CLI

@wholisphere.ai/cli is a self-contained command-line client. Use it locally or in CI to render VPATs from a scan, submit findings, and call live-assist capabilities.

Install

npm install -g @wholisphere.ai/cli
# or run without installing:
npx @wholisphere.ai/cli --help

The binary is wholisphere. Authenticate with an API key:

export WHOLISPHERE_API_KEY=whk_…

Commands

Command	What it does
wholisphere health	Check API reachability + auth
wholisphere scan <scanId>	Render a stored scan’s VPAT (--format json|openacr|openacr-html|openacr-markdown|html|markdown)
wholisphere vpat	Generate a VPAT from a findings JSON file (--in findings.json)
wholisphere findings	Submit or fetch findings for a product
wholisphere act-report <scanId>	Emit a W3C ACT/EARL report for a scan
wholisphere describe	Describe an image via the vision model
wholisphere simplify	Rewrite text at a reading level
wholisphere summarize	Summarize text into overview + bullets
wholisphere cache:invalidate	Invalidate an intent cache (e.g. after a deploy)

Run any command with --help for its flags.

Example: VPAT in CI

Render the latest scan’s VPAT as procurement-ready HTML and save it as a build artifact:

wholisphere scan "$SCAN_ID" --format html > vpat.html

For scanning a site in CI (rather than rendering an existing scan), use the GitHub Action or the @wholisphere.ai/scanner-runner package (wholisphere-scan), which fetches pages, runs the evaluators, and submits findings. Useful scan flags (see Scanning & Monitoring for details): --criteria / --skip-criteria / --levels to scan only the criteria your site can fail, --include-pdfs to grade linked PDF documents into the same VPAT (add --pdf-vision to opt into the rendered-page phase — deterministic rendered-text contrast, image-of-text detection, and a reading-order signal; also active whenever a vision LLM is configured), and — on the browser runner wholisphere-scan-browser — --flow <startUrl> for opt-in multi-step flow checks (SC 3.3.4 / 3.3.7).