Technology Stack

One database engine. Six storage modalities. Zero vendor lock-in. Every component is open source or open standard.

Four-Layer Architecture

OACIS is built on four layers. Each layer has a distinct role. The layers above depend on the layers below, but can be independently replaced.

Layer 4 — Model (Communication)

LLMs, dashboards, reports, natural language interfaces. Replaceable — the least trusted tier. AI is the last mile, not the first.

Layer 3 — Knowledge Graph (Truth)

PostgreSQL + AGE + pgvector + TimescaleDB. The competitive advantage — owned, not rented. Models are rented intelligence with diminishing returns. The graph is owned truth that compounds over time.

Layer 2 — Ontology (Meaning)

SUMO + CDM + CIM + FIBO + HL7 FHIR. Shared definitions — what things mean. When Genesys says "queue" and Cisco says "skill group," the ontology resolves them to the same concept.

Layer 1 — Trust Hierarchy (Governance)

7 trust levels × 6 confidentiality tags. What's reliable and who can see it. Enforcement via PostgreSQL Row-Level Security.

PostgreSQL — The Everything Database

A single PostgreSQL instance extended with specialized capabilities serves six storage modalities — eliminating the need for separate graph, vector, time-series, and analytics databases.

Apache AGE

Graph database with Cypher queries. Core knowledge graph — nodes, edges, traversals, path analysis, fraud ring detection.

pgvector

Vector similarity search. Semantic search, embedding-based deduplication, ontology nearest-neighbor matching.

TimescaleDB

Time-series analytics. Temporal versioning ("organizational time-lapse"), trend detection, drift scoring.

pg_duckdb

Columnar analytics & data lake. Lakehouse queries on Parquet, CSV, JSON, and Apache Iceberg on object storage.

Citus

Horizontal scaling. Distributed queries across shards for multi-tenant and large-scale deployments.

PostGIS

Geospatial queries. Jurisdiction mapping, geographic fraud velocity checks, location-based analytics.

Additional Extensions

tsvector

Full-text search (built-in)

pgAudit

Compliance audit logging

PgBouncer

Connection pooling

pg_cron

Job scheduling

pg_partman

Automated partitioning

postgres_fdw

Foreign data wrappers

Six Storage Modalities, One Engine

Traditional architectures require separate databases for each modality. OACIS consolidates them into a single platform with ACID transactions across all six.

Relational

PostgreSQL core. Structured entities, ACID transactions, tenant isolation via Row-Level Security.

Document

JSONB. Semi-structured configs, API payloads, flexible metadata — queryable with GIN indexes.

Time-Series

TimescaleDB. Temporal snapshots, trend analysis, organizational time-lapse with 10:1 compression.

Vector

pgvector. Embeddings, semantic search, similarity-based deduplication, ontology matching.

Graph

Apache AGE. Knowledge graph, Cypher traversals, relationship modeling, path analysis.

Data Lake

pg_duckdb. Parquet/Iceberg analytics, cold storage queries, data science workloads.

Temperature-Based Storage Tiering

Data moves through tiers based on access patterns and age:

🔴
Hot — PostgreSQL fast storage. Active operational data.
🟠
Warm — TimescaleDB compressed. Recent historical, compressed 10:1.
🔵
Cold — Iceberg on object storage. Analytical/archival via pg_duckdb.
Archive — Cheapest tier. Compliance retention, rarely accessed.

Application Stack

Every component is open source. Every format is open standard. No proprietary dependencies.

Backend

  • API — Flask + FastAPI
  • Task Queue — Celery + Redis
  • ORM — SQLAlchemy 2.0
  • Migrations — Alembic
  • Graph — Custom AGE client (Cypher)
  • Orchestration — Juju
  • Secrets — HashiCorp Vault

Frontend

  • Framework — React 19
  • Build — Vite
  • Language — TypeScript
  • Styling — Tailwind CSS
  • Serving — Nginx (production)

Infrastructure

  • Containers — Docker + Compose
  • Orchestration — Kubernetes / Juju
  • Monitoring — Prometheus + Grafana
  • CI/CD — GitHub Actions
  • Streaming — Apache Kafka
  • Search — Elasticsearch
  • Docs — Marker (PDF/OCR → MD)

Five AIO Subsystems

Five specialized subsystems share a single knowledge graph. A CVE discovered by SECURE-AIO automatically surfaces as a compliance gap in COMPLY-AIO and a risk factor in CX-AIO.

CX-AIO

Customer Experience. Contact center intelligence — 16-platform adapter, real-time analytics, fraud detection, quality management, workforce management.

BIZ-AIO

Business Operations. ERP, CRM, financial, HR, and operational data integration into the unified knowledge graph.

LEGAL-AIO

Legal Intelligence. Statute-to-case mapping, cross-jurisdictional analysis, regulatory change detection, deadline tracking.

COMPLY-AIO

Continuous Compliance. OSCAL/NIST/CIS/STIG framework management, continuous control verification, POA&M tracking.

SECURE-AIO

Threat Graph. 53 node types — SBOM ingestion, CVE correlation, OSCAL controls, attack path analysis, SASE/zero-trust, deepfake detection.

Want to See It in Action?

We demo the full stack — from raw data ingestion through the Conveyor pipeline to knowledge graph queries and dashboard visualization.

Request a Demo