Menú
Serve, scale, and safely roll out a RAG pipeline with health checks, caching, and canary releases.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
| name | production-rag-deployment |
| title | Production RAG Deployment |
| description | Serve, scale, and safely roll out a RAG pipeline with health checks, caching, and canary releases. |
| allowed-tools | ["Read","Grep","Glob","Bash"] |
| category | deployment |
| tags | ["serving","scaling","canary","observability"] |
This skill covers turning a working RAG pipeline into a reliable production service: exposing it behind an API, scaling retrieval and generation independently, caching hot paths, and rolling out changes without breaking users.
Prototype RAG code assumes one user, a warm index, and unlimited time. In production it faces concurrency, cold starts, provider timeouts, and cost limits. Without serving discipline, latency spikes, failures cascade, and a bad deploy hits every user at once.
Wrap retrieval and generation in an API with timeouts, retries, and health checks for each dependency.
Why: Explicit timeouts and health checks stop a slow vector store or LLM from stalling the whole request path.
Scale the retrieval and generation tiers independently and cache embeddings and frequent answers.
Why: Retrieval and generation bottleneck differently, and caching removes repeated cost and latency on common queries.
Release changes to a small traffic slice first and define a degraded path when a component is unavailable.
Why: Canary releases and fallbacks contain the blast radius of a bad deploy or a failing dependency.
For infrastructure patterns, see the Qdrant Cloud documentation and the Ray Serve LLM guide.
Use this skill when building, debugging, or improving Retrieval-Augmented Generation systems, including chunking, vector database selection, hybrid search, reranking, multimodal RAG, code documentation RAG, retrieval latency, and production RAG architecture.
Chunk nested documents into parent-child levels so retrieval can move from broad sections to fine-grained passages.
Use semantic boundaries and embedding similarity to chunk text for higher-relevance retrieval.
Route RAG chunking decisions across semantic, hierarchical, sliding-window, contextual-header, and framework-selection strategies.
Use overlapping windows to preserve context across chunk boundaries while controlling retrieval size.
Reduce retrieval latency with caching, batching, and index-level optimization.
