Kuulua

How search works & privacy

Effective October 16, 2026. Applies to submitted searches and follow-up choices in the food and places working preview.

How Kuulua builds an answer

Kuulua brings together source-backed local information, question interpretation, scoped retrieval and a presentation suited to the request.

  1. Understand the request. Identify the place, requested information and relevant conditions. OpenAI may assist with interpretation; the proposed interpretation is checked before it is used.
  2. Retrieve saved evidence. Use the available local records and their source links. A model interpretation does not create business facts, confirm availability or approve an unreviewed submission.
  3. Present the requested information. A question about food types should show published food descriptions with the named carts. Profiles, menus and source details provide further context where available.
  4. Keep limitations visible. Coverage is limited to the saved sample. Publication dates, missing evidence and corrections matter; a listing does not prove a complete menu or current opening.

The preview combines these functions. OpenAI is one interpretation component, not the source of local business information or the whole Kuulua service.

What Kuulua records

Kuulua runs submitted searches on its server and records private diagnostics: the question, recognized needs, answer coverage, unresolved conditions, dataset version, server time, and any follow-up choice used to interpret the request. These records help us find missing information and improve search. Search records do not establish business facts or prove that an offering is absent locally.

Question interpretation

Kuulua may send your submitted question and a small catalog of public names and search terms to OpenAI to interpret the request. The full question can be sent before diagnostic redaction, so please leave out personal or sensitive details. OpenAI interpretation does not establish business facts; Kuulua retrieves answers from its saved evidence. We request no stored response using the API’s store=false setting. OpenAI may retain abuse-monitoring content for up to 30 days by default, with exceptions under its policies, and does not use API content for training by default. See OpenAI data controls.

Validated interpretations may be cached in encrypted private storage for up to 24 hours. Each use is revalidated and retrieves evidence again. Expired entries are excluded immediately and included in scheduled cleanup. Private per-call cost metadata contains no question or response text and is kept for 90 days; monthly cost totals have no automatic expiry.

Keep personal details out of searches

We apply patterns to remove email addresses, links, phone-like numbers and some street addresses before saving diagnostics. These patterns are incomplete; they cannot reliably identify every name or personal detail. Saved diagnostic text is encrypted. Please do not include sensitive personal information.

Retention and access

Detailed server diagnostics expire after 30 days. Every 15 minutes, Kuulua processes private historical summaries before deleting expired details. Summaries retain daily execution counts, recognized catalog topics and combinations, answer coverage and diagnostic categories, separated by dataset version and search or follow-up. They exclude question text, personal identifiers and individual record IDs and currently have no automatic expiry. Processing or cleanup failures can delay physical deletion of details so unsummarized records are not lost; expired rows are excluded from our detailed export tool. Cloudflare, our hosting provider, may retain recovery copies under its own service retention. Access is limited through our hosting account and private local tools; there is no public search-history endpoint. Private review exports must be managed and deleted separately; server cleanup does not erase downloaded copies.

Follow-up questions and additional results pages use encrypted sessions that expire after 10 minutes and can be used once. Results-page sessions preserve the original search and scope so loading more does not create another submitted-search record or require another interpretation. Expired sessions are included in scheduled cleanup. Initial example results, page views, typing, and clearing a search do not create a submitted-search record. Automated verification records are kept separate from visitor searches.

Technical request information

Kuulua does not store IP addresses, browser user-agent strings, cookies, or advertising identifiers in search diagnostics. An address-derived, short-lived key is used for request limiting. Cloudflare processes requests and may keep infrastructure information under its own policies. We do not treat search counts as unique people or completed visits.

Use of search diagnostics

These diagnostics support search improvement. They do not subscribe you to marketing, create a social feed, automatically update business information, or send messages to businesses. An executed search record does not prove that your browser displayed or that you were satisfied with its result.

Kuulua is produced and owned by Waybound Strategies LLC.