RAG Fundamentals

Retrieval-augmented generation (RAG) lets your LLM answer questions grounded in your own documents rather than relying solely on its training data. These guides walk you from a minimal working pipeline all the way to hybrid search, conversation memory, and metadata filtering.

Every guide is self-contained, includes a Google Colab notebook, and ends with a complete runnable example.

Guides in this section

Guide	Time	Difficulty	What you'll build
RAG in 3 Lines	~5 min	Beginner	A working RAG pipeline with the absolute minimum code
Build a PDF Knowledge Base	~15 min	Beginner	Ingest a PDF, chunk it, store embeddings, and query
Multi-Format Document Ingestion	~20 min	Intermediate	Unified ingestion from PDF, DOCX, web, CSV, and directories
Choosing a Chunking Strategy	~15 min	Intermediate	Compare four splitters and pick the right one for your data
Streaming RAG Responses	~10 min	Beginner	Token-by-token streaming output and a FastAPI SSE endpoint
RAG with Conversation Memory	~15 min	Beginner	Multi-turn Q&A that remembers previous questions in a session
Metadata Filtering in Vector Search	~10 min	Intermediate	Scope retrieval by source, date, or category at query time
Hybrid BM25 + Vector Search	~20 min	Intermediate	Combine keyword and semantic search with a tunable alpha parameter

How RAG works in SynapseKit

Documents → Loader → Splitter → Embeddings → VectorStore
                                                   ↓
User query → Embed query → Retrieve top-k chunks → LLM → Answer

SynapseKit's RAGPipeline handles all of this with a single object. You choose the loader, splitter, vector store, embeddings, and LLM independently so every component is swappable.

Quickstart

import asyncio
from synapsekit import RAGPipeline
from synapsekit.llms.openai import OpenAILLM
from synapsekit.embeddings.openai import OpenAIEmbeddings
from synapsekit.vectorstores.memory import InMemoryVectorStore

rag = RAGPipeline(
    llm=OpenAILLM(model="gpt-4o-mini"),
    embeddings=OpenAIEmbeddings(model="text-embedding-3-small"),
    vectorstore=InMemoryVectorStore(),
)

async def main():
    await rag.aadd(["SynapseKit is a Python library for building LLM applications."])
    answer = await rag.aquery("What is SynapseKit?")
    print(answer)

asyncio.run(main())

Prerequisites for all guides

pip install synapsekit
export OPENAI_API_KEY="sk-..."

Ready to start? Begin with RAG in 3 Lines if you are new to RAG, or jump directly to the guide that matches your use case.