Bright Data compatible

The one-line change, in code that posts to /request

If your code sends the request itself, with curl, requests, fetch or axios, change the host and the token and leave the body alone. This is the SERP call from Bright Data's docs, on SpecStream:

curl -X POST https://api.specstream.ai/brightdata/request \
  -H "Authorization: Bearer $SPECSTREAM_KEY" -H "Content-Type: application/json" \
  -d '{"zone": "sdk_serp", "url": "https://www.google.com/search?q=pizza&hl=en&gl=us&brd_json=1", "format": "raw"}'

And the Web Unlocker call, which returns the page. Add "data_format": "markdown" for markdown.

curl -X POST https://api.specstream.ai/brightdata/request \
  -H "Authorization: Bearer $SPECSTREAM_KEY" -H "Content-Type: application/json" \
  -d '{"zone": "sdk_unlocker", "url": "https://en.wikipedia.org/wiki/Web_scraping", "format": "raw"}'

@brightdata/sdk: point its endpoint table at SpecStream

@brightdata/sdk 1.2.0 has no base address option: utils/constants.cjs holds https://api.brightdata.com and builds every endpoint from it. Replace that host in the endpoint table once, before you create the client. The rest of your code stays as it is. SpecStream's tests run Bright Data's own client this way.

const path = require("path");
const { bdclient } = require("@brightdata/sdk");

// The client takes no base address, so point its endpoint table at SpecStream first.
const constants = require(path.join(path.dirname(require.resolve("@brightdata/sdk")), "utils/constants.cjs"));
for (const [name, value] of Object.entries(constants.API_ENDPOINT)) {
  constants.API_ENDPOINT[name] = value.replace("https://api.brightdata.com", "https://api.specstream.ai/brightdata");
}

(async () => {
  const client = new bdclient({ apiKey: process.env.SPECSTREAM_API_KEY });
  const markdown = await client.scrapeUrl("https://en.wikipedia.org/wiki/Web_scraping", { dataFormat: "markdown" });
  const results = JSON.parse(await client.search.google("pizza restaurants", { country: "us" }));
  console.log(markdown.slice(0, 200), results.organic[0].link);
  await client.close();
})();

The client lists zones and creates a missing one before its first call. SpecStream answers both, so autoCreateZones can stay on. This works where the SDK is loaded with require.

What each field does here

url: a Google address with path /search, a Bing address with /search or /news/search, or a Yandex address with /search is a search. Any other address is a page read. zone is accepted and changes nothing, since one route serves both.

Search fields are read from the results address as Bright Data's client writes them: q, start, hl, gl, safe, tbs=qdr: and tbm=nws for Google, q, count, first, mkt and cc for Bing, text, numdoc, p and lang for Yandex. country in the body sets the country when the address names none.

Without brd_json=1, a search returns a plain HTML page of the results, each a link and its snippet, and with data_format markdown a markdown list. The parsed reply carries pagination.next_page_link when more results exist.

Refused with a 400: data_format screenshot, a method other than GET, and image or video searches. A page SpecStream cannot read answers 502, and a missing page 404. A busy call gets 429 with Retry-After and is not charged.

One price for a search and one for a page

A search of up to 10 results is $0.001. Each started block of 10 more results in the same call adds $0.0012, up to 100 results. A page is one Extract API page at $0.0008, the same for every site. There is no plan and no zone to set up, and each charged reply carries the balance left in X-Balance-Cents.

The most accurate web search API in the world

On SimpleQA Verified the SpecStream Search API scores 97.9 F-score on all 1,000 official questions, 978 of them correct. On WideSearch it scores 64.0 item F1 and 43.4 row F1 on all 200 official tasks. The Extract API scores 0.908 F1 on WCXB. Every score and its method is on the benchmarks page.

You decide what is kept, on each key

Create a key with zero_retention and a call made with it keeps no request text and no answer after the reply. spend_cap_cents is the most the key can spend, so a job that loops stops at the cap you set.

curl -X POST https://api.specstream.ai/v1/account/keys \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"name": "production", "scopes": ["account:read", "finds:spend"], "spend_cap_cents": 500000, "zero_retention": true}'

For the SDK's dataset calls, ask for the rows

The SDK's platform scrapers and dataset calls go to Bright Data's /datasets addresses, which this route does not answer. On SpecStream that job is Deep Search: a request in plain words returns every match on the live web as rows, at 5¢, 15¢ or $1 a row, with the most a run can cost quoted before it starts. Bright Data's rates for each product are on Bright Data pricing.

SerpApi

Search results as JSON on monthly plans: Starter is $25 a month for 1,000 searches, and cached, errored and failed searches are not counted. SpecStream answers 1,000 searches for $1, and SerpApi's own call too: see the SerpApi migration guide.

Firecrawl

Search and scrape billed in credits: 2 credits per 10 search results, and Hobby at $19 a month for 5,000 credits. On SpecStream a search is $0.001 and a page $0.0008 with no plan to buy, and Firecrawl's calls work on our address: see the Firecrawl migration guide.

ScraperAPI

A page API on monthly plans: Hobby is $49 a month for 100,000 API credits. SpecStream reads 100,000 pages for $80 with no monthly plan, and answers ScraperAPI's call: see the ScraperAPI migration guide.

Tavily

A search API billed in credits at $0.008 a credit on pay as you go. On SpecStream a search is $0.001, and Tavily's client runs on our address: see the Tavily migration guide.