Haystack industry newsletter

Discover, verify, and deduplicate a weekly news briefing.

What you’ll build

A Haystack research editor that turns a week of live industry coverage into a compact newsletter. It discovers stories, verifies important background, and avoids presenting syndicated copies as independent developments.

Default brief: Create a warehouse-robotics newsletter with one lead story, three briefs, why each matters, and source URLs.

How the agent works

  1. Discover the week. News search gathers a broad set of recent warehouse-robotics developments.
  2. Verify the background. Web search checks durable facts behind the strongest or most material claims.
  3. Edit and deduplicate. The Haystack agent collapses syndicated coverage and shapes the final issue within eight steps.

Core agent setup

This excerpt keeps the Haystack generator, tools, and editorial loop together. The complete script also validates credentials, normalizes the final message, and saves the newsletter.

agent = Agent(
    chat_generator=generator,
    system_prompt=(
        "Use live search, deduplicate stories, verify important claims, "
        "and include direct source URLs."
    ),
    tools=[
        news_search(provider="haystack"),
        web_search(provider="haystack", allowed_engines=["google_light", "bing"]),
    ],
    max_agent_steps=8,
)
result = agent.run(messages=[ChatMessage.from_user(PROMPT)])

View the complete main.py →

Run the recipe

Set SERPAPI_API_KEY and OPENAI_API_KEY in the repository-root .env, then run:

uv run --isolated --no-project --with 'serpapi-search-tools[haystack]' --with python-dotenv cookbook/haystack/main.py

Optional controls: OPENAI_MODEL, COOKBOOK_PROMPT, and COOKBOOK_OUTPUT_DIR.

Change the beat, not the wiring: Set COOKBOOK_PROMPT to cover another industry while keeping the discovery, verification, and deduplication loop intact.

Inspect the result

The recipe writes cookbook-output/haystack-industry-newsletter.md. Expect a lead story, three concise briefs, a “why it matters” angle for each item, and direct source URLs without duplicated syndicated stories.

The rendered newsletter is also printed to the terminal for a quick review.

SerpApi tools used

news_search discovers timely stories for the issue, and web_search verifies the durable background behind important claims. This two-pass use helps the editor reason about freshness and recognize syndicated copies of the same story.

Framework pattern: The editorial workflow follows the Haystack Cookbook newsletter agent and uses SerpApi news and web tools.