Privacy
No account required. Your current session is enough.
MVP privacy notice · October 7, 2026. This notice describes the public Midnight Repertory service.
What this build stores.
Seen or rejected titles, platform preferences and clarification or refinement context stay in this tab’s session storage. The build expires that context after six hours of inactivity and you can clear it with “Reset this session.” No personal taste profile or persistent watch history is built.
Search requests go to a Cloudflare Worker so it can return recommendations. Live search sends the necessary query, anonymous search context and stored movie evidence to OpenAI for model processing. The API request asks OpenAI not to store the response; this is not a promise of zero retention, and OpenAI may separately retain limited data for abuse monitoring under the applicable service and security policies.
Operational data.
Midnight Repertory stores the raw text of each submitted search in its first-party Cloudflare D1 database for retrieval-quality analysis, together with non-sensitive metadata: response type and error category, result and match-class counts, whether the review limit was reached, total server duration, and the deployment/retrieval mode. Production query analytics are retained for 30 days, with cleanup after writes and on a daily schedule.
The analytics table does not store IP addresses, account identifiers, session exclusions, platform preferences, private narrative evidence, model prompts, or model request and response bodies. Raw searches are not copied into ordinary Worker or provider logs and are not sent to third-party analytics. Hosting and model providers may separately process standard request metadata under their own service and security policies.
The counter records visits without attaching a personal identity. It counts sessions, not verified unique people. No third-party analytics script, advertising tracker or tracking cookie is included.
Links and images.
IMDb links open a separate website with its own privacy practices. This build uses no external poster requests. If posters are added later, their source and request behavior will be disclosed here.
Your data isn’t the product.
Individual taste profiles and personal watch histories are not sold as identifiable data. Any future commercialization of aggregate or anonymized usage data will be disclosed here before it starts.
Questions.
This notice will be updated if the service’s data practices or processors materially change.