Cookbook
Cookbook
Fact-check a claim
Find pages about a claim, then quote the sentence on each page that supports it.
1 min read · Updated 20 Nov 2026
On this page
Search for the claim, then ask Lineage to find the sentence on each page that supports it. You get a score and a quote per source, and the pages that could not be read are listed separately.
fact-check.py
import osimport requestsAPI = "https://fluxsearch.io/api/v1"HEADERS = {"Authorization": f"Bearer {os.environ['FLUX_API_KEY']}"}CLAIM = "CATL and HyperStrong agreed a 60 GWh sodium-ion battery supply deal."def post(endpoint: str, body: dict) -> dict: response = requests.post(f"{API}/{endpoint}", headers=HEADERS, json=body, timeout=180) response.raise_for_status() return response.json()search = post("search", {"query": CLAIM, "max_results": 4, "enrich": 0})urls = [result["url"] for result in search["results"]]checked = post("lineage", {"documents": [{"url": url} for url in urls], "queries": [CLAIM]})print(f"Claim: {CLAIM}\n")for entry in checked["data"]: url = urls[entry["document_index"]] best = max(entry["results"], key=lambda p: p["score"] or 0, default=None) if best is None: print(f"no support {url}") continue print(f"{best['score']:.2f} {url}") print(f" \"{best['text'].strip()}\"")for error in checked.get("errors", []): print(f"unreadable {error['url']} ({error['code']})")const API = "https://fluxsearch.io/api/v1";const CLAIM = "CATL and HyperStrong agreed a 60 GWh sodium-ion battery supply deal.";type Passage = { text: string; score: number | null };type Entry = { document_index: number; results: Passage[] };type Failure = { url: string; code: string };async function post<T>(endpoint: string, body: unknown): Promise<T> { const response = await fetch(`${API}/${endpoint}`, { method: "POST", headers: { Authorization: `Bearer ${process.env.FLUX_API_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify(body), }); if (!response.ok) throw new Error(`Flux ${response.status}: ${await response.text()}`); return (await response.json()) as T;}const search = await post<{ results: { url: string }[] }>("search", { query: CLAIM, max_results: 4, enrich: 0,});const urls = search.results.map((r) => r.url);const checked = await post<{ data: Entry[]; errors?: Failure[] }>("lineage", { documents: urls.map((url) => ({ url })), queries: [CLAIM],});console.log(`Claim: ${CLAIM}\n`);for (const entry of checked.data) { const url = urls[entry.document_index]; const best = [...entry.results].sort((a, b) => (b.score ?? 0) - (a.score ?? 0))[0]; if (!best) { console.log(`no support ${url}`); continue; } console.log(`${(best.score ?? 0).toFixed(2)} ${url}`); console.log(` "${best.text.trim()}"`);}for (const error of checked.errors ?? []) console.log(`unreadable ${error.url} (${error.code})`);What it prints
Output
Claim: CATL and HyperStrong agreed a 60 GWh sodium-ion battery supply deal.0.97 https://www.catl.com/en/news/6812.html "The two parties announced a three-year partnership covering 60GWh of sodium-ion battery supply, marking a significant milestone for the industrialization of sodium-ion batteries technology."0.98 https://www.prnewswire.com/news-releases/catl-and-hyperstrong-sign-the-worlds-largest-sodium-ion-energy-storage-cooperation-agreement-302755769.html "The two parties announced a three-year partnership covering 60GWh of sodium-ion battery supply, marking a significant milestone for the industrialization of sodium-ion batteries technology."no support https://www.reddit.com/r/electricvehicles/comments/1sx6okt/catl_secures_worlds_largest_sodiumion_battery/0.96 https://www.hyperstrong.com/en/news/company-news/95 "April 27, Beijing - Beijing HyperStrong Technology Co., Ltd. ("HyperStrong") and Contemporary Amperex Technology Co., Limited ("CATL") signed a strategic cooperation agreement on sodium-ion batteries for energy storage, establishing a three-year partnership totaling 60 GWh, marking a pivotal milestone toward the large-scale deployment of sodium-ion batteries in the energy storage industry."How it works
enrich: 0makes the search fast, because Lineage reads the pages next.- One Lineage request reads every page through
documents. Each entry indatacarries thedocument_indexof the page it came from. - A page that fails is listed in
errorsand is not charged. The others are still answered.