Recipe: Enterprise RAG Chatbot

Use Case: This configuration is ideal for projects building production-grade Retrieval-Augmented Generation (RAG) systems — such as internal knowledge bases, document Q&A, and enterprise chatbots. It directly addresses pain points like hallucination, missing citations, and inconsistent prompt versioning.

Configuration (.ai-rules.yaml)

Copy the following into your project's root .ai-rules.yaml:

profile: llm-eng
snippets:
  - rag
  - prompt-engineering
  - responsible-ai

Why this stack?

llm-eng (profile): LLM Engineering overlay enforces strict standards for non-deterministic LLM output — including prompt versioning, fallback handling, and structured output contracts — which are essential for reliable RAG pipelines.

rag: Requires the AI agent to implement hybrid search (dense + sparse), preserve document metadata, enforce chunk overlap strategies, and always include source citations in responses.

prompt-engineering: Mandates version-controlled prompt templates with explicit variable declarations. Prevents prompt drift across development iterations and ensures reproducible retrieval behavior.

responsible-ai: Adds guardrails for hallucination detection, bias auditing, and PII handling — critical when the RAG system surfaces sensitive enterprise documents to end users.