addie-models¶

ADMET (absorption, distribution, metabolism, excretion, toxicity) prediction. 31 base endpoints + a 22-model TDC state-of-the-art overlay covering CYP inhibition/substrate, clearance, half-life, hepatotoxicity (DILI), cardiotoxicity (hERG + DICTrank), Ames, permeability, solubility, and the Tox21 nuclear-receptor/stress-response panels.

Source: https://github.com/NovoMCP/addie-models
Weights: https://huggingface.co/NovoMCP/addie-models (~510 MiB, MIT) — pulled automatically on first boot.

Pre-reqs¶

Docker
CPU works; GPU (any NVIDIA) speeds up batch inference
~4 GB RAM
~1 GB disk for weights, downloaded from Hugging Face on first boot (no cloud credentials needed)

Deploy¶

Via docker-compose:

Uncomment the addie-models block in docker-compose.yml, then:

docker compose up addie-models

Standalone:

docker run -d \
  --name novomcp-addie \
  -p 8025:8025 \
  --restart unless-stopped \
  ghcr.io/novomcp/addie-models:latest
# first boot downloads the weights from Hugging Face, then serves on :8025

# With GPU: add  --gpus all

Wire into the engine¶

export ADDIE_MODELS_URL=http://localhost:8025

Verify¶

curl -s http://localhost:8025/health
# {"status":"healthy","models_loaded":31, ...}

curl -s -X POST http://localhost:8025/addie/process \
  -H 'Content-Type: application/json' \
  -d '{"molecules":[{"id":"aspirin","smiles":"CC(=O)Oc1ccccc1C(=O)O"}]}'

Tools that light up¶

predict_admet, ADMET across all endpoints per molecule
get_molecule_profile, fills the ADMET section (falls back to properties-only without this service)
screen_library, batch ADMET across a compound list
batch_profile, same, higher-level wrapper

Env vars¶

Var	Default	Purpose
`PORT`	`8025`	HTTP listen port
`STORAGE_BACKEND`	`HF`	Weights source: `HF` (Hugging Face) \| `S3` \| `AZURE`
`HF_MODEL_REPO`	`NovoMCP/addie-models`	Hugging Face weights repo (when `STORAGE_BACKEND=HF`)
`EXECUTOR_WORKERS`	`8`	Inference thread-pool size
`TDC_USE_GIN`	`true`	Use the GIN-featurized TDC ensembles

GPU is auto-detected at boot (PyTorch); no flag needed beyond --gpus all on the container.