Data Containers¶

Executive Summary¶

Data Containers defines the logical data stores in the Algosure C4 Level 2 architecture: PostgreSQL, Redis, object and document storage, search index, vector store / semantic memory store, and future analytics warehouse.

Why This Exists¶

Algosure must distinguish transactional source facts, short-lived coordination state, file evidence, search projections, semantic memory, and future analytics. This protects DDD ownership, tenant isolation, auditability, AIOS memory governance, and event-driven data flow.

Owner¶

The owner is the Chief Product Officer and Enterprise Architect.

Business Value¶

Clear data container responsibilities prevent data sprawl, reduce source-of-truth ambiguity, and help teams design secure, auditable, tenant-aware data flows.

Data Container View¶

flowchart TB
    Backend[Spring Boot Modulith Backend]
    AIOS[AI Orchestration Service]
    Integration[Integration Gateway]
    Analytics[Future Analytics Consumers]

    Postgres[(PostgreSQL)]
    Redis[(Redis)]
    ObjectStore[(Object / Document Storage)]
    Search[(Search Index)]
    Vector[(Vector Store / Semantic Memory Store)]
    Warehouse[(Future Analytics Warehouse)]

    Backend --> Postgres
    Backend --> Redis
    Backend --> ObjectStore
    Backend --> Search
    Backend --> Vector
    Backend --> Warehouse
    AIOS --> Vector
    AIOS --> ObjectStore
    Integration --> ObjectStore
    Search --> Backend
    Warehouse --> Analytics

Data Container Responsibilities¶

Container	Logical Responsibility
PostgreSQL	Stores authoritative transactional records, tenant state, domain-owned facts, workflow state, audit references, configuration, and relational business data.
Redis	Supports short-lived cache, coordination, rate limiting, temporary state, and performance-oriented access patterns that do not replace durable source facts.
Object / document storage	Stores tender packs, compliance evidence, uploaded attachments, generated documents, signed files, exports, source documents, and binary evidence.
Search index	Provides searchable projections over domain records, documents, suppliers, opportunities, compliance evidence, contracts, and operational knowledge.
Vector store / semantic memory store	Stores embeddings, semantic references, memory projections, contextual retrieval records, and AIOS-approved memory artifacts.
Future analytics warehouse	Supports historical analytics, cross-domain reporting, trend analysis, executive insight, and downstream management information in a future architecture wave.

Ownership Model¶

Data Type	Owning Architecture Area	Container Support
Source domain facts	Owning Domain in the Spring Boot Modulith backend	PostgreSQL
Workflow state	Workflow and owning Domain modules	PostgreSQL, Redis where short-lived
Audit evidence references	Security, tenant isolation, audit, and owning Domain	PostgreSQL and object / document storage
Documents and files	Owning Domain or workflow	Object / document storage
Searchable projections	Owning Domain publishes or permits projection	Search index
Semantic memory	AIOS and relevant source Domains under governance	Vector store / semantic memory store
Analytical history	Future analytics architecture under source-domain lineage	Future analytics warehouse

Data Boundary Rules¶

Rule	Meaning
PostgreSQL is not a shared free-for-all	Tables and records must reflect domain ownership and tenant boundaries.
Redis is not authoritative	Redis must not become the source of durable business facts, approvals, compliance state, or payment state.
Object storage needs metadata ownership	Files require domain-owned metadata, tenant context, classification, retention, and evidence rules.
Search is a projection	Search index content must be rebuildable from owned sources and must not become the only source of truth.
Vector memory is governed	Semantic memory requires source attribution, tenant scope, approval where required, retention rules, and explainability support.
Analytics is downstream	The future analytics warehouse must consume governed events, extracts, or data products; it does not own operational decisions.
Tenant isolation is mandatory	Every data container must preserve tenant context, authorization, retention, and audit expectations.

Event-Driven Data Flow¶

flowchart LR
    Domain[Domain Event]
    Projection[Projection Builder]
    Search[(Search Index)]
    Memory[Memory Governance]
    Vector[(Vector Store / Semantic Memory Store)]
    Analytics[Analytics Ingestion]
    Warehouse[(Future Analytics Warehouse)]

    Domain --> Projection
    Projection --> Search
    Domain --> Memory
    Memory --> Vector
    Domain --> Analytics
    Analytics --> Warehouse

Non-Implementation Boundary¶

This document does not define:

Database schema.
Index mappings.
Storage bucket structure.
Cache key design.
Embedding model.
Retention configuration.
Backup configuration.
Analytics platform.
Data pipeline tooling.