Fullstack RAG Pro

Boundary

This skill covers the end-to-end RAG architecture: document ingestion, chunking, embedding generation, vector database querying, and LLM synthesis.

When to use

• Building an AI chat application with custom knowledge.
• Indexing documents (PDFs, markdown) into a Vector DB.
• Implementing semantic search or hybrid search (semantic + keyword).
• Optimizing RAG context windows and retrieval quality.

Workflow

1. Ingestion Strategy: Determine document parsers and chunking logic (e.g., recursive character text splitter).
2. Embedding Selection: Choose embedding model (OpenAI text-embedding-3-small, local BGE, etc.).
3. Database Choice: Use pgvector, Pinecone, or Qdrant for vector storage.
4. Retrieval: Implement cosine similarity search with optional metadata filtering.
5. Synthesis: Format retrieved context cleanly into the LLM system prompt.