Haystack — Module 14.3

How This Lesson Fits the Module & Volume

Haystack (by deepset) is Volume 14’s orchestration capstone: production-minded pipelines and components for NLP and RAG—retrievers, readers, rankers, generators—composable as explicit graphs. It sits beside LangChain and LlamaIndex as a battle-tested search/RAG framework.

After this lecture you leave Vol. 14 with retrieval + stores + orchestration literacy, ready for Volume 15 AI Agents.

Learning Objectives

By the end of this lesson, students should be able to:

Describe Haystack pipelines as directed graphs of components.
Wire a minimal RAG pipeline (embedder, retriever, generator).
Contrast Haystack pipelines with LangChain LCEL and LangGraph agents.
Explain document stores and component I/O contracts.
Synthesize a Vol. 14 stack choice: store + orchestration + agent path.
Articulate how Vol. 14 skills feed Volume 15 agent design.

Definition

Haystack is an open-source framework for building production search and RAG systems. Applications are pipelines: components (embedders, retrievers, rankers, generators, routers) connected so data flows through an explicit, testable graph.

Framework Capstone Comparison

Framework	Best mental model	Vol 15 bridge
LangChain	Runnable app glue	Tools + chains → agents
LangGraph	Stateful agent graphs	Direct agent control plane
LlamaIndex	Indexes & query engines	Retrieval inside agents
CrewAI / AutoGen	Multi-agent teams/chats	Multi-agent systems
PydanticAI	Typed agent I/O	Reliable tool contracts
Haystack	Search/RAG pipelines	Retrieval subgraphs for agents

Minimal RAG Pipeline Sketch

from haystack import Pipeline
from haystack.components.embedders import (
    SentenceTransformersTextEmbedder,
    SentenceTransformersDocumentEmbedder,
)
from haystack.components.retrievers.in_memory import InMemoryEmbeddingRetriever
from haystack.components.generators import OpenAIGenerator
from haystack.components.builders import PromptBuilder
from haystack.document_stores.in_memory import InMemoryDocumentStore
from haystack.components.writers import DocumentWriter
from haystack import Document

document_store = InMemoryDocumentStore()
docs = [
    Document(content="SSO is configured under Admin > Security.", meta={"product": "auth"}),
    Document(content="Invoices live under Workspace > Billing.", meta={"product": "billing"}),
]

indexing = Pipeline()
indexing.add_component("doc_embedder", SentenceTransformersDocumentEmbedder())
indexing.add_component("writer", DocumentWriter(document_store))
indexing.connect("doc_embedder", "writer")
indexing.run({"doc_embedder": {"documents": docs}})

prompt = """Answer using the context.
Context:
{% for doc in documents %}{{ doc.content }}{% endfor %}
Question: {{ query }}
"""

rag = Pipeline()
rag.add_component("text_embedder", SentenceTransformersTextEmbedder())
rag.add_component("retriever", InMemoryEmbeddingRetriever(document_store))
rag.add_component("prompt", PromptBuilder(template=prompt))
rag.add_component("llm", OpenAIGenerator(model="gpt-4o-mini"))
rag.connect("text_embedder.embedding", "retriever.query_embedding")
rag.connect("retriever", "prompt.documents")
rag.connect("prompt", "llm")

out = rag.run({
    "text_embedder": {"text": "Where do I configure SSO?"},
    "prompt": {"query": "Where do I configure SSO?"},
})
print(out["llm"]["replies"][0])

Volume 14 → Volume 15 Bridge

14.1 RAG

Chunks & retrieval

14.2 Stores

Vectors at scale

14.3 Orch.

Pipelines & agents

15 Agents

Loops, tools, memory

Haystack pipelines are excellent retrieval subgraphs inside larger agents: an agent (LangGraph/PydanticAI/Crew) can call a Haystack RAG pipeline as a tool, then plan next actions with Volume 15 patterns—memory, reflection, multi-agent handoffs.

Strengths

Explicit, testable pipelines
Strong IR/RAG heritage
Clear component contracts
Production search mindset

Tradeoffs

Less “agent playground” than CrewAI
Pipeline design overhead
Need document store ops
Ecosystem overlaps others

Common Misconception

“One orchestration framework must do everything.” Mature stacks mix: Haystack or LlamaIndex for retrieval pipelines, LangGraph for agent control, PydanticAI for typed tool I/O, Qdrant/Milvus for vectors. Composition beats monoculture.

Knowledge Check

Short Answer: What is a Haystack pipeline? Answer: A graph of components connected to process data for search/RAG.
True/False: Components have defined inputs/outputs you connect. Answer: True.
Multiple Choice: Haystack is especially associated with: (a) only CSS, (b) production search/RAG pipelines, (c) CUDA drivers. Answer: (b).
Short Answer: Name one component type in a RAG pipeline. Answer: Embedder, retriever, ranker, prompt builder, generator (any).
True/False: Haystack replaces the need for Volume 15 agent concepts. Answer: False.
Multiple Choice: A good Vol 15 pattern is to treat RAG pipelines as: (a) illegal, (b) tools/subgraphs for agents, (c) only batch ETL. Answer: (b).
Short Answer: Why prefer explicit pipelines? Answer: Testability, clarity, production debugging.
Short Answer: Name two other 14.3 frameworks and their focus. Answer: e.g. LangGraph=state graphs; CrewAI=role crews; PydanticAI=typed agents (any two).
Multiple Choice: Mixing frameworks is: (a) always forbidden, (b) often pragmatic, (c) only for CNNs. Answer: (b).
True/False: Volume 14 ends by bridging into AI Agents (Vol. 15). Answer: True.

Key Takeaways

Haystack builds explicit RAG/search pipelines from composable components.
It caps Vol. 14 orchestration beside LangChain, LlamaIndex, and agent frameworks.
Retrieval pipelines become tools inside Volume 15 agents.
Choose stores (14.2) and orchestrators (14.3) by control, scale, and typing needs.
Next volume: AI Agent—loops, tools, memory, multi-agent systems.

Trainer’s Guide

Capstone: Teams propose a full stack—one vector DB from 14.2 + one primary orchestrator from 14.3 + how an agent in Vol 15 will call RAG as a tool. 5-minute architecture pitches.

Lab: Implement the indexing + RAG pipelines above on a 10-document corpus; swap InMemory for a Module 14.2 store if time allows.

Recap: Haystack pipelines close Volume 14’s orchestration story. Continue to Volume 15 — AI Agent.