This repository contains three practical AI applications that demonstrate how to build real-world AI systems using modern tools and frameworks. These projects are designed to help you move from theoretical ML concepts to building functional AI applications.
This repository supports the article: Stop Learning ML - Start Building AI Systems
A code analysis tool that uses GPT-4 to review code for bugs, performance issues, style improvements, and security concerns. Supports multiple programming languages including Python, JavaScript, Java, Go, and Rust.
Features:
- Real-time code analysis with streaming responses
- Multi-language support
- Structured JSON output with detailed feedback
- Security and performance recommendations
A conversational AI system that allows you to upload PDF documents and ask questions about their content. Built with LangChain and FAISS for efficient document processing and retrieval.
Features:
- PDF document processing and text extraction
- Intelligent text chunking and embedding
- Conversational chat interface
- Streaming responses for better UX
- Persistent chat history
A content moderation tool that uses Hugging Face's toxic-bert model to detect and flag inappropriate content in real-time.
Features:
- Toxicity detection and scoring
- Real-time content analysis
- Configurable blocking thresholds
- Clean, intuitive interface
- Python 3.11 or higher
- OpenAI API key (for Code Reviewer and Document Q&A)
- Internet connection (for model downloads)
# Clone the repository
git clone <repository-url>
cd document-qa
# Create and activate conda environment
conda env create -f environment.yaml
conda activate demo# Clone the repository
git clone <repository-url>
cd document-qa
# Create virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt-
Get your OpenAI API key:
- Visit OpenAI Platform
- Create a new API key
- Copy the key for use in the applications
-
Set your API key:
export OPENAI_API_KEY="your-api-key-here"
streamlit run code-reviewer.py- Upload or paste your code
- Select the programming language
- Get instant feedback on bugs, performance, style, and security
streamlit run document-qa.py- Upload a PDF document
- Start a conversation about the document content
- Ask questions and get AI-powered answers
streamlit run content-moderator.py- Enter text to check for inappropriate content
- Get toxicity scores and moderation recommendations
- Note: First run will download the toxic-bert model (~500MB)
- Streamlit: Web application framework
- OpenAI: GPT-4 API for code review and document Q&A
- LangChain: Framework for building LLM applications
- FAISS: Vector database for document similarity search
- PyPDF2: PDF text extraction
- Transformers: Hugging Face models for content moderation
- Frontend: Streamlit web interface
- Backend: Python with async processing
- AI Models:
- OpenAI GPT-4 for code review and Q&A
- Hugging Face toxic-bert for content moderation
- Vector Database: FAISS for document embeddings
- Multi-language support (Python, JavaScript, Java, Go, Rust)
- Streaming responses for real-time feedback
- Structured JSON output with categorized feedback
- Security vulnerability detection
- Performance optimization suggestions
- PDF document processing
- Intelligent text chunking (1000 chars with 200 char overlap)
- OpenAI embeddings for semantic search
- Conversational chat interface
- Persistent session state
- Real-time toxicity detection
- Configurable blocking thresholds
- Clean, intuitive interface
- Fast inference with optimized models
Feel free to submit issues, feature requests, or pull requests to improve these applications.
This project is open source and available under the MIT License.
- OpenAI for providing the GPT-4 API
- Hugging Face for the toxic-bert model
- Streamlit for the web framework
- LangChain for the LLM orchestration framework
If you found this repository helpful, please give it a ⭐ star!