Haystack Agents

• Key points are in the visual diagram above.

Teams already running Haystack for RAG can add agentic behavior — tool use, routing, multi-hop retrieval — without migrating to a new framework.

Haystack Agents turn a library search desk into a research librarian who not only finds books but also calls external databases and synthesizes answers.

Scenario

Enterprise search agent: user query → agent decides between Confluence retriever, Jira tool, or Slack search → synthesizes unified answer.

Solution

In Agent Framework Landscape, apply Haystack Agents to this scenario: Enterprise search agent: user query → agent decides between Confluence retriever, Jira tool, or Slack search → synthesizes unified answer. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Haystack Agents (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Haystack Agents happens in the code.

1from haystack.components.agents import Agent  # import dependencies2from haystack.tools import Tool  # import dependencies3
4search_tool = Tool(name="search_docs", description="Search internal wiki",5    function=wiki_search, parameters={"query": {"type": "string"}})6agent = Agent(chat_generator=generator, tools=[search_tool])7result = agent.run(messages=[{"role": "user", "content": "PTO policy 2026"}])

• Treating Haystack Agents as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for haystack agents

Quick recap — the most important points from this module.