OmniBioAI is a modular, AI-powered bioinformatics workbench designed to accelerate genomic research across:
- Local machines
- On-prem servers
- HPC environments (Slurm, Apptainer)
- Cloud infrastructure (AWS Batch, Azure Batch, Kubernetes)
With no mandatory cloud dependencies.
This repository is the workspace root of the OmniBioAI ecosystem — it assembles independently versioned components into a single runnable, production-grade stack.
OmniBioAI follows a four-plane architecture:
| Plane | Role | Key Components |
|---|---|---|
| Control | Orchestration, governance, APIs | Workbench, TES, ToolServer, Model Registry, LIMS |
| Compute | Ephemeral execution | Workflow runners, tool runtime containers, HPC adapters |
| Data | Artifacts, outputs, versioning | OmniObjects, model artifacts, workflow outputs |
| AI | Reasoning, retrieval, agents | RAG services, LLM integration, agent orchestration |
TES (Tool Execution Service) is the strict boundary between the control and compute planes.
Desktop/machine/
│
├── omnibioai/ # Workbench — Django platform, plugins, agents
├── omnibioai-tes/ # Tool Execution Service — HPC/cloud/local execution
├── omnibioai-toolserver/ # FastAPI ToolServer — validated async tool APIs
├── omnibioai-tool-runtime/ # Minimal cloud-agnostic container execution runtime
├── omnibioai-model-registry/ # Production model registry — versioning, provenance
├── omnibioai-lims/ # Lightweight Django LIMS — samples, metadata
├── omnibioai-rag/ # RAG assistant — Hugging Face + Ollama LLMs
├── omnibioai_sdk/ # Python SDK — API, object registry, notebooks
├── omnibioai-workflow-bundles/ # Engine-agnostic workflows — WDL, Nextflow, Snakemake, CWL
├── omnibioai-dev-docker/ # GPU dev environment — CUDA, JupyterLab, Ollama
├── omnibioai-control-center/ # Health dashboard, ecosystem report, orchestration
│
├── deploy/
│ ├── compose/ # Canonical Docker Compose files
│ ├── control-center/ # Control Center runtime (moved to omnibioai-control-center)
│ ├── scripts/ # Bootstrap utilities
│ ├── bundle/ # Offline image bundles
│ ├── hpc/ # Apptainer / Singularity assets
│ └── k8s/ # Kubernetes (post-beta)
│
├── db-init/ # Database initialisation scripts
├── data/ # Runtime data volumes
├── local_registry/ # Local model artifact storage
├── tmpdata/ # Temporary execution workspace
├── out/ # Report and analysis outputs
│ └── reports/
│ └── omnibioai_ecosystem_report.html # Generated ecosystem report
│
├── utils/ # Shared utilities
├── images/ # Container image assets
└── README.md
| Service | Port | Role |
|---|---|---|
| OmniBioAI Workbench | 8000 | UI, plugins, agents, AI tools |
| Tool Execution Service (TES) | 8081 | Workflow and tool orchestration |
| ToolServer | 9090 | Validated async tool APIs |
| Model Registry | 8095 | Versioned ML model artifacts |
| LIMS | 7000 | Sample and metadata management |
| Control Center | 7070 | Health dashboard, ecosystem report |
| MySQL | 3306 | Relational databases |
| Redis | 6379 | Celery task queue and caching |
All ports are configurable via .env.
