The warehouse is the personal data layer your server maintains locally: what AI reads through Recued, and what recipes compute over. It exists so a question like "what is the state of this deal" is answered in one pass over your actual data — without that data living in someone else's cloud.
#Collections
| Collection | Contents | How it fills |
|---|---|---|
| Messages from your mail accounts | Provider sync, such as IMAP | |
| Calendar | Events from your calendars | Provider sync |
| Contacts | A contact graph keyed by email address | Derived from mail and calendar, plus manual entry |
| Files | Files and message attachments | Watched paths and inbound attachments |
| CRM | Snapshots of deals, contacts, and accounts | Reconciled from vendors such as HubSpot and Salesforce |
| Memory | Audit entries with provenance links | Written by the engine as things run |
CRM rows are reference snapshots attached to platform-resident records — the CRM remains the system of record, and the warehouse holds what recipes and AI need to reason locally. Vendor events (webhooks or streaming) accelerate freshness; periodic reconciliation is the safety net.
#Enrichments
Enrichments are precomputed facts attached to records — a deal's health score, a contact's engagement level, a thread's signals. Some are deterministic aggregations; some are produced by AI. Every enrichment topic carries its own trust setting — off, manual, or automatic — and AI-produced topics never switch to automatic silently: Recued suggests after you have run one manually a few times, and you promote.
#Housekeeping
Maintenance runs between your activity on an idle-driven cadence you pick: compaction, cache eviction, enrichment refresh, CRM reconciliation. Background AI respects the per-topic trust settings and the global Pause-AI switch.
#Timelines
The links between collections power a per-entity timeline: one chronological feed for a contact or a deal across mail, meetings, CRM changes, annotations, and memory. See Audit and memory.
#Local by design
The warehouse lives in your server's data directory. It does not sync to the cloud, and no Recued cloud component can read it. Paired clients display it; the authority stays on the server.
#Related
- Connections — the credentials and endpoints that feed vendor-backed collections.