skills.rest
DucCuong159
DucCuong159COMMUNITY
Explore all skills by this author →
GitHub

llm-app-patterns

Implements RAG pipelines, agent architectures, and LLMOps patterns for LLM applications.

Updated Aug 11, 2026
Search Tags:#prompt engineering#llm#rag#ai agents#vector search#function calling#llmops
One-click install
npx skills add https://github.com/DucCuong159/Realtime-chatapp --skill llm-app-patterns-duccuong159
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-app-patterns
Source: https://github.com/DucCuong159/Realtime-chatapp/tree/main/.agent/skills/llm-app-patterns
Command: npx skills add https://github.com/DucCuong159/Realtime-chatapp --skill llm-app-patterns-duccuong159

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building production LLM applications requires solving recurring challenges like grounding responses in private data, orchestrating multi-step agent reasoning, controlling costs, and monitoring quality, and this Skill provides proven implementation patterns for each. ## Core Features & Use Cases - RAG Pipeline Design: Covers document chunking strategies, embedding model selection, vector database options (Pinecone, Weaviate, ChromaDB, pgvector), hybrid search, and generation with citations. - Agent Architectures: Provides reference implementations for ReAct, function calling, plan-and-execute, and multi-agent collaboration patterns. - LLMOps & Production Hardening: Includes prompt versioning and A/B testing, metrics tracking, distributed tracing, evaluation frameworks, caching, rate limiting, retry logic, and model fallback strategies. - Use Case: When designing a customer-support chatbot that must answer from company documentation, use this Skill to select a chunking strategy, wire hybrid retrieval, and add fallback models for quota failures. ## Quick Start Ask the AI to design a RAG pipeline with hybrid search and a fallback model strategy for your documentation chatbot.

Frequently Asked Questions about llm-app-patterns

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a RAG pipeline for question answering?▼

A RAG pipeline ingests documents through chunking and embedding, retrieves relevant context via vector or hybrid search, then generates an answer grounded in that context. Use a chunk size around 512 tokens with overlap, and return sources alongside the answer for citation.

What vector database should I use for semantic search?▼

Pinecone suits managed production workloads at billion-vector scale, pgvector fits existing Postgres infrastructure, Weaviate works for self-hosted multi-modal needs, and ChromaDB is appropriate for development and prototyping with thousands of vectors.

ReAct vs function calling: which agent pattern should I choose?▼

Function calling is lower complexity and cost when tools have structured schemas, while ReAct suits multi-step reasoning where the model must think through intermediate observations. Plan-and-execute fits complex tasks, and multi-agent fits research workflows.

How do I handle LLM rate limits and API failures in production?▼

Combine a request rate limiter with retry using exponential backoff for rate-limit and server errors, and configure a fallback chain of models so a failure on the primary model automatically tries alternates before raising an error.

When should I cache LLM responses?▼

Cache responses using a deterministic key built from the model, prompt, and parameters, but only store outputs generated with temperature zero since non-deterministic outputs should not be reused. Redis with a TTL is a typical backing store.

Related Skills

Explore More Software Engineering Skills→
neuron-coreneuron-core
2.0k

neuron-rag-specialist

Integrate vector stores, embeddings, and document loaders into Neuron AI agents.

Official
Advanced
Clay-HHKClay-HHK
1

langsmith-observability

Trace, evaluate, and monitor LLM applications using the LangSmith Python SDK.

Community
Intermediate
patricio0312revpatricio0312rev
5

guardrails-safety-filter-builder

Define a modular safety pipeline that detects unsafe prompts, redacts PII, and blocks policy-violating requests.

Community
Advanced
Scheune10-codingScheune10-coding

llm-framework-advisor

Recommend LLM agent frameworks based on use case and decision criteria.

Community
Advanced