Agent and RAG Grounding

Search with content

scrape_results: true scrapes each result page and returns its content in the formats you request. The rag and markdown formats are the usual choices for LLM context.

import requests

resp = requests.post(
    "https://api.alterlab.io/api/v1/search",
    headers={"X-API-Key": "YOUR_API_KEY"},
    json={
        "query": "postgres 17 release notes logical replication",
        "engine": "auto",
        "num_results": 5,
        "scrape_results": True,
        "formats": ["markdown"],
    },
).json()

context = "\n\n".join(
    f"[{i}] {r['title']} ({r['url']})\n{(r.get('content') or {}).get('markdown') or r['snippet']}"
    for i, r in enumerate(resp["results"], 1)
)

Check the Search API reference for the exact content shape of a scraped result.

Freshness and domain control

Unified search accepts freshness (day, week, month, year), include_domains, exclude_domains and max_per_domain so an agent can restrict grounding to recent pages from sources you trust. Use sources to mix verticals such as web and news in one request.

Keep citations

Every result carries url, title and published_at when known; the response has fetched_at. Carry these into your prompt and ask the model to cite by index, so each claim traces back to a page you can open. For multi-step investigations that need many searches and reads, use a research job instead of looping in your own agent.

Cost