Skill UI

An expert guide and framework for building production-grade, sophisticated AI agent systems. This skill specializes in utilizing the latest LangChain 0.1+ and LangGraph APIs to implement multi-agent orchestration, advanced Retrieval-Augmented Generation (RAG) pipelines (e.g., RAG Fusion, HyDE), and complex state management. It covers comprehensive topics from architectural design (Plan-and-Execute, ReAct Agents) to production deployment (FastAPI streaming, LangSmith observability, caching, and robust error handling).

This skill simulates an expert prompt engineer, specializing in optimizing prompts for Large Language Models (LLMs). It applies advanced techniques such as Constitutional AI, Chain-of-Thought reasoning, and model-specific tuning. Use this tool to transform basic instructions into production-ready, high-performance prompts, significantly boosting accuracy, reducing hallucinations, and optimizing token usage.

Master comprehensive strategies for evaluating Large Language Model (LLM) applications. This skill covers automated metrics (BLEU, ROUGE, BERTScore), structured human evaluation dimensions, and advanced LLM-as-Judge techniques. Use it to systematically measure model performance, compare prompts, detect regressions, and build confidence in production systems.

Practical guide to integrating Claude models: covers API client usage, streaming responses, tool calls, RAG pipelines, document chunking, cost-aware model selection, and anti-pattern warnings for efficient LLM workflows.

A comprehensive workflow for advanced AI applications using Mistral AI. This guide covers generating text and code embeddings (1024 dimensions), implementing semantic search via cosine similarity, and building robust RAG pipelines. It also demonstrates advanced tool augmentation using function calling, allowing LLMs to interact with external APIs and databases.

This comprehensive guide provides minimal working examples for integrating with the Mistral AI API. It demonstrates core LLM functionalities including basic chat completion, streaming responses, multi-turn conversations, and generating structured JSON output. Use this guide when starting a new Mistral integration, testing API connections, or learning advanced AI patterns using TypeScript or Python.

This guide details advanced methods for merging multiple specialized LLMs (e.g., combining math, coding, and chat capabilities) without the need for expensive retraining. Learn various techniques such as Linear Merging, SLERP, Task Arithmetic, TIES-Merging, and DARE. Use mergekit to create high-performing, specialized model variants quickly, significantly improving performance and reducing computational costs.

Compress large language models quickly using Wanda, SparseGPT, and N:M pruning to enforce hardware-friendly sparsity, enabling faster inference and deployment without retraining.

NVIDIA’s NeMo Guardrails adds runtime safety rails to LLM apps, covering jailbreak detection, hallucination and fact-checking, PII filtering, and toxicity checks via Colang 2.0 rules and integrations for production use on T4 GPUs.

High-performance RLHF framework that combines Ray, vLLM, and ZeRO-3 to accelerate PPO/GRPO/RLOO/DPO training for 7B-70B+ models on distributed GPU clusters, providing a single-stack workflow for reward models and policy optimization.

This skill provides advanced strategies for managing large language model (LLM) context windows and token usage via OpenRouter. Functions include context window lookup, context-aware model selection, conversation trimming, document chunking for RAG, and prompt caching. Use this when dealing with long conversations or large documents to maximize quality while minimizing API costs and preventing context limit overflow.

Implements standardized function/tool calling across various large language models (LLMs) via OpenRouter. This unified approach allows developers to build sophisticated, multi-step agents and complex workflows, regardless of the underlying provider (e.g., GPT-4o, Claude 3.5). It simplifies the process of achieving structured output and tool-augmented reasoning in production systems.