ChemInteractions [v0.2.0-BETA]

A high-performance, Rust-based chemical interaction engine fusing ReactionT5 Neural Networks with Rule-Based Chemical Logic. Optimized for deployment via Docker and verified for high-accuracy standard organic reactions.

---

🚀 Quick Start (Docker - Recommended)

The easiest way to run the full stack (Backend + Frontend + Database) is using Docker Compose.

1. Requirements

• Docker and Docker Compose installed.
• (Pre-configured) The project includes a pre-seeded SQLite database and the ReactionT5 ONNX model.

2. Stand up the Services

# Build and start in detached mode
docker compose up -d

3. Access the Dashboard

Navigate to: 👉 http://localhost:8080

---

🛠 Local Development Setup

If you prefer to run natively on your machine:

1. Requirements

• Rust 1.92+
• SQLite3
• ONNX Runtime: Managed by the ort crate.

2. Prepare the Database & SQLx

# Install sqlx-cli if you haven't
cargo install sqlx-cli --no-default-features --features sqlite

# Initialize the offline query cache (required for compilation)
cargo sqlx prepare -- --lib

3. Run the Server

cargo run --release --bin chem-interactions

---

🧪 Feature Examples

1. Scientific Input (SMILES/Names)

Reactant A	Reactant B	Conditions	Expected Result
CC(=O)O (Acetic Acid)	CCO (Ethanol)	reflux, H2SO4	Fischer Esterification → Ethyl Acetate
NC1=CC=C([N+](=O)[O-])C=C1	(None)	H2, Pd/C	Nitro Reduction → p-Phenylenediamine
C1=CC=CC=C1 (Benzene)	ClCC	AlCl3	Friedel-Crafts Alkylation → Ethylbenzene

2. Natural Language Input

The engine uses a heuristic parser to extract reactants and conditions from text:

• Query: "React Acetic Acid with Ethanol under reflux with an acid catalyst."
• Parse: Reactants: ["Acetic Acid", "Ethanol"], Conditions: { heat: true, catalyst: "Acid" }.

3. API Reference

POST /api/predict

curl -X POST http://localhost:8080/api/predict \
     -H "Content-Type: application/json" \
     -d '{
       "reactants": ["CC(=O)O", "CCO"],
       "conditions": "reflux with H2SO4"
     }'

GET /api/compound/:name

Resolves metadata from PubChem or Local Cache.

curl http://localhost:8080/api/compound/Aspirin

GET /api/reactions

Lists all available textbook rules in the Rule Brain.

curl http://localhost:8080/api/reactions

---

🛠 Project Structure

• /frontend: "Supreme Scientific" minimalist dashboard (Vanilla JS/CSS).
• /src/predictor: Fusion engine, Rule brain (textbook-based), and ML brain (ReactionT5).
• /knowledge_base: JSON-based textbook reaction rules (Fischer, Grignard, etc.).
• /models: ONNX weights for ReactionT5.
• /data: Persistent SQLite storage.

---

🏗 CI/CD & Dockerization

The project uses a multi-stage Docker build based on Ubuntu 24.04 to satisfy the glibc 2.38+ requirements of the ort ONNX runtime. Offline SQLx compilation is enabled via .sqlx metadata.