Data model
Practice this questionWANT TO UNDERSTAND THE CONCEPT FIRST?Learn Data model
| Record | Essential fields and purpose |
|---|---|
customers | Tenant, customer identity, permitted contact/profile fields; maintained by business service |
policies | Tenant/customer, product, region, validity, document version, endorsements; canonical applicability |
claims | Tenant/customer/policy, status, receipt, required documents; canonical claim state |
documents | Tenant, source ID/hash, MIME/size, ACL, version, parser state, publication status, timestamps |
chunks | Tenant/document/version, section/page, text, embedding model/dimension, vector, eligible metadata |
conversations/messages | Tenant/user/session, bounded content and roles, tool relationships, retention category |
agent_runs | Tenant/user, graph version, task, phase, deadline/budget, checkpoint reference, outcome |
claim_drafts | Draft ID/version, structured fields, source refs, content hash, creator, validation results |
approvals | Draft ID/hash, action, reviewer, scope, timestamp, expiry, decision; authenticated record |
operation_receipts | Tenant/action/key, payload hash, pending/completed/unknown state, external reference |
outbox | Business-linked event ID, payload, dispatch status; reliable async publication |
user_memories | Tenant/user, approved fact/preference, source, timestamp, expiry; separate from claim truth |
audit_events | Actor, action, object/version, decision, outcome, run/receipt reference; restricted immutable history |
- Enforce unique operation keys and optimistic draft versions.
- Keep large original files in protected object storage.
- Keep model-visible projections small.
- A document index and conversation cache are rebuildable derived data; claim and approval records need stronger durability.
Adapted from 13_AI.MD · “Data model”. Code examples are study fragments. Difficulty and core labels are editorial estimates; version-specific behavior should be checked against your project.