RAG Systems

Build Retrieval-Augmented Generation systems for grounded responses.

When to Use This Skill

Invoke this skill when:

• Building Q&A over custom documents
• Implementing semantic search
• Setting up vector databases
• Optimizing retrieval quality

Parameter Schema

Parameter	Type	Required	Description	Default
task	string	Yes	RAG goal	-
vector_db	enum	No	pinecone, weaviate, chroma, pgvector	chroma
embedding_model	string	No	Embedding model	text-embedding-3-small
chunk_size	int	No	Chunk size in chars	1000

Quick Start

from langchain_openai import OpenAIEmbeddings
from langchain_chroma import Chroma
from langchain_text_splitters import RecursiveCharacterTextSplitter

# 1. Split documents
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
chunks = splitter.split_documents(documents)

# 2. Create vector store
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
vectorstore = Chroma.from_documents(chunks, embeddings)

# 3. Retrieve
docs = vectorstore.similarity_search("query", k=5)

Chunking Strategy

Content Type	Size	Overlap	Rationale
Technical docs	500-800	100	Preserve code
Legal docs	1000-1500	200	Keep clauses
Q&A/FAQ	200-400	50	Atomic answers

Embedding Costs

Model	Cost/1M tokens
text-embedding-3-small	$0.02
text-embedding-3-large	$0.13
Cohere embed-v3	$0.10

Troubleshooting

Issue	Solution
Irrelevant results	Improve chunking, add reranking
Missing context	Increase k, use parent retriever
Hallucinations	Add "only use context" prompt
Slow retrieval	Add caching, reduce k

Best Practices

• Always include source attribution
• Use hybrid search (dense + BM25)
• Implement reranking for quality
• Evaluate with RAGAS metrics

Related Skills

• llm-integration - LLM for generation
• agent-memory - Memory retrieval
• ai-agent-basics - Agentic RAG

References

• LangChain RAG
• Pinecone Guide