Privacy Policy

What the search demand log collects

GPUSPECINDEX records one event when a visitor runs a catalog search. The event contains the selected search filters, the number of GPU variants returned, the time of the search, and a short pseudonymous visitor key. It does not contain a page URL, referrer, raw IP address, or stored user-agent string. The purpose is narrow: understanding which filter combinations and facet values people actually use, so future editorial comparisons are based on demonstrated demand instead of guesses. Search events are not used to build advertising profiles, determine identity, or infer interests outside this catalog.

How the visitor key works

The server temporarily reads the connecting IP address and user-agent to calculate `sha256(ip + user-agent + daily salt)`, then keeps only the first 16 hexadecimal characters. The raw inputs are not written to the events database. The randomly generated salt rotates daily, limiting the key’s usefulness across days. Even with those safeguards, a truncated salted hash is pseudonymous personal data rather than anonymous data, so this policy describes it plainly. The configured household WAN IP is excluded before hashing so operator searches do not enter the log.

Retention and sharing

Search events are retained for 90 days and then deleted by an automated host timer. GPUSPECINDEX uses no third-party analytics service. Event data is not sold, shared with advertisers, or combined with external profiles. Infrastructure providers may necessarily process network traffic to deliver the site, but the application’s demand database remains separate from the replaceable GPU catalog database and contains only the four fields described above.

Scope and changes

This policy covers the first-party search demand log. Normal server and network infrastructure may create short-lived operational logs needed for security and reliability; those logs are not the demand dataset used for content planning. If the collection purpose, event fields, retention period, or analytics providers change, this registry page must be updated before the changed collection begins. The current design intentionally favors aggregate questions—popular filter combinations, commonly selected facet values, and searches returning few results—over reconstructing a person’s browsing history.