Skill UI

LangSmith centralizes tracing, evaluation, and monitoring for LLM applications, letting teams debug prompts, run regression tests, collect latency/cost metrics, and integrate with OpenAI, Anthropic, LangChain, or LlamaIndex to keep production systems observable.

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).

A robust framework for building complex, long-horizon AI agents using LangGraph. It addresses common pitfalls in advanced agent design, such as unbounded state growth (Virtual FS cleanup) and subagent persona leakage (explicit message overriding). Ideal for tasks requiring multi-step planning, task delegation, and self-correction through reflection loops.

Master the complexities of building robust Retrieval Augmented Generation (RAG) pipelines using LangChain. This guide addresses common pitfalls in vector search, such as score semantics differences (L2 vs. Cosine), embedding dimension mismatches, and chunking issues for code/Markdown. Learn best practices for selecting between FAISS, Pinecone, Chroma, and PGVector, normalizing scores, and implementing hybrid keyword+vector search for maximum retrieval quality.

This harness provides comprehensive, reproducible evaluation pipelines for complex LLM chains and agents (LangChain/LangGraph 1.0). It integrates golden dataset management, LangSmith evaluation, RAGAS metrics, deepeval LLM-as-judge, and structured agent trajectory analysis. Use it when establishing quality benchmarks for new chains, diagnosing performance regressions after model switches, or implementing CI/CD gates to prevent quality drops.

Implement robust human-in-the-loop (HITL) approval gates for sophisticated LangGraph workflows. This skill enables the graph to pause execution before critical, irreversible actions (like sending emails or calling tools). It manages complex state serialization, provides clear resume semantics using `Command(resume=...)`, and integrates with external UIs (e.g., Slack) to await human decision, ensuring safe and controlled workflow execution.

This guide addresses critical LangGraph streaming pitfalls, including state overdraw, proxy buffering issues (Nginx, Cloud Run), and excessive event noise. It provides a production-grade pattern for implementing robust SSE/WebSocket endpoints using FastAPI. Learn how to correctly select the `stream_mode` (messages, updates, values) and implement server-side event filtering to reliably build live-token chat UIs and perform comprehensive state visualization.

Master advanced patterns for composing LangGraph subgraphs, which is essential for building robust, multi-agent systems like planner-executors or hierarchical teams. This skill addresses critical issues such as shared state contract propagation (P21) and callback scoping (P28). Learn to manage state across independent graph components using sophisticated dispatch patterns: compiled nodes, Send fan-out, and Command bubble-up.

This skill integrates LangChain 1.0 and LangGraph 1.0 into standard OpenTelemetry (OTEL) backends (e.g., Jaeger, Honeycomb, Tempo). It solves critical observability gaps by ensuring correct prompt content capture, propagating spans across complex subgraphs, and providing five LLM-specific Service Level Objectives (SLOs) like latency and cost. Ideal for multi-tenant, compliance-sensitive, or existing OTEL stack environments where LangSmith is unsuitable.

This skill provides advanced techniques for robust prompt engineering within LangChain applications. It solves common production issues, such as f-string template breakage when user input contains literal curly braces, and guides developers on consolidating scattered prompts. Crucially, it details how to use the LangSmith Prompt Hub to version, pin, and A/B test prompts using immutable commit hashes, ensuring stable, reproducible LLM deployments.