Giluma is a chapter analysis tool for fiction writers. When you submit a chapter, a room of specialist agents reads it and returns editorial observations. This policy explains what we collect, how we store it, and what we never do with it.

When you sign up for early access, we collect your name and email address, and any optional information you share (manuscript stage, genre, what you find hardest about writing).

When you use the editorial platform, we collect:

• The chapter text you submit for analysis
• The intent statement you provide (optional)
• The editorial observations our agents generate
• Your feedback on those observations (agree, disagree, add context, ignore)
• The Personal Book Model — a per-book memory of patterns, exceptions, and confirmed intentions you build over time

All data is stored in Supabase, hosted in the EU (Zurich, Switzerland). Data is encrypted at rest and in transit.

Your manuscript text, observations, and feedback are stored in a PostgreSQL database accessible only to the Giluma application. No third party has access to your manuscript content.

Your manuscript data is used for one purpose: editorial analysis within your session. The observations our agents generate are derived from your chapter and returned to you.

Your feedback is used to update your Personal Book Model, which shapes future sessions for your book — suppressing patterns you've already addressed, and remembering decisions you've confirmed as intentional.

We do not sell your data. We do not share your manuscript with advertisers, analytics platforms, or any third party. We do not use your chapter content for any purpose other than returning editorial observations to you.

Giluma uses the following third-party services:

• Groq — provides the inference infrastructure that powers Giluma's specialist agents. Groq receives your chapter text per request. We use Groq's API under terms that prohibit training on API inputs.
• Supabase — database infrastructure, hosted in Zurich.
• Resend — transactional email delivery for signup confirmations. Receives your email address only.
• Cloudflare — CDN and DNS for giluma.com. Processes HTTP requests but does not receive manuscript content.
• Plausible Analytics — privacy-first, cookieless analytics for site traffic. No personal data collected, no cross-site tracking.

You can request deletion of all your data at any time. Send a request to giluma.dev@gmail.com and we will remove your account, manuscript submissions, observations, and Personal Book Model within 30 days.

You can also request a copy of the data we hold for you. We will respond within 30 days.

For any privacy questions or data requests: giluma.dev@gmail.com