Skill UI

This guide outlines essential security practices for deploying production-grade LLM applications using LangChain. It covers secure secrets management, defending against prompt injection, implementing sandboxed tool execution, validating model outputs for sensitive data, and setting up comprehensive audit logging, ensuring robust and secure AI deployment.

This comprehensive guide details how to safely upgrade LangChain versions (e.g., from 0.1.x to 0.3.x+). It covers critical breaking changes across both TypeScript and Python implementations, including updating import paths, converting legacy LLMChain to the modern LCEL pipe syntax, updating agent creation, and migrating memory management. Essential for maintaining and modernizing existing LLM applications.

This comprehensive skill implements advanced event-driven patterns for LangChain applications. It provides custom callback handlers to observe every stage of chain execution (e.g., tool start/end, LLM token generation). Furthermore, it supports real-time data transmission via Server-Sent Events (SSE) and WebSockets, and integrates webhooks for centralized event logging and comprehensive tracing.

Track and improve production LLMs with Langfuse observability: trace API calls, manage prompts, score evaluations, integrate with LangChain/LlamaIndex/OpenAI, and keep dashboards on cost, latency, and quality.

Fetch execution traces from LangSmith Studio to inspect LangChain/LangGraph agent runs, analyze errors, tool calls, memory usage, and performance, and export sessions for deeper diagnostics using the langsmith-fetch CLI.

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 deep dive into handling typed content blocks (text, image, tool_use, thinking, document) in LangChain 1.0. This skill standardizes multi-modal message composition, tool-call iteration, and complex content handling across providers like Claude, GPT-4o, and Gemini, solving key API discrepancies.

This guide provides advanced patterns for composing robust, multi-step Large Language Model (LLM) workflows using LangChain 1.0 core runnables. Learn to implement parallel retrievals (RunnableParallel), conditional routing (RunnableBranch), and state management. It also addresses common pitfalls like silent KeyErrors due to mismatched dictionary shapes, providing solutions using debug probes and type annotations for reliable RAG, classification, and extraction pipelines.

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.