Project Overview

Project Overview

rust-ai-examples 🦀🤖

This project is designed to assist those who wish to work with the Rust language in the field of artificial intelligence and its sub-disciplines.

Goals

• Provide practical, runnable examples of AI/ML in Rust
• Demonstrate popular Rust ML libraries (Candle, Linfa, Burn, etc.)
• Cover various domains: NLP, Computer Vision, Traditional ML, Deep Learning
• Serve as a learning resource for Rust developers interested in AI

Libraries Used

Library	Purpose	Examples
Candle	Tensor operations, neural networks	Text Classification, Sentiment Analysis, OCR, Image Classification
Linfa	Classical ML algorithms	KNN, Decision Tree, Linear Regression, Language Detection
Tokenizers	Text tokenization	Text Classification, Sentiment Analysis
Hugging Face Hub	Model hosting	Sentiment Analysis, OCR, Image Classification

Example Categories

Machine Learning (Traditional)

• Decision Tree Classifier - Loan approval prediction
• House Price Prediction - Linear regression
• KNN Classifier - Iris dataset classification

Deep Learning

• Image Classification - ResNet-18 on ImageNet
• OCR - TrOCR for text recognition

Natural Language Processing

• Keyword Extraction - TF-IDF based extraction
• Language Detection - Character n-gram features
• Sentiment Analysis - DistilBERT for sentiment
• Text Classification - BERT for custom datasets
• Email Spam Classification - Transformer-based spam detection

Repository Structure

rust-ai-examples/
├── examples/              # All example projects (standalone Cargo crates)
│   ├── OCR/
│   ├── decision-tree-classifier/
│   ├── email-spam-classification/
│   ├── house-price-prediction/
│   ├── image-classification/
│   ├── keyword-extraction/
│   ├── knn-classifier/
│   ├── language-detection/
│   ├── sentiment-analysis/
│   └── text-classification/
├── src/content/docs/      # Documentation content (this site)
├── public/                # Static assets
├── package.json           # Node.js dependencies
├── astro.config.mjs       # Astro configuration
├── tsconfig.json          # TypeScript configuration
├── Cargo.toml             # (Not at root - each example has its own)
└── README.md              # Project README

Anatomy of an Example

Every example is an independent Cargo crate. They share a common shape, but the number of modules and required assets varies:

example-name/
├── Cargo.toml               # Its own manifest and dependencies
├── src/
│   ├── main.rs              # Entry point (usually the whole pipeline)
│   ├── model.rs             # Optional: model/classifier code
│   ├── dataset.rs           # Optional: data loading
│   └── ...                  # Optional: preprocessing, downloader
├── data/                    # Optional: committed datasets
└── models/                  # Optional: model files (committed or downloaded)

Example	Source files	Requires
Text Classification	src/main.rs	Local models/ BERT checkpoint
Sentiment Analysis	src/main.rs	Downloads models/ from Hugging Face
Email Spam Classification	(none yet)	BERT assets present; source missing
Language Detection	src/{main,model,dataset}.rs	data/languages.json
Keyword Extraction	src/main.rs	.txt files in documents/
KNN Classifier	src/main.rs	data/iris.csv
Decision Tree Classifier	src/main.rs	data/loans.csv
House Price Prediction	src/main.rs	data/house.csv
Image Classification	src/{main,model,preprocessing,downloader}.rs	Downloads models/ on first run
OCR	src/{main,image_processor}.rs	Download helpers only; inference missing

Completion Status

Example	Runnable?	Notes
Decision Tree Classifier	✅ Yes	Trains and evaluates on the committed CSV
House Price Prediction	✅ Yes	Trains and evaluates on the committed CSV
KNN Classifier	✅ Yes	Hand-written KNN with full metrics
Language Detection	✅ Yes	Trains on languages.json, then interactive
Keyword Extraction	✅ Yes	Reads documents/*.txt
Sentiment Analysis	✅ Yes	Downloads DistilBERT, then predict/evaluate
Text Classification	✅ Yes	Trains a linear head on a local BERT
Image Classification	✅ Yes	Downloads ResNet-18, then classifies an image
Email Spam Classification	❌ No	Assets only; no src/ yet
OCR	❌ No	No fn main; inference not implemented

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.