Storage

Placeholder. Where records live after the collector receives them.

• Input: raw records from collector.py.
• Output: queryable store for ../03-analyze/.
• Data it touches: everything measured; this is where retention and deletion apply.

Coverage (tentative)

• First: append-only JSONL per day, queried with DuckDB or SQLite. Zero services, enough for one team.
• Next: Postgres for orgs that already run one.
• Large orgs: ClickHouse, or the existing OTel backend (../04-auto-analysis/cloud-logs.md).

Pitfalls

• Each record repeats user email and ~10 identity attributes: hash or drop on ingest if the owner asks.
• Retention and deletion requests (GDPR) must work per user.
• Raw records stay unmodified; derived tables are rebuilt from them, never edited in place.
• Data residency: the store's region is the owner's choice, stated at deploy.