Data, privacy, and where your content lives
What we store
- Transcripts of every episode we’ve processed — the raw text output of transcription
- Embeddings — 512-dim vector representations of transcript chunks, used for semantic search
- Enrichment metadata — structured extraction (speakers, series, bible verses for sermons, topics)
- Usage counters — how many searches / chat turns your key has served this month
- Your account row — email, segment (church or podcast), display name, account creation timestamp
Where it lives
- Postgres on Supabase — structured data (feeds, episodes, chunks, embeddings via pgvector, embed keys, customer accounts)
- Cloudflare R2 — bulk transcript files, episode audio metadata
Who can access it
- You — via the customer console, scoped to feeds you own
- Mark and operators — via the internal admin tool, for support and platform maintenance
- Nobody else. Nothing goes to third-party training data. Nothing is shared with other customers. Model providers (OpenRouter → OpenAI embeddings, Gemini enrichment + chat) see request text scoped to your feed at query time only; they don’t retain or train on it under our arrangement.
How to delete your content
Delete your feed from Console → Feeds → <feed> → Settings → Delete feed. Everything — transcripts, embeddings, usage records, enrichment — goes with it via cascade delete. The operation is immediate and irreversible.
Compliance
- SOC 2 Type II on the auth + data layer (via Supabase)
- GDPR / CCPA — right-to-delete honored via the feed delete action above
- HIPAA — available on request; contact support if your feed contains health-adjacent content that requires it
This is a stub page — expanded content and formal privacy policy coming in a follow-up.