llm-app-patterns

To build a RAG pipeline, you need patterns for document ingestion, embedding, retrieval, and generation. This Skill provides production-ready architectures and code examples to implement Retrieval-Augmented Generation for grounding LLM responses in your custom data.

Designing agent architectures requires structured approaches like ReAct, function calling, and plan-and-execute. This Skill offers battle-tested patterns and code examples to deploy various agentic systems, including frameworks for multi-agent collaboration.

Prompt chaining connects sequential LLM calls where the output of one prompt feeds into the next. This Skill provides advanced prompt engineering techniques, including prompt templating, versioning, and chaining strategies to manage complex generation workflows.

Yes, these patterns are built for scalable, production-ready LLM applications. You can implement conversational interfaces with RAG pipelines to access company knowledge bases, utilizing LLMOps strategies like caching, rate limiting, and fallback mechanisms for reliable deployment.

Reliable LLM deployment requires observability, caching, rate limiting, and evaluation. This Skill integrates LLMOps monitoring patterns and fallback mechanisms to ensure robust, scalable performance for production environments.

Multi-agent collaboration suits complex tasks requiring specialized roles, while ReAct fits simpler reasoning and acting loops. This Skill provides architectural patterns for both, allowing you to choose the right agent architecture based on your application's task complexity.