Search, read, answer

/** * Search, read, answer: search the web, read the top results in a sandboxed browser, and answer with sources. * *   npx tsx node/index.ts            (BOXLINE_API_KEY; QUESTION for your own question) * * Writes output/result.json: the search results, the answer and the pages it came from. */import { mkdirSync, writeFileSync } from "node:fs";import { join } from "node:path";import { Boxline } from "@boxline/sdk";const out = process.env.OUTPUT_DIR ?? "output";const question = process.env.QUESTION ?? "In what year was the Python programming language first released?";const bx = new Boxline();// 1. Search, and open the top 3 results as Markdown (in a real browser inside a sandbox, never on this computer).//    The query goes to the search provider: keep passwords and personal data out of it.const found = await bx.search({ query: question, limit: 5, fetch: 3 });console.log(`${found.results.length} results${found.cached ? " (from the cache)" : ""}:`);for (const r of found.results) console.log(`  ${r.content ? "read" : "    "}  ${r.title} (${r.url})${r.error ? ` (${r.error.code})` : ""}`);// 2. Answer from the pages that loaded, with a schema so the answer and its sources come back as data.const read = found.results.filter((r) => r.content).map((r) => r.url);if (!read.length) throw new Error("none of the top results could be opened; try another question");const answer = await bx.extract<{ answer: string; sources: string[] }>({  urls: read,  prompt: `Answer this question in one or two sentences, using only these pages: "${question}". In "sources", list the URLs of the pages the answer comes from.`,  schema: {    type: "object",    properties: { answer: { type: "string" }, sources: { type: "array", items: { type: "string" } } },    required: ["answer", "sources"],  },});console.log(`\n${answer.data.answer}\nSources:\n${answer.data.sources.map((s) => `  - ${s}`).join("\n")}`);mkdirSync(out, { recursive: true });writeFileSync(  join(out, "result.json"),  JSON.stringify(    {      question,      results: found.results.map((r) => ({ title: r.title, url: r.url, read: Boolean(r.content) })),      answer: answer.data.answer,      sources: answer.data.sources,      model: answer.model,      usage: { modelUsd: answer.usage.costUsd, searches: found.cached ? 0 : 1 },    },    null,    2,  ),);

"""Search, read, answer: search the web, read the top results in a sandboxed browser, and answer with sources.    python python/main.py            (BOXLINE_API_KEY; QUESTION for your own question)Writes output/result.json: the search results, the answer and the pages it came from."""import jsonimport osfrom pathlib import Pathfrom boxline import Boxlineout = Path(os.environ.get("OUTPUT_DIR", "output"))question = os.environ.get("QUESTION", "In what year was the Python programming language first released?")bx = Boxline()# 1. Search, and open the top 3 results as Markdown (in a real browser inside a sandbox, never on this computer).#    The query goes to the search provider: keep passwords and personal data out of it.found = bx.search(question, limit=5, fetch=3)print(f"{len(found['results'])} results{' (from the cache)' if found['cached'] else ''}:")for r in found["results"]:    error = f" ({r['error']['code']})" if r.get("error") else ""    print(f"  {'read' if r.get('content') else '    '}  {r['title']} ({r['url']}){error}")# 2. Answer from the pages that loaded, with a schema so the answer and its sources come back as data.read = [r["url"] for r in found["results"] if r.get("content")]if not read:    raise SystemExit("none of the top results could be opened; try another question")answer = bx.extract(    urls=read,    prompt=f'Answer this question in one or two sentences, using only these pages: "{question}". '    'In "sources", list the URLs of the pages the answer comes from.',    schema={        "type": "object",        "properties": {"answer": {"type": "string"}, "sources": {"type": "array", "items": {"type": "string"}}},        "required": ["answer", "sources"],    },)data = answer["data"]print(f"\n{data['answer']}\nSources:")for s in data["sources"]:    print(f"  - {s}")out.mkdir(parents=True, exist_ok=True)result = {    "question": question,    "results": [{"title": r["title"], "url": r["url"], "read": bool(r.get("content"))} for r in found["results"]],    "answer": data["answer"],    "sources": data["sources"],    "model": answer["model"],    "usage": {"modelUsd": answer["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