc...",
  "eval_score": 0.91
}
python
from opentelemetry import trace
tracer = trace.get_tracer(__name__)
 
with tracer.start_as_current_span("claude.classify") as span:
    span.set_attribute("gen_ai.system", "anthropic")
    span.set_attribute("gen_ai.request.model", "claude-sonnet-4-6")
    span.set_attribute("gen_ai.request.temperature", 0.0)
    resp = client.messages.create(...)
    span.set_attribute("gen_ai.usage.input_tokens", resp.usage.input_tokens)
    span.set_attribute("gen_ai.usage.output_tokens", resp.usage.output_tokens)
    span.set_attribute("gen_ai.response.finish_reason", resp.stop_reason)
OpenTelemetry gen-AI semantik convention'ları ile span.
Boşluk doldur · text
Üretimde her LLM çağrısı için _____ ve prompt_version log'lanmalıdır. PII'ler her zaman _____ olarak yazılır. OpenTelemetry'nin yeni gen-AI namespace'i _____ ile başlar.
Quiz

Bu modülü değerlendirme zamanı

Buraya kadar öğrendiklerini quiz ile pekiştir. Süreli, puanlı ve geri bildirimli bir değerlendirmedir.

Quiz'e başla

Frequently Asked Questions

Any OTel-compatible (Datadog, Honeycomb, NewRelic, Tempo, Langfuse). The key is end-to-end pipeline tracing.

Yorumlar & Soru-Cevap

(0)
Yorum yazmak için giriş yap.
Yorumlar yükleniyor...

Related Content

10. Agent Design

What Is an Agent? Reactive vs Autonomous

Start Learning
10. Agent Design

Build Your First Agent with the Claude Agent SDK

Start Learning
1. Foundations — Welcome to Claude

What is Claude? The New Generation of AI Assistants

Start Learning

Connected pillar topics

Pillar topics this article maps to

Pillar Topic

LLMOps: Production-Grade LLM Operations

LLMOps is the engineering discipline that covers the development, deployment, monitoring, evaluation and cost management of LLM-powered applications — extending classic MLOps with prompt versioning, eval-driven CI and observability tailored for non-deterministic systems.