Haystack is an end-to-end framework that you can use to build powerful and production-ready pipelines with Large Language Models (LLMs) for different search use cases.

perform retrieval-augmented generation (RAG), question answering, or semantic document search, you can use the state-of-the-art LLMs and NLP models in Haystack to provide custom search experiences and make it possible for your users to query in natural language.

There are a lot of components you can use to build Haystack search systems:

• Node
  • it performs different kinds of tasks such as: Preprocessing data, Retrieving documents, Answering questions, Summarizing documents and Routing data.
  • These are often powered by the latest LLMs and transformer models.
• Pipeline
  • They connect these individual nodes in a specific order to accomplish a particular NLP task.
  • it has different types:
    • Querying Pipelines:
      • it is used to receive a query from the user and produce a result. They have access to a DocumentStore which stores a set of documents.
    • Indexing Pipelines
      • it is used to prepare your files for search. Their main objective is to convert your files into Haystack Documents, so that they can be saved in a DocumentStore.
• Agent
  • Agents are essentially LLMs equipped with a set of tools (which can be pipelines or individual Haystack nodes).
  • They are used to tackle complex tasks far beyond simple pipelines.
  • Unlike pipelines, which follow a fixed sequence of steps, agents work iteratively through a decision-making process.
  • They use the LLM's ability to generate text and understand prompts to:
    • Break down complex questions into smaller, actionable steps.
    • Choose the right tool for each step based on their understanding of the information and previous results.
    • Generate appropriate inputs for the chosen tool.
    • Interpret the output from the tool to decide if they have the answer or need to repeat the process.
  • Usage:
    • Handle multi-hop question answering:
      They can combine information from multiple sources and steps to answer complex questions that require reasoning and analysis.
    • Flexibility and adaptability:
      You can create custom tools and configure agents to fit specific needs and domains.

Summary:

• Retrievers act like librarians, scanning through a vast collection of books to find relevant ones based on your request.
• Readers serve as expert analysts, carefully examining those books to pinpoint the specific information you need.