Search

Retrieve relevance-ranked evidence with a live or historical cutoff.

Request parameters

Send a JSON body with query. Unknown parameters are rejected with invalid_parameter.

Pricing

Standard search costs 3 credits for up to 25 requested results, or 5 credits for 26–50. Reranking does not change the price. Advanced mode costs 5 credits per request with up to 20 results.

Time semantics

as_of_date and recency filter by indexing time. For a historical day, the upper bound is the end of that UTC day. Publication dates describe the original source and can be missing or differ from indexing time.

Put dates in the date fields. Writing “as of December 11” in the query does not set a cutoff. Boolean operators, quotes, and minus signs are treated as ordinary query text.

Make a request

curl -s https://api.ariselabs-search-api.com/v1/search \
  -H "Authorization: Bearer $ARISELABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "query": "TSMC CoWoS capacity 2027",
  "k": 5,
  "with_rerank": true,
  "as_of_date": "2026-08-20",
  "recency": "30d"
}'

import httpx, os

r = httpx.post(
    "https://api.ariselabs-search-api.com/v1/search",
    headers={"Authorization": f"Bearer {os.environ['ARISELABS_API_KEY']}"},
    json={
        "query": "TSMC CoWoS capacity 2027",
        "k": 5,
        "with_rerank": True,
        "as_of_date": "2026-08-20",
        "recency": "30d",
    },
    timeout=60,
)
r.raise_for_status()
page = r.json()
for row in page["data"]:
    print(row.get("published_at", "")[:10], row["title"])

const r = await fetch("https://api.ariselabs-search-api.com/v1/search", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.ARISELABS_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
  "query": "TSMC CoWoS capacity 2027",
  "k": 5,
  "with_rerank": true,
  "as_of_date": "2026-08-20",
  "recency": "30d"
}),
});
if (!r.ok) throw new Error(await r.text());
const page = await r.json();

Read the response

The data array is ranked by relevance. Standard rows contain the indexed text (up to 20,000 characters), source URL, title, and publication time when available. Advanced web results can carry snippet instead of text.

{
  "object": "search_result_page",
  "livemode": true,
  "data": [
    {
      "url": "https://example.com/blackwell-supply",
      "title": "Blackwell supply outlook",
      "text": "Example article text…",
      "published_at": "2026-09-06T13:22:11Z"
    }
  ],
  "search_metadata": {
    "query": "nvidia blackwell supply",
    "returned_count": 1,
    "degraded": [],
    "rerank_applied": false,
    "elapsed_ms": 284
  },
  "usage": {
    "credits": 3
  }
}

• returned_count gives the number of rows returned. collapsed_duplicates reports skipped copies.
• rerank_applied reports whether reranking succeeded. There is no score field; the ordering expresses relevance.
• degraded: [] means no reported degradation. partial_coverage means part of the index did not answer; rerank_capacity means standard search returned the first-stage ranking.
• Advanced mode can also report unavailable web retrieval or pool reranking in degraded. Check it before relying on full coverage.
• usage.credits is the charge for this request. livemode describes the API key environment; it does not indicate whether the query used a historical cutoff.

To retrieve the current contents of a specific URL, use Fetch.

Handle failures

• 400: fix the parameter named in the message.
• 402: add credits in the console.
• 429: wait for Retry-After, then retry with backoff if needed.
• 503: retry with backoff and honor Retry-After.

See the full error reference →