Table of Contents

Legal and Operational Policy

AI Reader API processes caller-supplied public URLs and converts retrieved content into AI-ready Markdown, text, or structured JSON.

This document is an operational product policy, not legal advice. Callers are responsible for confirming that their use is permitted for each target website, jurisdiction, and downstream workflow.

Caller Responsibility

Callers are responsible for:

  • Complying with applicable laws.
  • Complying with target website terms of service.
  • Respecting copyright, database rights, licensing terms, and attribution requirements.
  • Ensuring they have rights to store, transform, embed, summarize, redistribute, or commercially use processed outputs.
  • Avoiding requests that target private, local, internal, or metadata network resources.

OngsooLabs API does not grant rights to third-party content.

Target-Site Access Limits

AI Reader API is designed for safe processing of public webpages. It may refuse, fail, or limit requests when a target page:

  • Blocks automated retrieval.
  • Requires authentication or private session state.
  • Appears to target local, private, or cloud metadata network ranges.
  • Produces unusable or unsafe content.
  • Requires access behavior outside the product's AI data acquisition scope.

The service should not be used to bypass paywalls, authentication, authorization, bot protections, or technical access controls.

Content Accuracy

Web extraction is not guaranteed to be perfect.

Results can vary because of:

  • Target-site layout changes.
  • JavaScript rendering behavior.
  • Regional or network-dependent content.
  • Anti-automation controls.
  • Duplicated layout text.
  • Missing metadata.
  • Language or formatting ambiguity.

Use quality, usage, and execution metadata to decide whether a result is suitable for the caller's workflow.

Privacy and Logging

The reader should keep logs focused on operational telemetry.

Allowed aggregate telemetry includes:

  • HTTP method.
  • Target domain.
  • Success or failure.
  • Public error code.
  • HTTP status code.
  • Duration.
  • Execution profile.
  • Browser usage.
  • Cache status.
  • Retry count.
  • Reader mode.
  • Output option flags.
  • Character, token, and chunk counts.

Logs must not include:

  • Request bodies.
  • Response bodies.
  • Raw HTML.
  • Extracted content.
  • Cookies.
  • Custom headers.
  • Proxy data.
  • Raw query strings.

Cache Policy

Reader caching is a performance optimization, not archival storage.

Current cache policy:

  • Cache keys hash the target URL and include response-shaping options.
  • Cached reader entries can include the processed response and HTTP cache validation metadata.
  • Entries use a short operational TTL.
  • Requests with sensitive context skip reader result caching.
  • Cache failures should fall back to direct execution where possible.

The service should not market cache behavior as long-term storage, compliance retention, or customer-controlled archival history.

Abuse and Stability

OngsooLabs may limit or block traffic that threatens service stability, violates policy, or attempts to use the product outside its AI data acquisition scope.

Examples include:

  • High-volume crawling outside the active plan.
  • Repeated requests to blocked or unsafe targets.
  • Attempts to access internal network resources.
  • Automated stress testing without coordination.
  • Attempts to use the service as a generic proxy.

Support and Incident Handling

For persistent failures, callers should provide:

  • Request ID when available.
  • Endpoint and timestamp.
  • Public target domain.
  • Public error code.
  • Execution metadata such as profile, cache status, retry count, and duration.

Callers should not send secrets, cookies, private authorization headers, raw personal data, or proprietary source content in support requests unless a separate secure support process exists.