Domain Filtering: Control Sources in the Seltz API

Domain filtering is live: control which sources your agent reads

When you're building an AI agent that reasons over the news, the sources it reads decide whether you're reasoning over signal or noise. A claim from a primary outlet and the same claim from a press-release aggregator are not equivalent, even when the words match.

A customer building real-time AI workflows on top of Seltz wanted to restrict results to a defined set of publishers they trusted, and to keep everything else out. The query was right and the documents were good, but they needed the source of every result to be one they'd already vetted.

We shipped domain filtering.

How it works

Two new optional parameters on the search endpoint: include_domains and exclude_domains. Both take an array of domains. Pass either, both, or neither.

include_domains is an allowlist: results come back only from the domains you name. exclude_domains is a blocklist: every source is fair game except the ones you name.

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "include_domains": ["reuters.com", "apnews.com"]
  }'

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "include_domains": ["reuters.com", "apnews.com"]
  }'

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "include_domains": ["reuters.com", "apnews.com"]
  }'

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "exclude_domains": ["prnewswire.com"]
  }'

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "exclude_domains": ["prnewswire.com"]
  }'

bash

curl -X POST https://api.seltz.ai/v1/search \
  -H "x-api-key: $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic SpaceX compute deal",
    "max_results": 10,
    "exclude_domains": ["prnewswire.com"]
  }'

The response returns the matching documents, with url on every result regardless of whether you're filtering:

json
{
  "documents": [
    {
      "url": "https://www.reuters.com/technology/article",
      "content": "..."
    }
  ]
}

json
{
  "documents": [
    {
      "url": "https://www.reuters.com/technology/article",
      "content": "..."
    }
  ]
}

json
{
  "documents": [
    {
      "url": "https://www.reuters.com/technology/article",
      "content": "..."
    }
  ]
}

Even without filtering, you can use those URLs to make source decisions downstream: route by publisher, weight sources differently, or dedupe across outlets covering the same story.

Domains, not URLs

The filter operates on the publisher rather than the path: you choose which sources your agent reads, not which individual articles. You can allow or block a publication, but not a single section within one. The model stays simple: a domain is either in or out.

Both filters match at the domain level, including subdomains: naming reuters.com covers www.reuters.com and any subdomain under it, whether you're including or excluding, so you don't have to list every host a publisher serves from.

Try it in the playground

The Search Playground in your console now has Include Domains and Exclude Domains fields. Drop in a query, name the sources to include or exclude, and run it. Testing requires no code.

What's next

Domain filtering is the second filter to ship, after date filtering for the News API.

If there's a filter you want, tell us. Try domain filtering in the playground