Skill UI

This skill provides a comprehensive framework for optimizing, improving, and debugging Large Language Model (LLM) prompts. It leverages production trace data, system evaluations, and human annotations to extract prompts from spans. Users can run a data-driven optimization loop, ensuring the AI responses meet desired quality standards. Ideal for prompt tuning, system prompt refinement, and improving overall AI output quality.

This skill provides comprehensive tools to download, export, and deeply inspect existing Arize traces and spans. It is essential for understanding the full execution flow of complex LLM applications, debugging runtime errors, and performing root-cause analysis. Users can retrieve detailed data filtered by Trace ID, Span ID, or Session ID, enabling thorough analysis of conversational or multi-step agent behaviors.

The Phoenix CLI is a powerful tool for debugging complex LLM and agent workflows. It allows users to fetch detailed traces, spans, and sessions, analyze errors, and structure findings using open and axial coding techniques. It is essential for investigating complex LLM failures, building robust failure taxonomies, and guiding the next steps in application evaluation after initial instrumentation.

Phoenix Evals lets teams build evaluators for AI/LLM apps by wiring deterministic code with LLM nuance, running Python or TypeScript workloads, validating against humans, and iterating through error analysis, RAG, and production workflows.

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

A persistent, graph-based knowledge base designed to accumulate structured knowledge throughout the entire research lifecycle. It stores and connects diverse entities—including published papers, novel ideas, conducted experiments, and testable claims. Inspired by LLM Wiki patterns, it ensures knowledge compounding, meaning the wiki becomes progressively smarter with every piece of input, rather than requiring re-derivation for every query.

Modal is a powerful serverless cloud platform designed specifically for running AI and Machine Learning workloads. It allows users to deploy, serve, and scale complex models (like LLMs) without managing infrastructure. Key features include on-demand access to high-end GPUs (A100, H100), serverless function scaling, persistent storage, and the ability to define entire infrastructure using pure Python code, eliminating the need for separate YAML or Dockerfiles.

This guide provides comprehensive strategies to significantly reduce spending on the Anthropic Claude API. Key techniques include model routing (selecting the most cost-effective model for the task), prompt caching for repeated prompts, using Message Batches for asynchronous tasks, and implementing real-time spend tracking and budget alerts. Essential for managing large-scale, high-cost AI deployments and optimizing cloud LLM expenditures.

Automates comprehensive regression testing for all enabled cloud Large Language Model (LLM) providers. This includes verifying core connectivity, standard chat completion, structured JSON output, and Server-Sent Events (SSE) streaming capabilities. Essential for ensuring the reliability and stability of multi-cloud AI integrations in a CI/CD pipeline.

A comprehensive guide providing minimal working examples for core Cohere API v2 endpoints, including Chat completion, text embedding, and search reranking. This resource details basic setup, usage patterns, and includes runnable code examples in both TypeScript and Python, making it perfect for new AI integrations or learning advanced LLM functionality.

A comprehensive security auditing framework designed to systematically detect sensitive information disclosure from Large Language Models (LLMs). It uses multi-phase, escalating dialogue probes to test for leaks such as system prompts, API keys, PII, and internal configurations. Essential for red teaming and rigorous model security assessment.

This skill provides advanced sentiment analysis tailored for financial news and market texts. It supports two modes: high-speed, localized analysis using FinBERT, and advanced, reasoning-based analysis via LLM. Users can determine the sentiment polarity (positive, negative, or neutral), score, and specific reason for financial content, making it ideal for market intelligence and data enrichment.