DeepChem Server

A minimal cloud infrastructure for DeepChem

---

Table of Contents

1. Key Features
2. Requirements
3. Installation
   • Docker (recommended)
   • From Source
4. Quick Start
5. Project Structure
6. Contributing
7. License

---

Key Features

• FastAPI Backend – blazing-fast web framework with automatic, interactive API docs (Swagger UI & ReDoc).
• DeepChem Integration – built-in support for DeepChem featurizers and models.
• Flexible Datastore – pluggable storage layer (local disk by default) for datasets & models.
• Python SDK (pyds) – high-level client library for programmatic access & batch workflows.
• Containerised Deployment – first-class Docker support for easy, reproducible setups.

Requirements

DeepChem Server currently supports Python 3.7 – 3.10 and the requirements are listed in deepchem_server/requirements.txt.

---

Installation

Docker (recommended)

# 1. Clone the repo
$ git clone https://github.com/deepforestsci/deepchem-server.git
$ cd deepchem-server

# 2. Build & start (basic usage)
$ ./docker.sh

# Or specify a dockerfile (e.g., for GPU support)
$ ./docker.sh -f Dockerfile.gpu

# Show all options
$ ./docker.sh -h

Available options:

• -f, --dockerfile DOCKERFILE: Specify dockerfile to use (default: Dockerfile)
• -d, --data-dir DIRECTORY: Specify data directory (default: ./local_datastore)
• -h, --help: Show help message

Available Dockerfiles:

• Dockerfile: Standard CPU-based setup
• Dockerfile-GPU: GPU-accelerated setup with NVIDIA PyTorch base

The API will be live at http://localhost:8000 with interactive docs at http://localhost:8000/docs.

From Source

# 1. Clone the repo
$ git clone https://github.com/your-org/deepchem-server.git
$ cd deepchem-server

# 2. Create & activate virtualenv (optional but recommended)
$ python -m venv venv
$ source venv/bin/activate  # Windows: venv\Scripts\activate

# 3. Install dependencies
$ pip install -r deepchem_server/requirements.txt

# 4. Run the server (development mode)
$ bash start-dev-server.sh  # or uvicorn deepchem_server.main:app --reload

---

Quick Start

1. Upload data

   curl -X POST \
     -F "[email protected]" \
     http://localhost:8000/data/uploaddata

2. Featurize molecules

   curl -X POST \
     -H "Content-Type: application/json" \
     -d '{"dataset_id": "<ID>", "featurizer": "ecfp"}' \
     http://localhost:8000/primitive/featurize

3. Retrieve results

   curl -O http://localhost:8000/data/<ID>/download

For a fully-worked example, run the integration test:

python pyds/tests/test_upload_featurize.py

---

Project Structure

├── deepchem_server          # FastAPI application & core modules
│   ├── core/                # Business logic (featurization, datastore, …)
│   ├── routers/             # API routes grouped by domain
│   └── main.py              # ASGI entry-point
├── pyds/                   # Python client (SDK)
├── docs/                    # Sphinx documentation (see docs/source)
├── Dockerfile & docker.sh   # Container build & helper script
└── tests/                   # Py-test suites & sample assets

---

Contributing

Contributions are welcome — check out CONTRIBUTING.md (coming soon) and open a pull request. Please ensure new code is covered by tests and conforms to the existing style guidelines.

---

License

DeepChem Server is released under the MIT License. See the LICENSE file for details.