rag-builder

Overview

Retrieval-Augmented Generation (RAG) is the standard pattern for grounding LLMs in your data. This skill covers the full pipeline: document loading, chunking strategies, embedding, vector storage, retrieval, and answer synthesis.

Capabilities

• Design document chunking strategies (fixed, semantic, recursive)
• Select and configure embedding models for your use case
• Implement hybrid search (vector + keyword) for better retrieval
• Build answer synthesis with source attribution
• Evaluate RAG quality with RAGAS metrics

When to Use

Trigger phrases:

• "rag builder"

• "RAG pipeline design — document chunking, embedding strategies, retrieval optimiz"

• Building a chatbot over your documentation or knowledge base

• Need LLM answers grounded in factual, up-to-date data

• Document Q&A where hallucination is unacceptable

• Customer support automation over product docs

When NOT to Use

• Task is outside your authorization scope
• You need to implement controls (use implementing-* skills)
• Task is about analysis, not action (use analyzing-* skills)
• You don't have access to target systems
• Task requires compliance expertise (consult professionals)
• Task is about defense, not offense (use defensive skills)

Pseudo Code

# Example workflow for this skill
def execute(input_data):
    # Step 1: Validate input
    if not input_data:
        raise ValueError("Input data is required")

    # Step 2: Process core logic
    result = process(input_data)

    # Step 3: Validate output
    validate_output(result)

    return result

Chunking Strategy

def chunk_document(doc, strategy="recursive", chunk_size=512, overlap=50):
    if strategy == "fixed":
        return split_by_chars(doc, chunk_size, overlap)
    elif strategy == "recursive":
        return recursive_split(doc, separators=["\n\n", "\n", ". ", " "], chunk_size=chunk_size)
    elif strategy == "semantic":
        return semantic_split(doc, similarity_threshold=0.8)