- Docker Engine or Docker Desktop
- Docker Compose v2+
- Python 3.11+ (for report generation)
cp deploy/compose/.env.example deploy/compose/.env
docker compose \
--project-directory . \
--env-file deploy/compose/.env \
-f deploy/compose/docker-compose.yml \
up -dcurl http://127.0.0.1:8000 # Workbench
curl http://127.0.0.1:8081/health # TES
curl http://127.0.0.1:9090/health # ToolServer
curl http://127.0.0.1:8095/health # Model Registry
curl http://127.0.0.1:7070/health # Control Center
curl http://127.0.0.1:7070/summary # Ecosystem health summary (JSON)docker compose \
--project-directory . \
-f omnibioai-control-center/compose/docker-compose.control-center.yml \
up -dThe ecosystem report is a single interactive HTML file covering:
- Architecture — SVG lane diagram of all services and their connections
- Projects — Code line distribution across all repositories
- Languages — Language breakdown across the ecosystem
- Code Coverage — Per-repo pytest coverage with trend indicators
- Health Status — Live service and disk health from the Control Center
# With Control Center running (includes live health data)
python omnibioai-control-center/scripts/generate_report.py \
--root ~/Desktop/machine
# Without health data (faster, offline)
python omnibioai-control-center/scripts/generate_report.py \
--root ~/Desktop/machine \
--skip-health
# Custom Control Center URL
python omnibioai-control-center/scripts/generate_report.py \
--root ~/Desktop/machine \
--control-center-url http://localhost:7070- File:
out/reports/omnibioai_ecosystem_report.html - Browser: Open directly in any browser
- Control Center:
http://127.0.0.1:7070/report(served live when Control Center is running)
pip install cloc pandas
# pytest + pytest-cov for coverage collection (best-effort)The Control Center (omnibioai-control-center/) is the operational dashboard for the ecosystem:
GET /health— Control Center self-checkGET /services— Per-service health statusGET /summary— Full ecosystem summary (services + disk)GET /report— Serves the pre-generated ecosystem HTML report
Health checks cover HTTP endpoints, TCP ports (MySQL, Redis), and disk usage thresholds. Configuration lives in omnibioai-control-center/config/control_center.yaml.
| Mode | Control Plane | Compute Plane |
|---|---|---|
| Local dev | Docker Compose | Local Docker |
| On-prem | Docker Compose | Docker / TES |
| HPC | External VM | Apptainer via TES |
| Hybrid | VM | HPC + TES |
| Cloud | Kubernetes | Kubernetes |
- Single workspace root — all repos are siblings under one directory
- No absolute paths — fully portable across machines
- Strict service boundaries — control plane ≠ compute plane
- Restart-safe orchestration — ordered startup with health checks
- Container-native — OCI-compliant images throughout
- Environment-driven — all configuration via
.envand YAML - No forced cloud dependencies — runs fully offline and air-gapped
- Engine-agnostic workflows — WDL, Nextflow, Snakemake, CWL all supported
- Does not contain bioinformatics algorithms directly (these live in plugin repos)
- Does not vendor component repositories
- Does not enforce a single workflow engine
- Does not hide execution behind opaque AI calls
- Does not require external SaaS services
| Repository | Visibility | Description |
|---|---|---|
omnibioai |
Private | Workbench — plugin-based Django platform |
omnibioai-tes |
Private | Tool Execution Service — HPC/cloud/local backends |
omnibioai-toolserver |
Private | FastAPI tool execution APIs |
omnibioai-tool-runtime |
Private | Cloud-agnostic container execution contract |
omnibioai-model-registry |
Public | Production ML model registry |
omnibioai-lims |
Private | Laboratory Information Management System |
omnibioai-rag |
Private | RAG-powered bioinformatics assistant |
omnibioai_sdk |
Private | Python SDK — v1 complete |
omnibioai-workflow-bundles |
Private | Versioned engine-agnostic workflow bundles |
omnibioai-dev-docker |
Private | GPU AI development environment |
omnibioai-control-center |
Public | Health dashboard and ecosystem report |
| Component | Status |
|---|---|
| Multi-service orchestration | Stable |
| Tool Execution Service | Stable |
| ToolServer | Stable |
| Tool Runtime | Stable |
| Model Registry | Stable |
| LIMS | Stable |
| RAG assistant | Stable |
| Python SDK | v1 complete |
| Workflow bundles | Stable |
| Control Center | Active development |
| Ecosystem report | Active development |
| Kubernetes | Post-beta |
See individual repository LICENSE files.
Components are independently licensed.
omnibioai-model-registry and omnibioai-control-center are Apache 2.0.
OmniBioAI — reproducible bioinformatics at any scale, on any infrastructure.
© 2025 Manish Kumar. All rights reserved.