How Is the Haystack Documentation Organized?

Updated

The Haystack documentation at docs.haystack.deepset.ai is organized around a task-based structure: new users start with installation and the "Get Started" path, while experienced users jump directly to component references, pipeline APIs, or the API reference section. The fastest route depends on whether you are learning the framework or looking up a specific class or parameter.

Main Sections You Will Encounter

Haystack's docs follow a common pattern for Python frameworks, with these functional areas:

Section What it answers Who it is for
Installation / Get Started How to install Haystack and run a first pipeline New users
Concepts / Overview What pipelines, components, and document stores are Users building a mental model
Tutorials / Guides Step-by-step tasks (RAG, indexing, retrieval) Learners who prefer doing
Component Reference What each component does and its parameters Users assembling pipelines
API Reference Class and method signatures Users checking exact inputs/outputs

The exact labels and ordering can change between releases, so treat this as a map of types of content rather than a fixed table of contents.

If You Are New to Haystack

Start with the installation instructions, then follow the introductory tutorial before reading component pages. The reason: Haystack's core abstraction is the pipeline, and component reference pages assume you already know how components connect.

A workable first path:

  1. Install Haystack in a virtual environment.
  2. Run the minimal "first pipeline" example from the Get Started section.
  3. Read the concepts page on pipelines and components.
  4. Move to a tutorial that matches your use case (for example, retrieval-augmented generation or document indexing).

Verify success by getting the example pipeline to return output before you customize anything. Most early confusion comes from modifying a pipeline before the unmodified example runs.

If You Already Know the Framework

Go straight to the reference material:

  • Component reference — for the list of available components and their initialization parameters.
  • API reference — for exact method signatures, return types, and exceptions.
  • Pipeline API — for how to connect, run, and serialize pipelines.

When a tutorial and a reference page disagree, trust the reference page for the version you are running, since tutorials are sometimes written against an earlier release.

Finding the Right Page and Version

Two mechanisms matter most:

  • Search — use the documentation search for a class or component name rather than browsing, especially if you already know the identifier.
  • Version switching — Haystack has had breaking changes across major versions, so confirm the version selector matches the version you installed. A page for a different major version can describe APIs that no longer exist in your environment.

A quick check: compare the version shown in the docs against pip show haystack-ai (or the equivalent for your install) before copying any code.

Common Sticking Points

  • Mixing major versions. Code from a 1.x-era tutorial often will not run on 2.x, and vice versa. Check the version selector first.
  • Skipping concepts. Jumping to component pages without understanding pipelines leads to wiring mistakes that are hard to debug.
  • Assuming tutorials are current. Treat tutorials as guidance and reference pages as the source of truth for parameters.

If you tell me your goal — learning from scratch, upgrading a version, or looking up a specific component — I can point to the specific section to open first.