skills.rest
Tgoldi
TgoldiCOMMUNITY
Explore all skills by this author →
GitHub

llm-app-patterns

Implements RAG pipelines, agent architectures, prompt management, and LLMOps monitoring patterns.

1|Updated May 10, 2026
Search Tags:#prompt engineering#llm#rag#vector database#agents#observability#llmops
One-click install
npx skills add https://github.com/Tgoldi/claude-skills --skill llm-app-patterns-tgoldi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-app-patterns
Source: https://github.com/Tgoldi/claude-skills/tree/main/llm-app-patterns
Command: npx skills add https://github.com/Tgoldi/claude-skills --skill llm-app-patterns-tgoldi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building production LLM applications requires solving recurring design problems like retrieval grounding, agent orchestration, prompt versioning, and observability, and this Skill provides proven implementation patterns for each. ## Core Features & Use Cases - RAG Pipelines: Covers chunking strategies, embedding model selection, vector database options, hybrid retrieval, and generation with citations. - Agent Architectures: Provides ReAct, function calling, plan-and-execute, and multi-agent collaboration patterns with Python implementations. - Prompt IDE & LLMOps: Includes prompt templating, versioning, A/B testing, prompt chaining, metrics tracking, logging, evaluation, caching, rate limiting, and fallback strategies. - Use Case: When designing a document Q&A system, use the RAG section to pick a chunking strategy, select a vector database like pgvector or Pinecone, and implement hybrid search with reciprocal rank fusion. ## Quick Start Ask the assistant to design a RAG pipeline with hybrid search and citations for a document question-answering application.

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 with hybrid search?▼

Combine semantic vector search with BM25 keyword search and merge results using reciprocal rank fusion. Set the alpha parameter to balance the two: 1.0 for pure semantic, 0.0 for pure keyword, and 0.5 for a balanced mix.

What vector database should I use for RAG?▼

Pinecone suits managed production at billion-vector scale, Weaviate fits self-hosted multi-modal needs, ChromaDB works for prototyping, and pgvector integrates with existing Postgres infrastructure. Choose based on scale and operational constraints.

ReAct agent vs function calling: which should I use?▼

Function calling is simpler and lower cost for structured tool use with defined schemas. ReAct suits multi-step reasoning tasks where the model must think through intermediate steps before acting.

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

Use exponential backoff retries with tenacity for transient errors, a sliding-window rate limiter to stay under request quotas, and a fallback chain that switches to secondary models when the primary model fails.

What metrics should I track for LLM observability?▼

Track latency percentiles, tokens per request, cost per request, cache hit rate, error and timeout rates, plus quality signals like user satisfaction and hallucination rate. Log requests and responses with OpenTelemetry tracing.

Related Skills

Explore More Software Engineering Skills→
agenticgogolagenticgogol
1

technical-design

Generates design.md with architecture, API contracts, and tech choices from a clarified brief.

Community
Advanced
ContextJet-aiContextJet-ai
29

set-up-ab-testing

Split live traffic between control and experimental groups to compare LLM prompt and model outcomes.

Official
Advanced
DigiStrique-SolutionsDigiStrique-Solutions
1

eval-harness

Automate AI prompt and output evaluation against ground-truth fixtures.

Official
Advanced
bugrabilgebugrabilge
10

langfuse

Trace, evaluate, and manage prompts for LLM applications on Langfuse.

Community
Advanced