OMEGA Project
Architecture
OMEGA Core is a modular AI engineering evaluation platform running on a dedicated server. Services are containerised, separated by concern, and connected through isolated Docker networks. All architectural decisions are documented as ADRs.
Platform Topology
Platform Services
| Container | Image | Port | Binding | Stack |
|---|---|---|---|---|
| omegacore-ollama | ollama/ollama:latest | 11434 | LAN | ai |
| omegacore-open-webui | open-webui:v0.9.6 (pinned) | 3000 | LAN | frontends |
| omegacore-sillytavern | sillytavern/sillytavern:latest | 3002 | LAN | frontends |
| omegacore-librechat | danny-avila/librechat-dev:latest | 3004 | LAN | frontends |
| omegacore-postgres | pgvector/pgvector:pg16 | 5432 | localhost | databases |
| omegacore-qdrant | qdrant/qdrant:latest | 6333 | localhost | databases |
| omegacore-evidence | omegacore-evidence:latest | 8000 | LAN | evidence |
| omegacore-inference | omegacore-inference:latest | 8001 | LAN | inference |
| omegacore-influxdb | influxdb:2.7 | 8086 | LAN | observability |
| omegacore-ui | omegacore-ui:latest | 8080 | LAN | ui |
| omegacore-builder | omegacore-builder:latest | 8082 | internal only | ui |
| omegacore-frontend-manager | omegacore-frontend-manager:latest | 8090 | internal only | ui |
Services bound to localhost are reachable on the host only. Services bound to the LAN IP are accessible from the private network. No service is exposed to the public internet.
The Inference API (:8001) is the platform access point for AI research participants. Claude and GPT interact with OMEGA Core through this interface as part of the research programme.
omegacore-ui sits behind Authentik SSO (a self-hosted identity provider on the Proxmox side — see below) — no route on it is reachable without login, including the previously-open Model Builder. It never touches the Docker socket directly: a separate internal-only service, omegacore-frontend-manager, holds that privilege instead.
Docker Networks
| Network | Driver | Purpose |
|---|---|---|
| omega_internal | bridge | Cross-stack communication — container name resolution between stacks |
| ai_default | bridge | Internal to the AI stack — Ollama only now; frontends reach it via omega_internal instead |
| librechat_default | bridge | Internal to LibreChat's own stack (MongoDB, Meilisearch, its own pgvector, RAG API) |
| databases_default | bridge | Internal to the databases stack |
| evidence_default | bridge | Internal to the evidence stack |
Each compose stack uses its own isolated bridge network. omega_internalis an external network created manually on the host — it must exist before any stack is started. Services address each other by container name, not IP.
Public Infrastructure
| Component | Service | Purpose |
|---|---|---|
| Static site | AWS S3 | Hosts built site assets — deployed via GitHub Actions on merge to main |
| CDN | AWS CloudFront | Global distribution, HTTPS termination, OAC for private S3 access |
| TLS certificate | AWS ACM | Managed certificate for omegaproject.ai — auto-renewing |
| Encryption | AWS KMS | Server-side encryption for S3 objects |
| DNS (zone) | AWS Route 53 | Hosted zone for omegaproject.ai — aliases CloudFront distribution |
| DNS (authoritative) | Cloudflare | Authoritative nameservers — delegates to Route 53 NS records |
| Live status | S3 api/status.json | Platform monitoring bundle pushed every 5 min from OMEGA Core |
| CI/CD | GitHub Actions | Site deploy on site/** changes; Terraform apply on infrastructure/** changes |
Evidence API
The Evidence API (omegacore-evidence) is a read-only FastAPI service that exposes live platform state. It is the authoritative source of runtime evidence for auditing, AI-assisted review, and operational monitoring.
| Endpoint | Description |
|---|---|
| GET /health | API liveness check |
| GET /state/services | Live Docker container state and health |
| GET /state/databases | PostgreSQL connectivity and pgvector version |
| GET /state/models | Installed Ollama models |
| GET /state/system | CPU, memory, disk and load average |
| GET /state/report | Full platform report aggregating all collectors |
Inference API
The Inference API (omegacore-inference) is the platform access point for AI research participants. Claude and GPT interact with OMEGA Core through this interface as part of the AI security and evaluation research programme. It is LAN-only and not exposed to the public internet.
| Endpoint | Description |
|---|---|
| GET /health | API liveness check |
| GET /version | API and Ollama version information |
| GET /models | Available models on the platform |
| POST /query | Submit a prompt to a named model and receive a response |
Remote Access
Remote access to the OMEGA Core host is provided by Twingate, running as a Docker container on the Proxmox hypervisor — not on the OMEGA Core host directly. This isolates the access control plane from the platform itself.
Remote device → Twingate (Proxmox) → LAN → OMEGA Core host
Hardware
OMEGA Core runs on a MINISFORUM AI X1 Pro-470, purchased as a barebone unit and self-configured. Selected for AMD NPU capability, high memory bandwidth, PCIe 4.0 storage performance and future GPU expansion via OCuLink.
| Component | Specification |
|---|---|
| Host | MINISFORUM AI X1 Pro-470 (Barebone) |
| CPU | AMD Ryzen AI 9 HX470 — up to 86 TOPS |
| Memory | Crucial 64GB DDR5-5600 (2×32GB SODIMM) |
| Storage | WD_BLACK SN7100 2TB NVMe — 7,250 MB/s read, 6,900 MB/s write |
| Network | Dual 2.5GbE · WiFi 7 · Bluetooth 5.4 |
| AI Compute | AMD Radeon 890M iGPU — available for local model inference |
| Expansion | OCuLink port for external GPU (future) |
Design Decisions
All major architectural choices are documented as Architecture Decision Records. ADRs record what was decided, why, and what alternatives were considered.
| ADR | Decision |
|---|---|
| ADR-001 | Ubuntu Server as the platform host |
| ADR-002 | Docker for all service containerisation |
| ADR-003 | PostgreSQL with pgvector for relational and vector storage |
| ADR-004 | Qdrant as dedicated vector database |
| ADR-005 | Ollama for local model hosting and inference |
| ADR-006 | InfluxDB for time-series observability metrics |
| ADR-007 | Prompt Engineering Workbench — two-stage Model Builder design |
| ADR-008 | Evaluation frontends — running multiple chat UIs side by side, not picking one upfront |
| ADR-009 | Dev environment pattern for Open WebUI — superseded by ADR-011 |
| ADR-010 | Authentik for centralised SSO across OMEGA Core frontends |
| ADR-011 | Backup-before-upgrade, replacing the standing dev instance from ADR-009 |
| ADR-012 | omega-ui restructure — Status/Builder/Frontends behind login, Docker control isolated in its own service |
Full ADR documentation is maintained in the project wiki.