MultiClaw-MLFlow 🧪🦞

MultiClaw-MLFlow is AIML Solutions’ model lifecycle and experiment governance lane.

It gives MultiClaw teams a reproducible system for training, evaluating, registering, and auditing models across quant and agentic workflows.

What this repo does

• Runs MLflow tracking + model registry
• Stores metadata in Postgres and artifacts in MinIO (S3-compatible)
• Provides baseline PyTorch + HF tracked experiment flow
• Defines architecture and runbook standards for expansion

Verified status

• mlflow-db, mlflow-minio, mlflow-tracking stack validated
• Tracking UI available at http://localhost:5000
• Artifact bucket mlflow-artifacts verified
• Sample training run completed with logged metrics + artifacts

Core docs

• docs/architecture.md
• docs/runbook.md
• docs/EXPERIMENT_STANDARD.md
• docs/ROADMAP.md

Quick start

cd infra
cp .env.example .env
docker compose up -d

MLFLOW_TRACKING_URI=http://127.0.0.1:5000 \
MLFLOW_S3_ENDPOINT_URL=http://127.0.0.1:9000 \
AWS_ACCESS_KEY_ID=minio \
AWS_SECRET_ACCESS_KEY=minio_dev_change_me \
python3 services/training/sample_mlflow_hf_torch_run.py

Contributing

See CONTRIBUTING.md.

License

MIT — see LICENSE.