How AuthorNX handles data
A plain description of what the pilot does today. No certifications are claimed.
Workspace isolation
Every mission, opportunity, book, pen name and audit record belongs to one organization. PocketBase access rules enforce who can read or change it, and automated tests check that one workspace cannot see another.
Credentials
DataForSEO and OpenRouter keys live only in the worker's server environment. They are never sent to the browser or stored in records anyone can read.
Spending
Paid research is off by default. Every billable call reserves its maximum estimated cost against mission, daily and account caps first, and is refused when the cost is unknown.
Provider data
Responses are kept only as long as each provider's terms allow and are used internally. Open Library and Wikimedia data are open; DataForSEO and Stack Exchange terms are recorded with each observation.
AI
AI output is labeled as model inference, never as evidence or a measured number. Imported text is treated as data, never as instructions.
Publishing
AuthorNX never uploads to KDP, changes a listing, spends on advertising or publishes anything. People do those steps after approval.