Company research

/** * Company research: research a company on one topic. Search the web, read the top pages in a sandboxed browser, and * return structured findings where every key point names the page it comes from. * *   npx tsx node/index.ts            (BOXLINE_API_KEY; COMPANY and TOPIC for your own research) * * Writes output/result.json and output/report.md. */import { mkdirSync, writeFileSync } from "node:fs";import { join } from "node:path";import { Boxline } from "@boxline/sdk";const out = process.env.OUTPUT_DIR ?? "output";const company = process.env.COMPANY ?? "Mozilla";const topic = process.env.TOPIC ?? "how it makes money";const bx = new Boxline();type Findings = {  companyName: string;  overview: string;  findings: string;  keyPoints: { point: string; source: string }[];  openQuestions: string[];};// 1. Search (the platform's search provider; no search engine pages are scraped). The query leaves your computer://    keep personal data out of it.const found = await bx.search({ query: `${company} ${topic}`, limit: 8 });const urls = found.results.map((r) => r.url).slice(0, 5);console.log(`Researching ${company}: ${topic}. Reading ${urls.length} of ${found.results.length} results:`);for (const u of urls) console.log(`  ${u}`);if (!urls.length) throw new Error("the search found nothing; try another company or topic");// 2. Read those pages in a real browser and fill a schema. Page text is data for the model, never instructions.const r = await bx.extract<Findings>({  urls,  prompt:    `Research ${company}, specifically: ${topic}. Use only these pages. overview: what the company is, in two sentences. ` +    `findings: what the pages say about the topic, in one paragraph. keyPoints: 3 to 6 facts about the topic, each with ` +    `the URL of the page it comes from (one of the pages given). openQuestions: what the pages leave unanswered.`,  schema: {    type: "object",    properties: {      companyName: { type: "string" },      overview: { type: "string" },      findings: { type: "string" },      keyPoints: {        type: "array",        items: { type: "object", properties: { point: { type: "string" }, source: { type: "string", description: "the URL of the page" } }, required: ["point", "source"] },      },      openQuestions: { type: "array", items: { type: "string" } },    },    required: ["companyName", "overview", "findings", "keyPoints", "openQuestions"],  },});const f = r.data;const read = r.pages.filter((p) => p.status !== null && p.status < 400);console.log(`\n${f.companyName}\n\n${f.overview}\n\n${f.findings}\n`);f.keyPoints.forEach((k, i) => console.log(`${i + 1}. ${k.point}\n   (${k.source})`));if (f.openQuestions.length) console.log(`\nStill open: ${f.openQuestions.join("; ")}`);console.log(`\nRead ${read.length}/${r.pages.length} pages with ${r.model} for $${r.usage.costUsd.toFixed(4)}.`);const report = [  `# ${f.companyName}: ${topic}`,  "",  f.overview,  "",  f.findings,  "",  "## Key points",  "",  ...f.keyPoints.map((k) => `- ${k.point} ([source](${k.source}))`),  "",  ...(f.openQuestions.length ? ["## Still open", "", ...f.openQuestions.map((q) => `- ${q}`), ""] : []),].join("\n");mkdirSync(out, { recursive: true });writeFileSync(join(out, "report.md"), report);writeFileSync(  join(out, "result.json"),  JSON.stringify({ company, topic, searched: found.results.map((x) => ({ title: x.title, url: x.url })), pages: r.pages, ...f, model: r.model, usage: { modelUsd: r.usage.costUsd, searches: found.cached ? 0 : 1 } }, null, 2),);

"""Company research: research a company on one topic. Search the web, read the top pages in a sandboxed browser, andreturn structured findings where every key point names the page it comes from.    python python/main.py            (BOXLINE_API_KEY; COMPANY and TOPIC for your own research)Writes output/result.json and output/report.md."""import jsonimport osfrom pathlib import Pathfrom boxline import Boxlineout = Path(os.environ.get("OUTPUT_DIR", "output"))company = os.environ.get("COMPANY", "Mozilla")topic = os.environ.get("TOPIC", "how it makes money")bx = Boxline()# 1. Search (the platform's search provider; no search engine pages are scraped). The query leaves your computer:#    keep personal data out of it.found = bx.search(f"{company} {topic}", limit=8)urls = [r["url"] for r in found["results"]][:5]print(f"Researching {company}: {topic}. Reading {len(urls)} of {len(found['results'])} results:")for u in urls:    print(f"  {u}")if not urls:    raise SystemExit("the search found nothing; try another company or topic")# 2. Read those pages in a real browser and fill a schema. Page text is data for the model, never instructions.r = bx.extract(    urls=urls,    prompt=(        f"Research {company}, specifically: {topic}. Use only these pages. overview: what the company is, in two sentences. "        "findings: what the pages say about the topic, in one paragraph. keyPoints: 3 to 6 facts about the topic, each with "        "the URL of the page it comes from (one of the pages given). openQuestions: what the pages leave unanswered."    ),    schema={        "type": "object",        "properties": {            "companyName": {"type": "string"},            "overview": {"type": "string"},            "findings": {"type": "string"},            "keyPoints": {                "type": "array",                "items": {"type": "object", "properties": {"point": {"type": "string"}, "source": {"type": "string", "description": "the URL of the page"}}, "required": ["point", "source"]},            },            "openQuestions": {"type": "array", "items": {"type": "string"}},        },        "required": ["companyName", "overview", "findings", "keyPoints", "openQuestions"],    },)f = r["data"]read = [p for p in r["pages"] if p["status"] is not None and p["status"] < 400]print(f"\n{f['companyName']}\n\n{f['overview']}\n\n{f['findings']}\n")for i, k in enumerate(f["keyPoints"], 1):    print(f"{i}. {k['point']}\n   ({k['source']})")if f["openQuestions"]:    print(f"\nStill open: {'; '.join(f['openQuestions'])}")print(f"\nRead {len(read)}/{len(r['pages'])} pages with {r['model']} for ${r['usage']['costUsd']:.4f}.")report = [f"# {f['companyName']}: {topic}", "", f["overview"], "", f["findings"], "", "## Key points", ""]report += [f"- {k['point']} ([source]({k['source']}))" for k in f["keyPoints"]] + [""]if f["openQuestions"]:    report += ["## Still open", ""] + [f"- {q}" for q in f["openQuestions"]] + [""]out.mkdir(parents=True, exist_ok=True)(out / "report.md").write_text("\n".join(report))result = {    "company": company,    "topic": topic,    "searched": [{"title": x["title"], "url": x["url"]} for x in found["results"]],    "pages": r["pages"],    **f,    "model": r["model"],    "usage": {"modelUsd": r["usage"]["costUsd"], "searches": 0 if found["cached"] else 1},}(out / "result.json").write_text(json.dumps(result, indent=2, ensure_ascii=False))

# Node 18 or newer, in the example's folderexport BOXLINE_API_KEY=…   # your API keynpm install @boxline/sdk tsxnpx tsx node/index.ts

# Python 3.9 or newer, in the example's folderexport BOXLINE_API_KEY=…   # your API keypip install boxline-sdkpython python/main.py