Skill UI

Guide for instrumenting LangChain applications with LangSmith tracing, Prometheus metrics, OpenTelemetry spans, structured logging, Grafana dashboards, and alerting to monitor health and performance in staging or production.

Guides setup of Langfuse tracing for LLM pipelines, covering automatic OpenAI wrapping, manual spans, streaming responses, and parent-child trace linking for end-to-end observability.

Provides a minimal Langfuse trace example that initializes the client, records spans and generations, flushes data, and logs a trace URL—ideal when starting integrations, validating setup, or learning trace patterns.

Guides Langfuse SDK best practices—singleton clients, trace lifecycle updates, nested spans, Python decorators, session tracking, and scoring—for reliable LLM observability.

This skill manages the schema for human feedback (annotation configs) within the Arize platform. It allows users to define label types (categorical, continuous, freeform) and programmatically apply these annotations to project spans using the Python SDK. Essential for establishing standardized human review processes and improving model evaluation.

This skill generates highly precise deep links to various resources within the Arize platform, including traces, spans, sessions, datasets, and evaluators. It requires collecting specific resource IDs (e.g., project_id, dataset_id) and managing time ranges (startA/endA) to ensure the user is taken to the exact view, optimizing the user experience for ML practitioners and data scientists.

Defines and runs LLM-as-judge evaluators on Arize via ax commands, handling evaluator creation, continuous monitoring, column mapping, trigger runs, and evaluation spans or experiments with configurable templates, models, and query filters.

Downloads spans, traces, and sessions from Arize via the ax CLI to debug LLM applications, using IDs, filters, sampling rules, and bulk/export safeguards while writing output to .arize-tmp-traces.

Phoenix CLI streamlines debugging Phoenix-based LLM observability: it fetches traces, spans, sessions, datasets, experiments, prompts, and GraphQL data while piping JSON to tools like jq so you can triage errors, inspect performance, and verify deployments quickly.

Guides hooking Phoenix AI services with OpenInference tracing, covering setup, spans, instrumentation, and production deployment for LLM workflows using Python or TypeScript.