Skip to content

TECH_STACK.md

Latest commit

 

History

History
244 lines (180 loc) · 5.17 KB

TECH_STACK.md

File metadata and controls

244 lines (180 loc) · 5.17 KB

Technology Stack

Complete technology stack and dependencies for the Safety-by-Design Entrainer Selection Framework

📋 Table of Contents

Core Technologies

Python Environment

Component Version Purpose
Python 3.11+ Core runtime
pip Latest Package management
venv Built-in Virtual environment

Data Science Stack

Library Version Purpose
NumPy ≥1.24 Numerical computing
Pandas ≥2.0 Data manipulation
SciPy ≥1.11 Scientific computing
Matplotlib ≥3.7 Visualization
Seaborn ≥0.12 Statistical visualization

Phase-Specific Dependencies

Phase I: Domain Mapping

Library Purpose
pubchempy PubChem API access
rdkit Molecular fingerprints & clustering
scikit-learn K-means clustering
hdbscan Density-based clustering

Phase II: Multi-Vector Selection

Engine A: Graph-RAG

Library Purpose
google-generativeai Gemini API integration
neo4j Graph database driver
chromadb Vector embeddings storage
langchain RAG orchestration

Engine B: TRIZ Multi-Agent

Library Purpose
google-generativeai Gemini API for agents
pydantic Data validation
asyncio Async agent coordination

Engine C: Cheminformatics

Library Purpose
rdkit Molecular descriptors
mordred Extended descriptors
scikit-learn Diversity selection

Phase III: Graph Traversal

Library Purpose
neo4j Graph database operations
networkx Graph algorithms
rdkit Tanimoto similarity

Phase IV: Bayesian Optimization

Library Version Purpose
botorch ≥0.9 Multi-objective BO
gpytorch ≥1.11 Gaussian processes
torch ≥2.0 Deep learning backend

Phase V: Process Simulation

Component Purpose
pythoncom COM automation (Windows)
win32com DWSIM interface
thermo UNIFAC calculations

Infrastructure

Databases

Database Version Purpose
Neo4j Community 5.x Graph storage for molecular relationships
ChromaDB Latest Vector embeddings for RAG
SQLite Built-in Local caching and results storage

External APIs

API Purpose Rate Limits
PubChem PUG REST Chemical data retrieval 5 req/sec
Google Gemini LLM for Graph-RAG & TRIZ Varies by tier
ChemSpider Supplementary chemical data API key required

Process Simulation

Software Version Platform
DWSIM 8.x Windows (COM automation)

Development Tools

Code Quality

Tool Purpose
black Code formatting
isort Import sorting
flake8 Linting
mypy Type checking
pre-commit Git hooks

Testing

Tool Purpose
pytest Test framework
pytest-cov Coverage reporting
pytest-asyncio Async test support
hypothesis Property-based testing

Documentation

Tool Purpose
mkdocs Documentation site
mkdocs-material Theme
mkdocstrings API documentation

Version Requirements

Minimum Requirements

Python >= 3.11
Neo4j >= 5.0
DWSIM >= 8.0 (Windows only)

Recommended Versions

Python 3.11.x (tested)
Neo4j 5.15.0
DWSIM 8.6.8

Hardware Requirements

Component Minimum Recommended
RAM 8 GB 16 GB
Storage 10 GB 50 GB
GPU Not required CUDA-capable (for BoTorch)

Installation Notes

RDKit Installation

RDKit requires conda or a pre-built wheel:

# Option 1: Conda (recommended)
conda install -c conda-forge rdkit

# Option 2: pip (if available for your platform)
pip install rdkit

Neo4j Setup

# Using Docker
docker run -d \
  --name neo4j \
  -p 7474:7474 -p 7687:7687 \
  -e NEO4J_AUTH=neo4j/password \
  neo4j:5-community

DWSIM Setup (Windows)

  1. Download DWSIM from dwsim.org
  2. Install with COM automation support enabled
  3. Register COM objects (usually automatic)

Dependency Graph

entrainer-selection
├── Core
│   ├── numpy, pandas, scipy
│   └── pydantic, pyyaml
├── Cheminformatics
│   ├── rdkit
│   ├── pubchempy
│   └── mordred
├── Machine Learning
│   ├── torch
│   ├── botorch
│   └── gpytorch
├── Graph/RAG
│   ├── neo4j
│   ├── chromadb
│   └── langchain
├── LLM
│   └── google-generativeai
└── Simulation
    ├── thermo
    └── pywin32 (Windows)