Web Search Configuration

Reka Research adds real-time web results to your AI app — so answers stay fresh, relevant, and grounded in current information.

Web search is on by default. Use these parameters to control how and where your agent searches:

1web_search: {
2 // Optional: Max number of web searches per request
3 max_uses: 10,
4
5 // Optional: Only search from these domains
6 allowed_domains: ["example.com", "safe_domain.org"],
7
8 // Optional: Exclude results from these domains
9 blocked_domains: ["unsafe_domain.com"],
10}

Limit usage

Use max_uses to limit how many searches the agent can run per request. Once the limit’s reached, it reason based on with what it has.

1completion = client.chat.completions.create(
2 model="reka-flash-research",
3 messages=[
4 {
5 "role": "user",
6 "content": "Who won the UEFA Nations League 2025?",
7 },
8 ],
9 extra_body={
10 "research": {
11 "web_search": {
12 "max_uses": 3,
13 },
14 }
15 },
16)

Filter sources

Choose where the agent can — or can’t — search:

  • allowed_domains: Only search from this list
  • blocked_domains: Never search from this list
1completion = client.chat.completions.create(
2 model="reka-flash-research",
3 messages=[
4 {
5 "role": "user",
6 "content": "Who won the UEFA Nations League 2025?",
7 },
8 ],
9 extra_body={
10 "research": {
11 "web_search": {
12 "allowed_domains": ["espn.com", "uefa.com"],
13 },
14 }
15 },
16)

Domain filtering

Control where your agent is allowed to search — and where it isn’t.

  • Use plain domains: Skip http://, https://, or www. — just write example.com.
  • Top-level filtering: Listing nytimes.com applies to all its subdomains too.
  • One list only: You can’t use allowed_domains and blocked_domains at the same time — doing so will trigger a validation error.

Performance impact

  • Filtering adds minimal overhead — but may slightly increase response time.
  • Fewer domains mean fewer results. Be mindful: strict filters can reduce answer quality.