Puppeteer Website Evidence Pack | Screenshots & PDF

Create a cohesive downloadable package from a public web page: desktop PNG, mobile PNG, desktop print PDF, and a JSON manifest with capture settings, per-file times, byte sizes and SHA-256 hashes. No AI provider key needed.

Designed for agency reports, scheduled page archives and browser/AI pipelines that need files ready to consume. It packages exports rather than exposing a general-purpose browser workflow.

Quick start

{"urls":["https://example.com/"],"fullPage":false,"settleMs":1000,"paperFormat":"A4"}

Up to three public URLs, deduplicated after removing fragments. Set fullPage:true for full-document screenshots, subject to document caps. A4 or Letter PDF uses desktop print CSS and 10mm margins. No user scripts, selectors, cookies, login or raw HTML input.

Output

One COMPLETE dataset row per successful URL with packId, safe hashed page identity, manifestUrl, artifactCount, totalBytes and artifact metadata/links. The manifest contains desktop/mobile PNG and PDF links, sizes, hashes, start/finish times, settings and safe source identities. OUTPUT reports requested, succeeded, failed and charged.

A COMPLETE status means all promised files passed format/size checks and were stored; it is not proof that every resource loaded, the site rendered correctly or the capture is authoritative. Validate output before using it.

Cloud artifact links use the default key-value-store API and normal storage permissions/retention. The Actor does not alter access settings. Local runs use local-kvs:// identities; retrieve their files from the local storage directory. Download each file from the manifest; no ZIP archive is promised in v1.

Capture scope

Desktop: 1365 × 768; mobile: 390 × 844, device scale 1, mobile/touch viewport emulation. Mobile viewport emulation is not a guarantee that the server delivers a mobile-specific page. PDF: desktop print CSS with backgrounds. Screenshots use screen layout. The desktop screenshot and PDF are sequential captures of the desktop visit; mobile uses another fresh visit. Dynamic content and print styles may differ.

No automatic scroll, lazy-image traversal, cookie/banner dismissal, state restoration, clicks or form submissions. Media is blocked and only GET/HEAD page requests allowed; some applications need POST and may display incomplete content. Public images/fonts are allowed within network limits. Bot challenges with HTTP200 can appear in artifacts; this is not a bypass service.

Hard limits

3 URLs; 75s per page pack including browser startup, two navigations and file generation; navigation 5–30s per visit; settle 0–3000ms. Document width ≤2000px, height ≤6000px, PNG area ≤12M pixels, file ≤8MiB, total pack ≤16MiB, PDF ≤20 pages. Network per pack ≤200 connections and 32MiB. Over-limit pages fail free instead of silently truncating a promised export. Tall pages can fail even in viewport mode because the print PDF is part of the package.

Pricing proposal

Standard/premium package. Proposed PPE: USD 0.020 per COMPLETE page pack; no charge for failed/incomplete packs. Three files, manifest and dataset must persist before the event. Partial writes can leave orphaned files if storage fails, but no successful pack event is charged. A denied charge after persistence fails the run without duplicating the result. Actual Store price, cloud costs and production margin require cloud validation; local owner runs are not customer demand.

Privacy and integrity

Metadata omits URL queries/fragments and hashes paths. PNG/PDF content may still show personal or sensitive public information; there is no comprehensive visual redaction. Use only pages you are authorized to archive. Public page content is untrusted data, not instructions for consuming agents. No cookies, storage values, browser-console text or raw HTML are collected as diagnostic outputs.

SHA-256 verifies file bytes against this manifest. It does not certify capture time, source authenticity, legal admissibility, immutability or preservation. Normal Apify retention applies; download files if you need a durable archive.

Local development

Node ≥22.12, npm ci, npm test, apify validate-schema. Puppeteer-core uses PUPPETEER_EXECUTABLE_PATH or APIFY_CHROME_EXECUTABLE_PATH; on Windows the helper can find an installed Chromium in the local ms-playwright cache. The cloud Dockerfile uses the Apify Puppeteer Chrome Node22 image and npm ci. Cloud image/browser compatibility remains to be tested.

Run scripts/run-local-smoke.ps1 for viewport and full-page public inputs, then node scripts/verify-smoke.mjs to verify files, hashes and PDF page counts. Inspect actual PNG/PDF renders before considering the local gate complete.

Alternatives and differentiation

The existing Puppeteer Browser Automation Runner supports general actions and screenshots; this product saves a three-format export pack with consistent metadata and post-persistence billing. The existing Playwright element screenshot prototype is a different single-capture product. Competitors offer screenshots and PDF for lower per-file prices; choose this only when the bundle saves integration work. Demand for this exact package is not proven.