OmniBioAI Model Registry

OmniBioAI Model Registry is a production-grade lifecycle management system for AI/ML models within the OmniBioAI ecosystem.

It provides:

• Immutable model versioning (write-once versions)
• Cryptographic integrity verification (SHA256)
• Provenance-friendly metadata capture
• Staged promotion workflows (latest → staging → production)
• Deterministic resolution by stable reference (model@alias or model@version)
• Local-first design with a clean path to future backends (S3/Azure/on-prem)

iThe registry is implemented as a standalone Python library and includes:

• a CLI (omr)
• a minimal REST service (FastAPI)

---

Why This Exists

Biomedical AI requires:

• Reproducibility
• Auditability
• Governance
• Offline / air-gapped deployment
• Cross-infrastructure execution parity

Traditional ML tooling often assumes:

• cloud-first infrastructure
• mutable artifacts
• weak provenance guarantees

OmniBioAI Model Registry is designed differently.

It treats AI models as scientific artifacts that must be immutable, verifiable, and reproducible across environments.

---

Role in the OmniBioAI Architecture

OmniBioAI follows a four-plane architecture:

Plane	Responsibility
Control Plane	UI, registries, metadata, governance
Compute Plane	Workflow execution, HPC/cloud adapters
Data Plane	Artifacts, datasets, outputs
AI Plane	Reasoning, RAG, agents, interpretation

The Model Registry belongs to the Control Plane and provides:

• AI artifact governance
• deterministic inference references
• promotion and audit workflows
• infrastructure-independent model resolution

---

Core Design Principles

1) Immutability

Each model version is write-once:

• no overwrites
• no silent mutation
• full historical trace

This guarantees scientific reproducibility.

2) Integrity Verification

Every model package includes a SHA256 manifest:

• sha256sums.txt hashes the package contents (excluding itself)

This enables:

• bit-level reproducibility
• tamper detection
• trustworthy deployment in regulated environments

3) Provenance-Friendly Metadata

Each model stores structured metadata via model_meta.json, such as:

• training code version (git commit)
• dataset reference (e.g., DVC / object store ref)
• hyperparameters and preprocessing
• creator and timestamp

4) Promotion Workflow

Models move through controlled stages:

latest → staging → production

All promotions are:

• explicit
• append-only
• audited (audit/promotions.jsonl)

5) Storage Abstraction

v0.1.0 supports:

• local filesystem backend (localfs)

Planned:

• S3 / Azure Blob / enterprise on-prem backends

---

Repository Structure

omnibioai-model-registry/
├── omnibioai_model_registry/
│   ├── api.py
│   ├── config.py
│   ├── refs.py
│   ├── storage/
│   ├── package/
│   ├── audit/
│   ├── cli/
│   └── service/
├── tests/
├── pyproject.toml
└── README.md

---

Canonical Model Package Layout

Registered models follow a strict, portable structure:

<OMNIBIOAI_MODEL_REGISTRY_ROOT>/
tasks/<task>/models/<model_name>/
versions/<version>/
model.pt
model_genes.txt
label_map.json
model_meta.json
metrics.json
feature_schema.json
sha256sums.txt
aliases/
latest.json
staging.json
production.json
audit/
promotions.jsonl

This guarantees:

• deterministic loading
• integrity validation
• cross-environment portability

---

Install, Build, and Use as a Python Package

1) Configure registry root

The registry requires a root directory:

export OMNIBIOAI_MODEL_REGISTRY_ROOT=~/Desktop/machine/local_registry/model_registry

2) Install (editable) for development

From this repository root:

pip install -e .

Verify:

python -c "import omnibioai_model_registry as m; print('OK', m.__file__)"
omr --help

3) Build a wheel (distribution)

Install build tooling:

pip install build

Build:

python -m build

Artifacts are written to dist/:

• dist/omnibioai_model_registry-0.1.0-py3-none-any.whl
• dist/omnibioai_model_registry-0.1.0.tar.gz

Install the wheel:

pip install dist/*.whl

---

CLI Usage (omr)

Register a model package

omr register \
  --task celltype_sc \
  --model human_pbmc \
  --version 2026-02-14_001 \
  --artifacts /tmp/model_pkg \
  --set-alias latest

Resolve a model reference

omr resolve --task celltype_sc --ref human_pbmc@latest

Promote a version to production

omr promote --task celltype_sc --model human_pbmc --version 2026-02-14_001 --alias production

Verify integrity

omr verify --task celltype_sc --ref human_pbmc@production

Show metadata

omr show --task celltype_sc --ref human_pbmc@production --json

---

Python API Usage

from omnibioai_model_registry import register_model, resolve_model, promote_model

register_model(
    task="celltype_sc",
    model_name="human_pbmc",
    version="2026-02-14_001",
    artifacts_dir="/tmp/model_pkg",
    metadata={
        "framework": "pytorch",
        "model_type": "classifier",
        "provenance": {
            "git_commit": "abc123",
            "training_data_ref": "s3://bucket/datasets/pbmc_v1",
            "trainer_version": "0.1.0",
        },
    },
    set_alias="latest",
    actor="manish",
    reason="initial training",
)

# Resolve by alias (or version)
path = resolve_model("celltype_sc", "human_pbmc@latest", verify=True)
print("Resolved model dir:", path)

# Promote to production
promote_model(
    task="celltype_sc",
    model_name="human_pbmc",
    alias="production",
    version="2026-02-14_001",
    actor="manish",
    reason="validated metrics",
)

---

Minimal REST Service (FastAPI)

Run locally

pip install -r omnibioai_model_registry/service/requirements.txt
uvicorn omnibioai_model_registry.service.app.main:app --host 0.0.0.0 --port 8095

Test:

curl -s http://127.0.0.1:8095/health | python -m json.tool

Endpoints:

• POST /v1/register
• GET /v1/resolve
• POST /v1/promote
• POST /v1/verify
• GET /v1/show

---

Testing

pip install -e ".[dev]"
pytest -q

---

Relationship to OmniBioAI Ecosystem

This registry is a control-plane component of OmniBioAI.

Companion repositories:

• omnibioai → AI-powered bioinformatics workbench
• omnibioai-tes → execution orchestration across local/HPC/cloud
• omnibioai-rag → reasoning and literature intelligence
• omnibioai-lims → laboratory data management
• omnibioai-workflow-bundles → reproducible pipelines
• omnibioai-sdk → Python client access

The Model Registry provides the AI artifact governance layer shared by all.

---

Roadmap

Near Term

• additional storage backends (S3 / Azure)
• expanded metadata validation + schemas
• model listing and metadata search APIs

Mid Term

• RBAC and governance controls
• richer registry service APIs (auth, pagination, filtering)
• comparison and promotion policies

Long Term

• enterprise biomedical AI governance platform
• regulatory-ready audit and lineage
• deeper integration with experiment tracking and clinical pipelines

---

Status

• ✅ Immutable and verifiable model storage
• ✅ Audit-ready promotion workflow
• ✅ CLI + minimal REST service
• ✅ Local-first, cloud-ready design

OmniBioAI Model Registry establishes the foundation for trustworthy, reproducible biomedical AI deployment.