AMV TRADE BUDDY · OPERATOR LEDGER
Privacy policy
This ledger explains what the service processes, why it is needed, who provides infrastructure, and when operational records are removed.
Information the service processes
Market pages process public numeric source observations, source URLs, source timestamps, and normalized scores. The Trade Brief processes the selected item variants and current market evidence needed to evaluate that trade.
Contact and data-request tickets store the request type, supplied name, reply email, affected page URL, supporting source URL, message, consent time, status, and operational timestamps.
Anonymous limits and cookies
We use a signed anonymous cookie to enforce daily AI and request limits without requiring an account. It is Secure, HttpOnly, SameSite=Lax, scoped to this site, and lasts up to one year.
The server stores purpose-specific HMAC digests for AI and operator-request limits, so the two stored client digests cannot be linked from the cookie value. On Vercel only, the platform's overwritten forwarded address is also converted immediately to a separate purpose-specific HMAC digest for abuse resistance. We do not store raw IP addresses or forwarded network headers in the application database.
After an eligible visitor has used the regular daily AI allowance, we offer one daily self-attested reward for sharing the Trade Brief on an approved platform. We retain the AI-purpose HMAC digest, selected platform, and claim time for that daily reward. We do not store a social account or third-party post, and the reward is not proof that a post was made.
AI processing and telemetry
OpenAI processes bounded trade evidence to choose evidence priority, strategy, uncertainty focus, and reply tone. OpenAI does not write the market numbers, verdict, explanation, or final reply shown on the site; those are validated and rendered by the application. Requests are sent with provider storage disabled.
AI telemetry records a request ID, anonymous digest, model, bounded status and error code, latency, token counts when available, side item counts, market timestamp, and source count. It does not store raw prompts, full item lists, model prose, emails, cookies, raw IPs, credentials, or connection strings.
Processors and infrastructure
Vercel hosts the website and server functions. Neon hosts the PostgreSQL database. OpenAI provides the bounded strategy model. Each provider processes data under its own terms and privacy commitments. Public value-source sites receive ordinary crawler requests carrying the project's declared crawler identity.
Retention and choices
Daily quota digests and daily share-reward records retain only the current and previous UTC dates. AI telemetry is deleted after 30 days. Resolved and rejected operator requests are deleted after 90 days. Open or in-review requests remain while reasonably needed to investigate and reply. Market snapshots are retained to provide dated history and source provenance.
You may ask to access, correct, or delete information supplied in a ticket, subject to verification and legitimate operational needs. Use the data-request desk and include the relevant request ID if available.